To see the other types of publications on this topic, follow the link: Applied learning.
Dissertations / Theses on the topic 'Applied learning'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'Applied learning.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
1
Mauricio, Palacio Sebastián. "Machine-Learning Applied Methods." Doctoral thesis, Universitat de Barcelona, 2020. http://hdl.handle.net/10803/669286.
Full textAbstract:
The presented discourse followed several topics where every new chapter introduced an economic prediction problem and showed how traditional approaches can be complemented with new techniques like machine learning and deep learning. These powerful tools combined with principles of economic theory is highly increasing the scope for empiricists. Chapter 3 addressed this discussion. By progressively moving from Ordinary Least Squares, Penalized Linear Regressions and Binary Trees to advanced ensemble trees. Results showed that ML algorithms significantly outperform statistical models in terms of predictive accuracy. Specifically, ML models perform 49-100\% better than unbiased methods. However, we cannot rely on parameter estimations. For example, Chapter 4 introduced a net prediction problem regarding fraudulent property claims in insurance. Despite the fact that we got extraordinary results in terms of predictive power, the complexity of the problem restricted us from getting behavioral insight. Contrarily, statistical models are easily interpretable. Coefficients give us the sign, the magnitude and the statistical significance. We can learn behavior from marginal impacts and elasticities. Chapter 5 analyzed another prediction problem in the insurance market, particularly, how the combination of self-reported data and risk categorization could improve the detection of risky potential customers in insurance markets. Results were also quite impressive in terms of prediction, but again, we did not know anything about the direction or the magnitude of the features. However, by using a Probit model, we showed the benefits of combining statistic models with ML-DL models. The Probit model let us get generalizable insights on what type of customers are likely to misreport, enhancing our results. Likewise, Chapter 2 is a clear example of how causal inference can benefit from ML and DL methods. These techniques allowed us to capture that 70 days before each auction there were abnormal behaviors in daily prices. By doing so, we could apply a solid statistical model and we could estimate precisely what the net effect of the mandated auctions in Spain was. This thesis aims at combining advantages of both methodologies, machine learning and econometrics, boosting their strengths and attenuating their weaknesses. Thus, we used ML and statistical methods side by side, exploring predictive performance and interpretability.
Several conditions can be inferred from the nature of both approaches. First, as we have observed throughout the chapters, ML and traditional econometric approaches solve fundamentally different problems. We use ML and DL techniques to predict, not in terms of traditional forecast, but making our models generalizable to unseen data. On the other hand, traditional econometrics has been focused on causal inference and parameter estimation. Therefore, ML is not replacing traditional techniques, but rather complementing them. Second, ML methods focus in out-of-sample data instead of in-sample data, while statistical models typically focus on goodness-of-fit. It is then not surprising that ML techniques consistently outperformed traditional techniques in terms of predictive accuracy. The cost is then biased estimators. Third, the tradition in economics has been to choose a unique model based on theoretical principles and to fit the full dataset on it and, in consequence, obtaining unbiased estimators and their respective confidence intervals. On the other hand, ML relies on data driven selection models, and does not consider causal inference. Instead of manually choosing the covariates, the functional form is determined by the data. This also translates to the main weakness of ML, which is the lack of inference of the underlying data-generating process. I.e. we cannot derive economically meaningful conclusions from the coefficients. Focusing on out-of-sample performance comes at the expense of the ability to infer causal effects, due to the lack of standard errors on the coefficients. Therefore, predictors are typically biased, and estimators may not be normally distributed. Thus, we can conclude that in terms of out-sample performance it is hard to compete against ML models. However, ML cannot contend with the powerful insights that the causal inference analysis gives us, which allow us not only to get the most important variables and their magnitude but also the ability to understand economic behaviors.
2
Galimberti, Jaqueson Kingeski. "Adaptive learning for applied macroeconomics." Thesis, University of Manchester, 2013. https://www.research.manchester.ac.uk/portal/en/theses/adaptive-learning-for-applied-macroeconomics(cde517d7-d552-4a53-a442-c584262c3a8f).html.
Full textAbstract:
The literature on bounded rationality and learning in macroeconomics has often used recursive algorithms to depict the evolution of agents' beliefs over time. In this thesis we assess this practice from an applied perspective, focusing on the use of such algorithms for the computation of forecasts of macroeconomic variables. Our analysis develops around three issues we find to have been previously neglected in the literature: (i) the initialization of the learning algorithms; (ii) the determination and calibration of the learning gains, which are key parameters of the algorithms' specifications; and, (iii) the choice of a representative learning mechanism. In order to approach these issues we establish an estimation framework under which we unify the two main algorithms considered in this literature, namely the least squares and the stochastic gradient algorithms. We then propose an evaluation framework that mimics the real-time process of expectation formation through learning-to-forecast exercises. To analyze the quality of the forecasts associated to the learning approach, we evaluate their forecasting accuracy and resemblance to surveys, these latter taken as proxy for agents' expectations. In spite of taking these two criteria as mutually desirable, it is not clear whether they are compatible with each other: whilst forecasting accuracy represents the goal of optimizing agents, resemblance to surveys is indicative of actual agents behavior. We carry out these exercises using real-time quarterly data on US inflation and output growth covering a broad post-WWII period of time. Our main contribution is to show that a proper assessment of the adaptive learning approach requires going beyond the previous views in the literature about these issues. For the initialization of the learning algorithms we argue that such initial estimates need to be coherent with the ongoing learning process that was already in place at the beginning of our sample of data. We find that the previous initialization methods in the literature are vulnerable to this requirement, and propose a new smoothing-based method that is not prone to this critic. Regarding the learning gains, we distinguish between two possible rationales to its determination: as a choice of agents; or, as a primitive parameter of agents learning-to-forecast behavior. Our results provide strong evidence in favor of the gain as a primitive approach, hence favoring the use of surveys data for their calibration. In the third issue, about the choice of a representative algorithm, we challenge the view that learning should be represented by only one of the above algorithms; on the basis of our two evaluation criteria, our results suggest that using a single algorithm represents a misspecification. That motivate us to propose the use of hybrid forms of the LS and SG algorithms, for which we find favorable evidence as representatives of how agents learn. Finally, our analysis concludes with an optimistic assessment on the plausibility of adaptive learning, though conditioned to an appropriate treatment of the above issues. We hope our results provide some guidance on that respect.
3
Serafini, Sara. "Machine Learning applied to OCR tasks." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2019.
Find full textAbstract:
The content of this thesis describes the work done during a six-month
internship at Datalogic, in its research laboratories in Pasadena (CA). The aim of my research was to implement and evaluate a classifier as part of an industrial OCR system for learning purposes and to see how well it could work in comparison to current best Datalogic products, since it might be simpler/faster, it might be a good alternative for implementing on an embedded system (where current Datalogic products may not be able to run fast enough).
4
Mohd, Nawi Abdullah. "Applied Drama in English Language Learning." Thesis, University of Canterbury. School of Literacies and Arts in Education, 2014. http://hdl.handle.net/10092/9584.
Full textAbstract:
This thesis is a reflective exploration of the use and impact of using drama pedagogies in the English as a Second Language (ESL)/ English as a Foreign Language (EFL) classroom. It stems from the problem of secondary school English language learning in Malaysia, where current teaching practices appear to have led to the decline of the standard of English as a second language in school leavers and university graduates (Abdul Rahman, 1997; Carol Ong Teck Lan, Anne Leong Chooi Khaun, & Singh, 2011; Hazita et al., 2010; Nalliah & Thiyagarajah, 1999). This problem resonates with my own experiences at school, as a secondary school student, an ESL teacher and, later, as a teacher trainer. Consequently, these experiences led me to explore alternative or supplementary teaching methodologies that could enhance the ESL learning experience, drawing initially from drama techniques such as those advocated by Maley and Duff (1983), Wessels (1987), and Di Pietro (1983), and later from process drama pedagogies such as those advocated by Greenwood (2005); Heathcote and Bolton (1995); Kao and O'Neill (1998), and Miller and Saxton (2004). This thesis is an account of my own exploration in adapting drama pedagogies to ESL/EFL teaching. It examines ways in which drama pedagogies might increase motivation and competency in English language learning.
The main methodology of the study is that of reflective practice (e.g. Griffiths & Tann, 1992; Zeichner & Liston, 1996). It tracks a learning journey, where I critically reflect on my learning, exploring and implementing such pedagogical approaches as well as evaluate their impact on my students’ learning. These critical reflections arise from three case studies, based on three different contexts: the first a New Zealand English for Speakers of Other Languages (ESOL) class in an intermediate school, the second a Malaysian ESL class in a rural secondary school, and the third an English proficiency class of adult learners in a language school. Data for the study were obtained through the following: research journal and reflective memo; observation and field notes; interview; social media; students’ class work; discussion with co-researchers; and through the literature of the field.
A major teaching methodology that emerges from the reflective cycles is that of staging the textbook, where the textbook section to be used for the teaching programme is distilled, and the key focuses of the language, skills, vocabulary, and themes to be learnt are identified and extracted. A layer of drama is matched with these distilled elements and then ‘staged’ on top of the textbook unit, incorporating context-setting opportunities, potential for a story, potential for tension or complication, and the target language elements.
The findings that emerge through critical reflection in the study relate to the drama methodologies that I learn and acquire, the impact of these methodologies on students, the role of culture in the application of drama methodologies, and language learning and acquisition. These findings have a number of implications. Firstly, they show how an English Language Teaching (ELT) practitioner might use drama methodologies and what their impact is on student learning. While the focus is primarily on the Malaysian context, aspects of the findings may resonate internationally. Secondly, they suggest a model of reflective practice that can be used by other ELT practitioners who are interested in using drama methodologies in their teaching. Thirdly, these findings also point towards the development of a more comprehensive syllabus for using drama pedagogies, as well as the development of reflective practice, in the teacher training programmes in Malaysia.
The use of drama pedagogies for language learning is a field that has not been researched in a Malaysian context. Therefore, this account of reflective practice offers a platform for further research and reflection in this context.
5
Drake, Adam C. "Practical Improvements in Applied Spectral Learning." BYU ScholarsArchive, 2010. https://scholarsarchive.byu.edu/etd/2546.
Full textAbstract:
Spectral learning algorithms, which learn an unknown function by learning a spectral representation of the function, have been widely used in computational learning theory to prove many interesting learnability results. These algorithms have also been successfully used in real-world applications. However, previous work has left open many questions about how to best use these methods in real-world learning scenarios. This dissertation presents several significant advances in real-world spectral learning. It presents new algorithms for finding large spectral coefficients (a key sub-problem in spectral learning) that allow spectral learning methods to be applied to much larger problems and to a wider range of problems than was possible with previous approaches. It presents an empirical comparison of new and existing spectral learning methods, showing among other things that the most common approach seems to be the least effective in typical real-world settings. It also presents a multi-spectrum learning approach in which a learner makes use of multiple representations when training. Empirical results show that a multi-spectrum learner can usually match or exceed the performance of the best single-spectrum learner. Finally, this dissertation shows how a particular application, sentiment analysis, can benefit from a spectral approach, as the standard approach to the problem is significantly improved by incorporating spectral features into the learning process.
6
Domeniconi, Federico. "Deep Learning Techniques applied to Photometric Stereo." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2020. http://amslaurea.unibo.it/20031/.
Full textAbstract:
La tesi si focalizza sullo studio dello stato dell’arte della fotometria stereo con deep learning: Self-calibrating Deep Photometric Stereo Networks. Il modello è composto è composto di due reti, la prima predice la direzione e l’intensità delle luci, la seconda predice le normali della superficie. L’obiettivo della tesi è individuare i limiti del modello e capire se possa essere modifcato per avere buone prestazioni anche in scenari reali. Il progetto di tesi è basato su fine-tuning, una tecnica supervisionata di transfer learning. Per questo scopo un nuovo dataset è stato creato acquisendo immagini in laboratorio. La ground-truth è ottenuta tramite una tecnica di distillazione. In particolare la direzione delle luci è ottenuta utilizzando due algoritmi di calibrazione delle luci e unendo i due risultati. Analogamente le normali delle superfici sono ottenute unendo i risultati di vari algoritmi di fotometria stereo. I risultati della tesi sono molto promettenti. L’errore nella predizione della direzione e dell’intensità delle luci è un terzo dell’errore del modello originale. Le predizioni delle normali delle superfici possono essere analizzate solo qualitativamente, ma i miglioramenti sono evidenti. Il lavoro di questa tesi ha mostrato che è possibile applicare transfer-learning alla fotometria stereo con deep learning. Perciò non è necessario allenare un nuovo modello da zero ma è possibile approfittare di modelli già esistenti per migliorare le prestazioni e ridurre il tempo di allenamento.
7
Weintraub, Ben Julian. "Learning control applied to a model helicopter." Thesis, Massachusetts Institute of Technology, 1994. http://hdl.handle.net/1721.1/49921.
Full text
8
Idowu, Samuel O. "Applied Machine Learning in District Heating System." Licentiate thesis, Luleå tekniska universitet, Datavetenskap, 2018. http://urn.kb.se/resolve?urn=urn:nbn:se:ltu:diva-68486.
Full textAbstract:
In an increasingly applied domain of pervasive computing, sensing devices are being deployed progressively for data acquisition from various systems through the use of technologies such as wireless sensor networks. Data obtained from such systems are used analytically to advance or improve system performance or efficiency. The possibility to acquire an enormous amount of data from any target system has made machine learning a useful approach for several large-scale analytical solutions. Machine learning has proved viable in the area of the energy sector, where the global demand for energy and the increasingly accepted need for green energy is gradually challenging energy supplies and the efficiency in its consumption. This research, carried out within the area of pervasive computing, aims to explore the application of machine learning and its effectiveness in the energy sector with dependency on sensing devices. The target application area readily falls under a multi-domain energy grid which provides a system across two energy utility grids as a combined heat and power system. The multi-domain aspect of the target system links to a district heating system network and electrical power from a combined heat and power plant. This thesis, however, focuses on the district heating system as the application area of interest while contributing towards a future goal of a multi-domain energy grid, where improved efficiency level, reduction of overall carbon dioxide footprint and enhanced interaction and synergy between the electricity and thermal grid are vital goals. This thesis explores research issues relating to the effectiveness of machine learning in forecasting heat demands at district heating substations, and the key factors affecting domestic heat load patterns in buildings. The key contribution of this thesis is the application of machine learning techniques in forecasting heat energy consumption in buildings, and our research outcome shows that supervised machine learning methods are suitable for domestic thermal load forecast. Among the examined machine learning methods which include multiple linear regression, support vector machine, feed forward neural network, and regression tree, the support vector machine performed best with a normalized root mean square error of 0.07 for a 24-hour forecast horizon. In addition, weather and time information are observed to be the most influencing factors when forecasting heat load at heating network substations. Investigation on the effect of using substation's operational attributes, such as the supply and return temperatures, as additional input parameters when forecasting heat load shows that the use of substation's internal operational attributes has less impact.
9
Martin, del Campo Barraza Sergio. "Unsupervised feature learning applied to condition monitoring." Doctoral thesis, Luleå tekniska universitet, Institutionen för system- och rymdteknik, 2017. http://urn.kb.se/resolve?urn=urn:nbn:se:ltu:diva-63113.
Full textAbstract:
Improving the reliability and efficiency of rotating machinery are central problems in many application domains, such as energy production and transportation. This requires efficient condition monitoring methods, including analytics needed to predict and detect faults and manage the high volume and velocity of data. Rolling element bearings are essential components of rotating machines, which are particularly important to monitor due to the high requirements on the operational conditions. Bearings are also located near the rotating parts of the machines and thereby the signal sources that characterize faults and abnormal operational conditions. Thus, bearings with embedded sensing, analysis and communication capabilities are developed. However, the analysis of signals from bearings and the surrounding components is a challenging problem due to the high variability and complexity of the systems. For example, machines evolve over time due to wear and maintenance, and the operational conditions typically also vary over time. Furthermore, the variety of fault signatures and failure mechanisms makes it difficult to derive generally useful and accurate models, which enable early detection of faults at reasonable cost. Therefore, investigations of machine learning methods that avoid some of these difficulties by automated on-line adaptation of the signal model are motivated. In particular, can unsupervised feature learning methods be used to automatically derive useful information about the state and operational conditions of a rotating machine? What additional methods are needed to recognize normal operational conditions and detect abnormal conditions, for example in terms of learned features or changes of model parameters? Condition monitoring systems are typically based on condition indicators that are pre-defined by experts, such as the amplitudes in certain frequency bands of a vibration signal, or the temperature of a bearing. Condition indicators are used to define alarms in terms of thresholds; when the indicator is above (or below) the threshold, an alarm indicating a fault condition is generated, without further information about the root cause of the fault. Similarly, machine learning methods and labeled datasets are used to train classifiers that can be used for the detection of faults. The accuracy and reliability of such condition monitoring methods depends on the type of condition indicators used and the data considered when determining the model parameters. Hence, this approach can be challenging to apply in the field where machines and sensor systems are different and change over time, and parameters have different meaning depending on the conditions. Adaptation of the model parameters to each condition monitoring application and operational condition is also difficult due to the need for labeled training data representing all relevant conditions, and the high cost of manual configuration. Therefore, neither of these solutions is viable in general. In this thesis I investigate unsupervised methods for feature learning and anomaly detection, which can operate online without pre-training with labeled datasets. Concepts and methods for validation of normal operational conditions and detection of abnormal operational conditions based on automatically learned features are proposed and studied. In particular, dictionary learning is applied to vibration and acoustic emission signals obtained from laboratory experiments and condition monitoring systems. The methodology is based on the assumption that signals can be described as a linear superposition of noise and learned atomic waveforms of arbitrary shape, amplitude and position. Greedy sparse coding algorithms and probabilistic gradient methods are used to learn dictionaries of atomic waveforms enabling sparse representation of the vibration and acoustic emission signals. As a result, the model can adapt automatically to different machine configurations, and environmental and operational conditions with a minimum of initial configuration. In addition, sparse coding results in reduced data rates that can simplify the processing and communication of information in resource-constrained systems. Measures that can be used to detect anomalies in a rotating machine are introduced and studied, like the dictionary distance between an online propagated dictionary and a set of dictionaries learned when the machine is known to operate in healthy conditions. In addition, the possibility to generalize a dictionary learned from the vibration signal in one machine to another similar machine is studied in the case of wind turbines. The main contributions of this thesis are the extension of unsupervised dictionary learning to condition monitoring for anomaly detection purposes, and the related case studies demonstrating that the learned features can be used to obtain information about the condition. The cases studies include vibration signals from controlled ball bearing experiments and wind turbines; and acoustic emission signals from controlled tensile strength tests and bearing contamination experiments. It is found that the dictionary distance between an online propagated dictionary and a baseline dictionary trained in healthy conditions can increase up to three times when a fault appears, without reference to kinematic information like defect frequencies. Furthermore, it is found that in the presence of a bearing defect, impulse-like waveforms with center frequencies that are about two times higher than in the healthy condition are learned. In the case of acoustic emission analysis, it is shown that the representations of signals of different strain stages of stainless steel appear as distinct clusters. Furthermore, the repetition rates of learned acoustic emission waveforms are found to be markedly different for a bearing with and without particles in the lubricant, especially at high rotational speed above 1000 rpm, where particle contaminants are difficult to detect using conventional methods. Different hyperparameters are investigated and it is found that the model is useful for anomaly detection with as little as 2.5 % preserved coefficients.
10
Wang, Shihai. "Boosting learning applied to facial expression recognition." Thesis, University of Manchester, 2009. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.511940.
Full text
11
Finnman, Peter, and Max Winberg. "Deep reinforcement learning compared with Q-table learning applied to backgammon." Thesis, KTH, Skolan för datavetenskap och kommunikation (CSC), 2016. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-186545.
Full textAbstract:
Reinforcement learning attempts to mimic how humans react to their surrounding environment by giving feedback to software agents based on the actions they take. To test the capabilities of these agents, researches have long regarded board games as a powerful tool. This thesis compares two approaches to reinforcement learning in the board game backgammon, a Q-table and a deep reinforcement network. It was determined which approach surpassed the other in terms of accuracy and convergence rate towards the perceived optimal strategy. The evaluation is performed by training the agents using the self-learning approach. After variable amounts of training sessions, the agents are benchmarked against each other and a third, random agent. The results derived from the study indicate that the convergence rate of the deep learning agent is far superior to that of the Q-table agent. However, the results also indicate that the accuracy of Q-tables is greater than that of deep learning once the former has mapped the environment.
12
Amar, Gilad. "Deep learning for supernovae detection." Master's thesis, University of Cape Town, 2017. http://hdl.handle.net/11427/27090.
Full textAbstract:
In future astronomical sky surveys it will be humanly impossible to classify the tens of thousands of candidate transients detected per night. This thesis explores the potential of using state-of-the-art machine learning algorithms to handle this burden more accurately and quickly than trained astronomers. To this end Deep Learning methods are applied to classify transients using real-world data from the Sloan Digital Sky Survey. Using cutting-edge training techniques several Convolutional Neural networks are trained and hyper-parameters tuned to outperform previous approaches and find that human labelling errors are the primary obstacle to further improvement. The tuning and optimisation of the deep models took in excess of 700 hours on a 4-Titan X GPU cluster.
13
Goldenberg, David. "Adaptive learning and cryptography." W&M ScholarWorks, 2010. https://scholarworks.wm.edu/etd/1539623564.
Full textAbstract:
Significant links exist between cryptography and computational learning theory. Cryptographic functions are the usual method of demonstrating significant intractability results in computational learning theory as they can demonstrate that certain problems are hard in a representation independent sense. On the other hand, hard learning problems have been used to create efficient cryptographic protocols such as authentication schemes, pseudo-random permutations and functions, and even public key encryption schemes.;Learning theory / coding theory also impacts cryptography in that it enables cryptographic primitives to deal with the issues of noise or bias in their inputs. Several different constructions of "fuzzy" primitives exist, a fuzzy primitive being a primitive which functions correctly even in the presence of "noisy", or non-uniform inputs. Some examples of these primitives include error-correcting blockciphers, fuzzy identity based cryptosystems, fuzzy extractors and fuzzy sketches. Error correcting blockciphers combine both encryption and error correction in a single function which results in increased efficiency. Fuzzy identity based encryption allows the decryption of any ciphertext that was encrypted under a "close enough" identity. Fuzzy extractors and sketches are methods of reliably (re)-producing a uniformly random secret key given an imperfectly reproducible string from a biased source, through a public string that is called the "sketch".;While hard learning problems have many qualities which make them useful in constructing cryptographic protocols, such as their inherent error tolerance and simple algebraic structure, it is often difficult to utilize them to construct very secure protocols due to assumptions they make on the learning algorithm. Due to these assumptions, the resulting protocols often do not have security against various types of "adaptive" adversaries. to help deal with this issue, we further examine the inter-relationships between cryptography and learning theory by introducing the concept of "adaptive learning". Adaptive learning is a rather weak form of learning in which the learner is not expected to closely approximate the concept function in its entirety, rather it is only expected to answer a query of the learner's choice about the target. Adaptive learning allows for a much weaker learner than in the standard model, while maintaining the the positive properties of many learning problems in the standard model, a fact which we feel makes problems that are hard to adaptively learn more useful than standard model learning problems in the design of cryptographic protocols. We argue that learning parity with noise is hard to do adaptively and use that assumption to construct a related key secure, efficient MAC as well as an efficient authentication scheme. In addition we examine the security properties of fuzzy sketches and extractors and demonstrate how these properties can be combined by using our related key secure MAC. We go on to demonstrate that our extractor can allow a form of related-key "hardening" for protocols in that, by affecting how the key for a primitive is stored it renders that protocol immune to related key attacks.
14
Du, Buisson Lise. "Machine learning in astronomy." Master's thesis, University of Cape Town, 2015. http://hdl.handle.net/11427/15502.
Full textAbstract:
The search to find answers to the deepest questions we have about the Universe has fueled the collection of data for ever larger volumes of our cosmos. The field of supernova cosmology, for example, is seeing continuous development with upcoming surveys set to produce a vast amount of data that will require new statistical inference and machine learning techniques for processing and analysis. Distinguishing between real objects and artefacts is one of the first steps in any transient science pipeline and, currently, is still carried out by humans - often leading to hand scanners having to sort hundreds or thousands of images per night. This is a time-consuming activity introducing human biases that are extremely hard to characterise. To succeed in the objectives of future transient surveys, the successful substitution of human hand scanners with machine learning techniques for the purpose of this artefact-transient classification therefore represents a vital frontier. In this thesis we test various machine learning algorithms and show that many of them can match the human hand scanner performance in classifying transient difference g, r and i-band imaging data from the SDSS-II SN Survey into real objects and artefacts. Using principal component analysis and linear discriminant analysis, we construct a grand total of 56 feature sets with which to train, optimise and test a Minimum Error Classifier (MEC), a naive Bayes classifier, a k-Nearest Neighbours (kNN) algorithm, a Support Vector Machine (SVM) and the SkyNet artificial neural network.
15
Chan, Cho-kui, and 陳祖鉅. "Collaborative learning on Internet: learning applied mathematics through newsgroup on the net." Thesis, The University of Hong Kong (Pokfulam, Hong Kong), 1998. http://hub.hku.hk/bib/B31959957.
Full text
16
Chan, Cho-kui. "Collaborative learning on Internet : learning applied mathematics through newsgroup on the net /." Hong Kong : University of Hong Kong, 1998. http://sunzi.lib.hku.hk/hkuto/record.jsp?B20057428.
Full text
17
Yang, Lili. "Machine learning methodologies applied to fire risk management." Thesis, University of Derby, 2004. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.407044.
Full text
18
Tummala, Akhil. "Self-learning algorithms applied in Continuous Integration system." Thesis, Blekinge Tekniska Högskola, Institutionen för datalogi och datorsystemteknik, 2018. http://urn.kb.se/resolve?urn=urn:nbn:se:bth-16675.
Full textAbstract:
Context: Continuous Integration (CI) is a software development practice where a developer integrates a code into the shared repository. And, then an automated system verifies the code and runs automated test cases to find integration error. For this research, Ericsson’s CI system is used. The tests that are performed in CI are regression tests. Based on the time scopes, the regression test suites are categorized into hourly and daily test suits. The hourly test is performed on all the commits made in a day, whereas the daily test is performed at night on the latest build that passed the hourly test. Here, the hourly and daily test suites are static, and the hourly test suite is a subset of the daily test suite. Since the daily test is performed at the end of the day, the results are obtained on the next day, which is delaying the feedback to the developers regarding the integration errors. To mitigate this problem, research is performed to find the possibility of creating a learning model and integrating into the CI system, which can then create a dynamic hourly test suite for faster feedback. Objectives: This research aims to find the suitable machine learning algorithm for CI system and investigate the feasibility of creating self-learning test machinery. This goal is achieved by examining the CI system and, finding out what type data is required for creating the learning model for prioritizing the test cases. Once the necessary data is obtained, then the selected algorithms are evaluated to find the suitable learning algorithm for creating self-learning test machinery. And then, the investigation is done whether the created learning model can be integrated into the CI workflow to create the self-learning test machinery. Methods: In this research, an experiment is conducted for evaluating the learning algorithms. For this experimentation, the data is provided by Ericsson AB, Gothenburg. The dataset consists of the daily test information and the test case results. The algorithms that are evaluated in this experiment are Naïve Bayes, Support vector machines, and Decision trees. This evaluation is done by performing leave-one-out cross-validation. And, the learning algorithm performance is calculated by using the prediction accuracy. After obtaining the accuracies, the algorithms are compared to find the suitable machine learning algorithm for CI system. Results: Based on the Experiment results it is found that support vector machines have outperformed Naïve Bayes and Decision tree algorithms in performance. But, due to the challenges present in the current CI system, the created learning model is not feasible to integrate into the CI. The primary challenge faced by the CI system is, mapping of test case failure to its respective commit is no possible (cannot find which commit made the test case to fail). This is because the daily test is performed on the latest build which is the combination of commits made in that day. Another challenge present is low data storage. Due to this low data storage, problems like the curse of dimensionality and class imbalance has occurred. Conclusions: By conducting this research, a suitable learning algorithm is identified for creating a self-learning machinery. And, also identified the challenges facing to integrate the model in CI. Based on the results obtained from the experiment, it is recognized that support vector machines have high prediction accuracy in test case result classification compared to Naïve Bayes and Decision trees.
19
Shukla, Manu. "Algorithmic Distribution of Applied Learning on Big Data." Diss., Virginia Tech, 2020. http://hdl.handle.net/10919/100603.
Full textAbstract:
Machine Learning and Graph techniques are complex and challenging to distribute. Generally, they are distributed by modeling the problem in a similar way as single node sequential techniques except applied on smaller chunks of data and compute and the results combined. These techniques focus on stitching the results from smaller chunks as the best possible way to have the outcome as close to the sequential results on entire data as possible. This approach is not feasible in numerous kernel, matrix, optimization, graph, and other techniques where the algorithm needs access to all the data during execution. In this work, we propose key-value pair based distribution techniques that are widely applicable to statistical machine learning techniques along with matrix, graph, and time series based algorithms. The crucial difference with previously proposed techniques is that all operations are modeled on key-value pair based fine or coarse-grained steps. This allows flexibility in distribution with no compounding error in each step. The distribution is applicable not only in robust disk-based frameworks but also in in-memory based systems without significant changes. Key-value pair based techniques also provide the ability to generate the same result as sequential techniques with no edge or overlap effects in structures such as graphs or matrices to resolve.
This thesis focuses on key-value pair based distribution of applied machine learning techniques on a variety of problems. For the first method key-value pair distribution is used for storytelling at scale. Storytelling connects entities (people, organizations) using their observed relationships to establish meaningful storylines. When performed sequentially these computations become a bottleneck because the massive number of entities make space and time complexity untenable. We present DISCRN, or DIstributed Spatio-temporal ConceptseaRch based StorytelliNg, a distributed framework for performing spatio-temporal storytelling. The framework extracts entities from microblogs and event data, and links these entities using a novel ConceptSearch to derive storylines in a distributed fashion utilizing key-value pair paradigm. Performing these operations at scale allows deeper and broader analysis of storylines. The novel parallelization techniques speed up the generation and filtering of storylines on massive datasets. Experiments with microblog posts such as Twitter data and GDELT(Global Database of Events, Language and Tone) events show the efficiency of the techniques in DISCRN.
The second work determines brand perception directly from people's comments in social media. Current techniques for determining brand perception, such as surveys of handpicked users by mail, in person, phone or online, are time consuming and increasingly inadequate. The proposed DERIV system distills storylines from open data representing direct consumer voice into a brand perception. The framework summarizes the perception of a brand in comparison to peer brands with in-memory key-value pair based distributed algorithms utilizing supervised machine learning techniques. Experiments performed with open data and models built with storylines of known peer brands show the technique as highly scalable and accurate in capturing brand perception from vast amounts of social data compared to sentiment analysis.
The third work performs event categorization and prospect identification in social media. The problem is challenging due to endless amount of information generated daily. In our work, we present DISTL, an event processing and prospect identifying platform. It accepts as input a set of storylines (a sequence of entities and their relationships) and processes them as follows: (1) uses different algorithms (LDA, SVM, information gain, rule sets) to identify themes from storylines;
(2) identifies top locations and times in storylines and combines with themes to generate events that are meaningful in a specific scenario for categorizing storylines; and (3) extracts top prospects as people and organizations from data elements contained in storylines. The output comprises sets of events in different categories and storylines under them along with top prospects identified. DISTL utilizes in-memory key-value pair based distributed processing that scales to high data volumes and categorizes generated storylines in near real-time.
The fourth work builds flight paths of drones in a distributed manner to survey a large area taking images to determine growth of vegetation over power lines allowing for adjustment to terrain and number of drones and their capabilities. Drones are increasingly being used to perform risky and labor intensive aerial tasks cheaply and safely. To ensure operating costs are low and flights autonomous, their flight plans must be pre-built. In existing techniques drone flight paths are not automatically pre-calculated based on drone capabilities and terrain information. We present details of an automated flight plan builder DIMPL that pre-builds flight plans for drones tasked with surveying a large area to take photographs of electric poles to identify ones with hazardous vegetation overgrowth. DIMPL employs a distributed in-memory key-value pair based paradigm to process subregions in parallel and build flight paths in a highly efficient manner.
The fifth work highlights scaling graph operations, particularly pruning and joins. Linking topics to specific experts in technical documents and finding connections between experts are crucial for detecting the evolution of emerging topics and the relationships between their influencers in state-of-the-art research. Current techniques that make such connections are limited to similarity measures. Methods based on weights such as TF-IDF and frequency to identify important topics and self joins between topics and experts are generally utilized to identify connections between experts. However, such approaches are inadequate for identifying emerging keywords and experts since the most useful terms in technical documents tend to be infrequent and concentrated in just a few documents. This makes connecting experts through joins on large dense graphs challenging. We present DIGDUG, a framework that identifies emerging topics by applying graph operations to technical terms. The framework identifies connections between authors of patents and journal papers by performing joins on connected topics and topics associated with the authors at scale. The problem of scaling the graph operations for topics and experts is solved through dense graph pruning and graph joins categorized under their own scalable separable dense graph class based on key-value pair distribution. Comparing our graph join and pruning technique against multiple graph and join methods in MapReduce revealed a significant improvement in performance using our approach.Doctor of Philosophy
Distribution of Machine Learning and Graph algorithms is commonly performed by modeling the core algorithm in the same way as the sequential technique except implemented on distributed framework. This approach is satisfactory in very few cases, such as depth-first search and subgraph enumerations in graphs, k nearest neighbors, and few additional common methods. These techniques focus on stitching the results from smaller data or compute chunks as the best possible way to have the outcome as close to the sequential results on entire data as possible. This approach is not feasible in numerous kernel, matrix, optimization, graph, and other techniques where the algorithm needs to perform exhaustive computations on all the data during execution. In this work, we propose key-value pair based distribution techniques that are exhaustive and widely applicable to statistical machine learning algorithms along with matrix, graph, and time series based operations. The crucial difference with previously proposed techniques is that all operations are modeled as key-value pair based fine or coarse-grained steps. This allows flexibility in distribution with no compounding error in each step. The distribution is applicable not only in robust disk-based frameworks but also in in-memory based systems without significant changes. Key-value pair based techniques also provide the ability to generate the same result as sequential techniques with no edge or overlap effects in structures such as graphs or matrices to resolve.
20
Lamb, Darren Hayes. "Project based learning in an applied construction curriculum." CSUSB ScholarWorks, 2003. https://scholarworks.lib.csusb.edu/etd-project/2188.
Full textAbstract:
This project addresses the integration of a career and technical (vocational) construction curriculum with academic curriculum. Career and technical (vocational) curriculum in the past has been developed to address specific content. This construction curriculum inegrates inherent academic aspects.
21
Adebonojo, Leslie G., and F. R. Jelovsek. "Learning Principles as Applied to Computer-Assisted Instruction." Digital Commons @ East Tennessee State University, 1993. https://dc.etsu.edu/etsu-works/6312.
Full text
22
Kim, Beomjoon. "Efficient imitation learning and inverse reinforcement learning with application to navigation in human environments." Thesis, McGill University, 2014. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=121555.
Full textAbstract:
A key skill for mobile robots is the ability to navigate efficiently through their environment. In the case of social or assistive robots, this involves navigating through human crowds. Typical performance criteria, such as reaching the goal using the shortest path, are not appropriate in such environments, where it is more important for the robot to move in a socially adaptive manner such as respecting comfort zones of the pedestrians. This thesis investigates the Learning from Demonstration framework to address the socially adaptive path planning problem. Learning from Demonstration is a practical framework for learning complex policies using demonstration trajectories produced by an expert. We propose two approaches based on Learning from Demonstrations. First approach is based on Inverse Reinforcement Learning, in which we compactly represent the socially adaptive path planning behaviours as a cost function that we learn. The second approach is based on imitation learning, in which we use supervised learning to learn such behaviours, and then provide theoretical guarantees on its performance. We evaluate our approach by deploying it on a real robotic wheelchair platform in various scenarios, and comparing the robot trajectories to human trajectories.Une compétence essentielle au bon fonctionnement des robots mobiles est la capacité à naviguer efficacement dans leur environnement. Ainsi, pour les robots sociaux ou d'assistance, il est essentiel de pouvoir naviguer parmi des foules humaines. Les critres de performance typiques, tels qu'atteindre un endroit ciblé par le chemin le plus court, ne sont pas appropriés dans de tels environnements, où il est plutôt important de se déplacer d'une manire socialement adaptée en respectant, par exemple, les zones de confort des piétons. Cette thèse examine un système d'apprentissage par démonstration ayant pour but de résoudre le problème de planification de trajectoire adaptée à un environment humain. L'apprentissage par démonstration est un cadre pratique permettant l'acquisition de controlleurs complexes en utilisant des trajectoires de démonstration provenant d'un expert. Nous proposons deux approches basées sur l'apprentissage par démonstration. La première approche est basée sur l'apprentissage par renforcement inverse, dans lequel nous représentons de façon compacte les comportements de planification de trajectoire socialement adaptative en fonction des cots appris. La deuxième approche est fondée sur l'apprentissage par imitation, où nous utilisons l'apprentissage supervisé pour aquérir ces comportements, et fournissons subséquemment des garanties théoriques sur sa performance. Nous évaluons notre approche en la déployant sur un véritable fauteuil roulant robotisé dans différents scénarios et la comparons à des trajectoires humaines.
23
Zhang, Bo. "Machine Learning on Statistical Manifold." Scholarship @ Claremont, 2017. http://scholarship.claremont.edu/hmc_theses/110.
Full textAbstract:
This senior thesis project explores and generalizes some fundamental machine learning algorithms from the Euclidean space to the statistical manifold, an abstract space in which each point is a probability distribution. In this thesis, we adapt the optimal separating hyperplane, the k-means clustering method, and the hierarchical clustering method for classifying and clustering probability distributions. In these modifications, we use the statistical distances as a measure of the dissimilarity between objects. We describe a situation where the clustering of probability distributions is needed and useful. We present many interesting and promising empirical clustering results, which demonstrate the statistical-distance-based clustering algorithms often outperform the same algorithms with the Euclidean distance in many complex scenarios. In particular, we apply our statistical-distance-based hierarchical and k-means clustering algorithms to the univariate normal distributions with k = 2 and k = 3 clusters, the bivariate normal distributions with diagonal covariance matrix and k = 3 clusters, and the discrete Poisson distributions with k = 3 clusters. Finally, we prove the k-means clustering algorithm applied on the discrete distributions with the Hellinger distance converges not only to the partial optimal solution but also to the local minimum.
24
Jones, Piet. "Structure learning of gene interaction networks." Thesis, Stellenbosch : Stellenbosch University, 2014. http://hdl.handle.net/10019.1/86650.
Full textAbstract:
Thesis (MSc)--Stellenbosch University, 2014.ENGLISH ABSTRACT: There is an ever increasing wealth of information that is being generated regarding biological systems, in particular information on the interactions and dependencies of genes and their regulatory process. It is thus important to be able to attach functional understanding to this wealth of information. Mathematics can potentially provide the tools needed to generate the necessary abstractions to model the complex system of gene interaction. Here the problem of uncovering gene interactions is cast in several contexts, namely uncovering gene interaction patterns using statistical dependence, cooccurrence as well as feature enrichment. Several techniques have been proposed in the past to solve these, with various levels of success. Techniques have ranged from supervised learning, clustering analysis, boolean networks to dynamical Bayesian models and complex system of di erential equations. These models attempt to navigate a high dimensional space with challenging degrees of freedom. In this work a number of approaches are applied to hypothesize a gene interaction network structure. Three di erent models are applied to real biological data to generate hypotheses on putative biological interactions. A cluster-based analysis combined with a feature enrichment detection is initially applied to a Vitis vinifera dataset, in a targetted analysis. This model bridges a disjointed set of putatively co-expressed genes based on signi cantly associated features, or experimental conditions. We then apply a cross-cluster Markov Blanket based model, on a Saccharomyces cerevisiae dataset. Here the disjointed clusters are bridged by estimating statistical dependence relationship across clusters, in an un-targetted approach. The nal model applied to the same Saccharomyces cerevisiae dataset is a non-parametric Bayesian method that detects probeset co-occurrence given a local background and inferring gene interaction based on the topological network structure resulting from gene co-occurance. In each case we gather evidence to support the biological relevance of these hypothesized interactions by investigating their relation to currently established biological knowledge. The various methods applied here appear to capture di erent aspects of gene interaction, in the datasets we applied them to. The targetted approach appears to putatively infer gene interactions based on functional similarities. The cross-cluster-analysis-based methods, appear to capture interactions within pathways. The probabilistic-co-occurrence-based method appears to generate modules of functionally related genes that are connected to potentially explain the underlying experimental dynamics.
AFRIKAANSE OPSOMMING: Daar is 'n toenemende rykdom van inligting wat gegenereer word met betrekking tot biologiese stelsels, veral inligting oor die interaksies en afhanklikheidsverhoudinge van gene asook hul regulatoriese prosesse. Dit is dus belangrik om in staat te wees om funksionele begrip te kan heg aan hierdie rykdom van inligting. Wiskunde kan moontlik die gereedskap verskaf en die nodige abstraksies bied om die komplekse sisteem van gene interaksies te modelleer. Hier is die probleem met die beraming van die interaksies tussen gene benader uit verskeie kontekste uit, soos die ontdekking van patrone in gene interaksie met behulp van statistiese afhanklikheid , mede-voorkoms asook funksie verryking. Verskeie tegnieke is in die verlede voorgestel om hierdie probleem te benader, met verskillende vlakke van sukses. Tegnieke het gewissel van toesig leer , die groepering analise, boolean netwerke, dinamiese Bayesian modelle en 'n komplekse stelsel van di erensiaalvergelykings. Hierdie modelle poog om 'n hoë dimensionele ruimte te navigeer met uitdagende grade van vryheid. In hierdie werk word 'n aantal benaderings toegepas om 'n genetiese interaksie netwerk struktuur voor te stel. Drie verskillende modelle word toegepas op werklike biologiese data met die doel om hipoteses oor vermeende biologiese interaksies te genereer. 'n Geteikende groeperings gebaseerde analise gekombineer met die opsporing van verrykte kenmerke is aanvanklik toegepas op 'n Vitis vinifera datastel. Hierdie model verbind disjunkte groepe van vermeende mede-uitgedrukte gene wat gebaseer is op beduidende verrykte kenmerke, hier eksperimentele toestande . Ons pas dan 'n tussen groepering Markov Kombers model toe, op 'n Saccharomyces cerevisiae datastel. Hier is die disjunkte groeperings ge-oorbrug deur die beraming van statistiese afhanklikheid verhoudings tussen die elemente in die afsondelike groeperings. Die nale model was ons toepas op dieselfde Saccharomyces cerevisiae datastel is 'n nie- parametriese Bayes metode wat probe stelle van mede-voorkommende gene ontdek, gegee 'n plaaslike agtergrond. Die gene interaksie is beraam op grond van die topologie van die netwerk struktuur veroorsaak deur die gesamentlike voorkoms gene. In elk van die voorgenome gevalle word ons hipotese vermoedelik ondersteun deur die beraamde gene interaksies in terme van huidige biologiese kennis na te vors. Die verskillende metodes wat hier toegepas is, modelleer verskillende aspekte van die interaksies tussen gene met betrekking tot die datastelle wat ons ondersoek het. In die geteikende benadering blyk dit asof ons vermeemde interaksies beraam gebaseer op die ooreenkoms van biologiese funksies. Waar die a eide gene interaksies moontlik gebaseer kan wees op funksionele ooreenkomste tussen die verskeie gene. In die analise gebaseer op die tussen modelering van gene groepe, blyk dit asof die verhouding van gene in bekende biologiese substelsels gemodelleer word. Dit blyk of die model gebaseer op die gesamentlike voorkoms van gene die verband tussen groepe van funksionele verbonde gene modelleer om die onderliggende dinamiese eienskappe van die experiment te verduidelik.
25
Roots, Lindsey. "Learning to trust : experimental evidence of social learning in a real-world social network of player A's in a trust game." Master's thesis, University of Cape Town, 2013. http://hdl.handle.net/11427/5704.
Full text
26
Abou-Moustafa, Karim. "Metric learning revisited: new approaches for supervised and unsupervised metric learning with analysis and algorithms." Thesis, McGill University, 2012. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=106370.
Full textAbstract:
In machine learning one is usually given a data set of real high dimensional vectors X, based on which it is desired to select a hypothesis θ from the space of hypotheses Θ using a learning algorithm. An immediate assumption that is usually imposed on X is that it is a subset from the very general embedding space Rp which makes the Euclidean distance ∥•∥2 to become the default metric for the elements of X. Since various learning algorithms assume that the input space is Rp with its endowed metric ∥•∥2 as a (dis)similarity measure, it follows that selecting hypothesis θ becomes intrinsically tied to the Euclidean distance. Metric learning is the problem of selecting a specific metric dX from a certain family of metrics D based on the properties of the elements in the set X. Under some performance measure, the metric dX is expected to perform better on X than any other metric d 2 D. If the learning algorithm replaces the very general metric ∥•∥2 with the metric dX , then selecting hypothesis θ will be tied to the more specific metric dX which carries all the information on the properties of the elements in X. In this thesis I propose two algorithms for learning the metric dX ; the first for supervised learning settings, and the second for unsupervised, as well as for supervised and semi-supervised settings. In particular, I propose algorithms that take into consideration the structure and geometry of X on one hand, and the characteristics of real world data sets on the other. However, if we are also seeking dimensionality reduction, then under some mild assumptions on the topology of X, and based on the available a priori information, one can learn an embedding for X into a low dimensional Euclidean space Rp0, p0 << p, where the Euclidean distance better reveals the similarities between the elements of X and their groupings (clusters). That is, as a by-product, we obtain dimensionality reduction together with metric learning. In the supervised setting, I propose PARDA, or Pareto discriminant analysis for discriminative linear dimensionality reduction. PARDA is based on the machinery of multi-objective optimization; simultaneously optimizing multiple, possibly conflicting, objective functions. This allows PARDA to adapt to the class topology in the lower dimensional space, and naturally handles the class masking problem that is inherent in Fisher's discriminant analysis framework for multiclass problems. As a result, PARDA yields significantly better classification results when compared with modern techniques for discriminative dimensionality reduction. In the unsupervised setting, I propose an algorithmic framework, denoted by ?? (note the different notation), that encapsulates spectral manifold learning algorithms and gears them for metric learning. The framework ?? captures the local structure and the local density information from each point in a data set, and hence it carries all the information on the varying sample density in the input space. The structure of ?? induces two distance metrics for its elements, the Bhattacharyya-Riemann metric dBR and the Jeffreys-Riemann metric dJR. Both metrics reorganize the proximity between the points in X based on the local structure and density around each point. As a result, when combining the metric space (??, dBR) or (??, dJR) with spectral clustering and Euclidean embedding, they yield significant improvements in clustering accuracies and error rates for a large variety of clustering and classification tasks.Dans cette thèse, je propose deux algorithmes pour l'apprentissage de la métrique dX; le premier pour l'apprentissage supervisé, et le deuxième pour l'apprentissage non-supervisé, ainsi que pour l'apprentissage supervisé et semi-supervisé. En particulier, je propose des algorithmes qui prennent en considération la structure et la géométrie de X d'une part, et les caractéristiques des ensembles de données du monde réel d'autre part. Cependant, si on cherche également la réduction de dimension, donc sous certaines hypothèses légères sur la topologie de X, et en même temps basé sur des informations disponibles a priori, on peut apprendre une intégration de X dans un espace Euclidien de petite dimension Rp0 p0 << p, où la distance Euclidienne révèle mieux les ressemblances entre les éléments de X et leurs groupements (clusters). Alors, comme un sous-produit, on obtient simultanément une réduction de dimension et un apprentissage métrique. Pour l'apprentissage supervisé, je propose PARDA, ou Pareto discriminant analysis, pour la discriminante réduction linéaire de dimension. PARDA est basé sur le mécanisme d'optimisation à multi-objectifs; optimisant simultanément plusieurs fonctions objectives, éventuellement des fonctions contradictoires. Cela permet à PARDA de s'adapter à la topologie de classe dans un espace dimensionnel plus petit, et naturellement gère le problème de masquage de classe associé au discriminant Fisher dans le cadre d'analyse de problèmes à multi-classes. En conséquence, PARDA permet des meilleurs résultats de classification par rapport aux techniques modernes de réduction discriminante de dimension. Pour l'apprentissage non-supervisés, je propose un cadre algorithmique, noté par ??, qui encapsule les algorithmes spectraux d'apprentissage formant an algorithme d'apprentissage de métrique. Le cadre ?? capture la structure locale et la densité locale d'information de chaque point dans un ensemble de données, et donc il porte toutes les informations sur la densité d'échantillon différente dans l'espace d'entrée. La structure de ?? induit deux métriques de distance pour ses éléments: la métrique Bhattacharyya-Riemann dBR et la métrique Jeffreys-Riemann dJR. Les deux mesures réorganisent la proximité entre les points de X basé sur la structure locale et la densité autour de chaque point. En conséquence, lorsqu'on combine l'espace métrique (??, dBR) ou (??, dJR) avec les algorithmes de "spectral clustering" et "Euclidean embedding", ils donnent des améliorations significatives dans les précisions de regroupement et les taux d'erreur pour une grande variété de tâches de clustering et de classification.
27
Meyer, Imke. "Instrumental conditioning and learning in Poroderma Pantherinum." Master's thesis, University of Cape Town, 2017. http://hdl.handle.net/11427/24515.
Full textAbstract:
Recent research has shown that higher cognitive functions and learning occur in teleosts and elasmobranchs. Very little is known about the cognitive abilities of benthic sharks and no research has been published on the learning ability of the endemic Leopard catshark (Poroderma pantherinum) species of South Africa. This species is listed as data deficient on the IUCN red list and known threats due to anthropogenic impacts include fatalities because of bycatch and depredation in the small-scale commercial fishing industry. It has been suggested that sharks can be attracted to fishing boats through the sound of outboard motors and an association can be formed between the sound and easy prey on the hooks of fisherman. This pilot study examined the learning capacity of Leopard catsharks in Hermanus, South Africa, by using a series of food-reward tests based on instrumental conditioning. A target with black and white stripes was used as the discriminative stimulus, while an auditory cue acted as a bridging stimulus for the food-reward task. Sharks were collected by hand whilst diving and shore-angling in Hermanus from August 2015 to November 2015 and acclimatised before the onset of experiments. Four juvenile sharks were trained through operant conditioning using visual and auditory stimuli for ten days each, consisting of six trials per day. A fifth juvenile shark, acting as the control, was trained through the same method without the auditory stimulus to test the influence thereof on the learning rates of Leopard catsharks. This study showed that Leopard catsharks have the ability to associate a visual stimulus with a food reward through the aid of an auditory cue. The auditory cue was also shown to increase learning rates significantly as an association was formed between the presence of food in front of the target and the auditory signal. The individual sharks in this study displayed differing levels of stress and learning rates. It is suggested that even though Leopard catsharks show high diversity in learning rates and adaptation to stress, they possess the ability to learn and adapt rapidly to changing environments. The results possibly have important implications for the understanding of learning and conditioning in Leopard catsharks and the likely anthropogenic threats caused through learned behaviour in benthic sharks.
28
Blake, Damien, and mikewood@deakin edu au. "From risk to relationship: Redefining pedagogy through applied learning reform." Deakin University. School of Social and Cultural Studies in Education, 2004. http://tux.lib.deakin.edu.au./adt-VDU/public/adt-VDU20060517.150434.
Full textAbstract:
The Victorian Certificate of Applied Learning (VCAL) emerged to provide more relevant curriculum programs that would cater for increasing retention rates of post-compulsory students. It is also an example of the new learning arising from contemporary debates and reforms that highlight inadequacies of the more traditional modes of learning.
This thesis focuses on the pedagogical and sociological issues emerging from the VCAL being introduced as an alternative learning pathways for at-risk students within a traditional secondary school culture. Through the eyes of an insider-researcher, the thesis argues for a deeper understanding of applied learning as a re-engaging pedagogy by studying the schooling experience of VCAL students and teachers.
The thesis concludes that traditional academic modes of teaching contribute to the social construction of at-risk students and argues that secondary school pedagogy needs to be redefined as a cultural phenomenon requiring teachers to be reflexively aware of their role in bridging the gap between students life experiences and the curriculum.
29
Faulkner, Ryan. "Dyna learning with deep belief networks." Thesis, McGill University, 2011. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=97177.
Full textAbstract:
The objective of reinforcement learning is to find "good" actions in an environment where feedback is provided through a numerical reward, and the current state (i.e. sensory input) is assumed to be available at each time step. The notion of "good" is defined as maximizing the expected cumulative returns over time. Sometimes it is useful to construct models of the environment to aid in solving the problem. We investigate Dyna-style reinforcement learning, a powerful approach for problems where not much real data is available. The main idea is to supplement real trajectories with simulated ones sampled from a learned model of the environment. However, in large state spaces, the problem of learning a good generative model of the environment has been open so far. We propose to use deep belief networks to learn an environment model. Deep belief networks (Hinton, 2006) are generative models that have been effective in learning the time dependency relationships among complex data. It has been shown that such models can be learned in a reasonable amount of time when they are built using energy models. We present our algorithm for using deep belief networks as a generative model for simulating the environment within the Dyna architecture, along with very promising empirical results.L'objectif de l'apprentissage par renforcement est de choisir de bonnes actions dansun environnement où les informations sont fournies par une récompense numérique, etl'état actuel (données sensorielles) est supposé être disponible à chaque pas de temps. Lanotion de "correct" est définie comme étant la maximisation des rendements attendus cumulatifsdans le temps. Il est parfois utile de construire des modèles de l'environnementpour aider à résoudre le problème. Nous étudions l'apprentissage par renforcement destyleDyna, une approche performante dans les situations où les données réelles disponiblesne sont pas nombreuses. L'idée principale est de compléter les trajectoires réelles aveccelles simulées échantillonnées partir d'un modèle appri de l'environnement. Toutefois,dans les domaines à plusieurs états, le problème de l'apprentissage d'un bon modèlegénératif de l'environnement est jusqu'à présent resté ouvert. Nous proposons d'utiliserles réseaux profonds de croyance pour apprendre un modèle de l'environnement. Lesréseaux de croyance profonds (Hinton, 2006) sont des modèles génératifs qui sont efficaces pourl'apprentissage des relations de dépendance temporelle parmi des données complexes. Ila été démontré que de tels modèles peuvent être appris dans un laps de temps raisonnablequand ils sont construits en utilisant des modèles de l'énergie. Nous présentons notre algorithmepour l'utilisation des réseaux de croyance profonds en tant que modèle génératifpour simuler l'environnement dans l'architecture Dyna, ainsi que des résultats empiriquesprometteurs.
30
Agrawal, Punit. "Program navigation analysis using machine learning." Thesis, McGill University, 2009. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=32599.
Full textAbstract:
Developers invest a large portion of their development time exploring program source code to find task-related code elements and to understand the context of their task. The task context is usually not recorded at the end of the task and is forgotten over time. Similarly, it is not possible to share the task context with other developers working on related tasks. Proposed solutions to automatically record the summary of the code investigation suffer from methodological limitations related to the techniques and the data sources used to generate the summary as well as the granularity at which it is generated. To overcome these limitations, we investigate the use of machine learning techniques, in particular decision tree learning, to predict automatically the task context from session navigation transcripts obtained from developers performing tasks on the source code. We conducted a user study to collect navigation transcripts from developers engaged in source code exploration tasks. We used the data from the user study to train and test decision tree classifiers. We compared the decision tree algorithm with two existing approaches, and found that it compares positively in most cases. Additionally, we developed an Eclipse plug-in that generates automatically a developer session summary using the decision tree classifier learned from the data collected during the user study. We provide qualitative analysis of the effectiveness of this plug-in.Les d\'eveloppeurs de logiciels investissent une grande partie de leur temps \`a explorer le code source pour trouver des \'el\'ements du code reli\'es \`a leurs t\^aches, et aussi pour mieux comprendre le contexte de leur t\^ache. Le contexte de leur t\^ache n'est g\'en\'eralement pas enregistr\'ee \`a la fin de leur s\'eance d'exploration de code et est oubli\'e au fil du temps. De m\^eme, il n'est pas possible de partager le contexte de leur t\^ache avec d'autres d\'eveloppeurs travaillant sur des t\^aches reli\'ees. Les solutions propos\'ees pour enregistrer automatiquement le r\'esum\'e de leur exploration du code souffrent de limitations m\'ethodologiques li\'ees aux techniques et aux sources de donn\'ees utilis\'ees pour g\'en\'erer le r\'esum\'e, ainsi qu'\`a la granularit\'e \`a laquelle il est g\'en\'er\'e. Pour surmonter ces limitations, nous \'etudions l'emploi de techniques d'apprentissage machine, en particulier l'arbre de d\'ecision d'apprentissage, pour pr\'evoir automatiquement le contexte de la t\^ache \`a partir des transcriptes de navigation d'une session d'exploration de code du d\'eveloppeur. Nous avons effectu\'e une \'etude de cas afin de recueillir des transcriptions de navigation g\'en\'er\'es par des d\'eveloppeurs lors de l'exploration du code source. Nous avons utilis\'e les donn\'ees de cette \'etude pour tester les classifications de l'arbre de d\'ecision. Nous avons compar\'e l'algorithme \`a arbre \`a d\'ecision avec deux approches existantes, et avons d\'emontr\'e que cette nouvelle approche se compare favorablement dans la plupart des cas. Additionnellement, nous avons d\'evelopp\'e un plug-in Eclipse qui g\'en\`ere automatiquement un
31
Bordianu, Gheorghita. "Learning influence probabilities in social networks." Thesis, McGill University, 2013. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=114597.
Full textAbstract:
Social network analysis is an important cross-disciplinary area of research, with applications in fields such as biology, epidemiology, marketing and even politics. Influence maximization is the problem of finding the set of seed nodes in an information diffusion process that guarantees maximum spread of influence in a social network, given its structure. Most approaches to this problem make two assumptions. First, the global structure of the network is known. Second, influence probabilities between any two nodes are known beforehand, which is rarely the case in practical settings. In this thesis we propose a different approach to the problem of learning those influence probabilities from past data, using only the local structure of the social network. The method is grounded in unsupervised machine learning techniques and is based on a form of hierarchical clustering, allowing us to distinguish between influential and the influenceable nodes. Finally, we provide empirical results using real data extracted from Facebook.L'analyse des réseaux sociaux est un domaine d'études interdisciplinaires qui comprend des applications en biologie, épidémiologie, marketing et même politique. La maximisation de l'influence représente un problème où l'on doit trouver l'ensemble des noeuds de semence dans un processus de diffusion de l'information qui en même temps garantit le maximum de propagation de son influence dans un réseau social avec une structure connue. La plupart des approches à ce genre de problème font appel à deux hypothèses. Premièrement, la structure générale du réseau social est connue. Deuxièmement, les probabilités des influences entre deux noeuds sont connues à l'avance, fait qui n'est d'ailleurs pas valide dans des circonstances pratiques. Dans cette thèse, on propose un procédé différent visant la problème de l'apprentissage de ces probabilités d'influence à partir des données passées, en utilisant seulement la structure locale du réseau social. Le procédé se base sur l'apprentissage automatique sans surveillance et il est relié à une forme de regroupement hiérarchique, ce qui nous permet de faire la distinction entre les noeuds influenceurs et les noeuds influencés. Finalement, on fournit des résultats empiriques en utilisant des données réelles extraites du réseau social Facebook.
32
Skubch, Hendrik. "Hierarchical strategy learning for FLUX agents : an applied technique /." Saarbrücken : VDM Verlag Dr. Müller, 2007. http://deposit.d-nb.de/cgi-bin/dokserv?id=3057454&prov=M&dok_var=1&dok_ext=htm.
Full text
33
Pagés, Sara Elizabeth. "Applied Securities Analysis: A Look Inside the Learning Process." Thesis, The University of Arizona, 2011. http://hdl.handle.net/10150/144933.
Full text
34
Andersson, Carl. "Deep learning applied to system identification : A probabilistic approach." Licentiate thesis, Uppsala universitet, Avdelningen för systemteknik, 2019. http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-397563.
Full textAbstract:
Machine learning has been applied to sequential data for a long time in the field of system identification. As deep learning grew under the late 00's machine learning was again applied to sequential data but from a new angle, not utilizing much of the knowledge from system identification. Likewise, the field of system identification has yet to adopt many of the recent advancements in deep learning. This thesis is a response to that. It introduces the field of deep learning in a probabilistic machine learning setting for problems known from system identification. Our goal for sequential modeling within the scope of this thesis is to obtain a model with good predictive and/or generative capabilities. The motivation behind this is that such a model can then be used in other areas, such as control or reinforcement learning. The model could also be used as a stepping stone for machine learning problems or for pure recreational purposes. Paper I and Paper II focus on how to apply deep learning to common system identification problems. Paper I introduces a novel way of regularizing the impulse response estimator for a system. In contrast to previous methods using Gaussian processes for this regularization we propose to parameterize the regularization with a neural network and train this using a large dataset. Paper II introduces deep learning and many of its core concepts for a system identification audience. In the paper we also evaluate several contemporary deep learning models on standard system identification benchmarks. Paper III is the odd fish in the collection in that it focuses on the mathematical formulation and evaluation of calibration in classification especially for deep neural network. The paper proposes a new formalized notation for calibration and some novel ideas for evaluation of calibration. It also provides some experimental results on calibration evaluation.
35
Allberg, Petrus. "Applied machine learning in the logistics sector : A comparative analysis of supervised learning algorithms." Thesis, Blekinge Tekniska Högskola, Institutionen för programvaruteknik, 2018. http://urn.kb.se/resolve?urn=urn:nbn:se:bth-16656.
Full textAbstract:
BackgroundMachine learning is an area that is being explored with great haste these days, which inspired this study to investigate how seven different supervised learning algorithms perform compared to each other. These algorithms were used to perform classification tasks on logistics consignments, the classification is binary and a consignment can either be classified as missed or not. ObjectivesThe goal was to find which of these algorithms perform well when used for this classification task and to see how the results varied with different sized datasets. Importance of the features which were included in the datasets has been analyzed with the intention of finding if there is any connection between human errors and these missed consignments. MethodsThe process from raw data to a predicted classification has many steps including data gathering, data preparation, feature investigation and more. Through cross-validation, the algorithms were all trained and tested upon the same datasets and then evaluated based on the metrics recall and accuracy. ResultsThe scores on both metrics increase with the size of the datasets, and when comparing the seven algorithms, two does not perform equally compared to the other five, which all perform moderately the same. Conclusions Any of the five algorithms mentioned prior can be chosen for this type of classification, or to further study based on other measurements, and there is an indication that human errors could play a part on whether a consignment gets classified as missed or not.
36
Duala-Ekoko, Ekwa. "Using structural relationships to facilitate API learning." Thesis, McGill University, 2012. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=107667.
Full textAbstract:
Application Programming Interfaces (APIs) allow software developers to reuse code libraries, frameworks, or services without the need of having to implement relevant functionalities from scratch. The benefits of reusing source code or services through APIs have encouraged the adoption of APIs as the building blocks of modern-day software systems. However, leveraging the benefits of APIs require a developer to frequently learn how to use unfamiliar APIs --- a process made difficult by the increasing size of APIs, and the increase in the number of APIs with which a developer has to work. In this dissertation, we investigated some of the challenges developers encounter when working with unfamiliar APIs, and we designed and implemented new programming tools to assist developers in learning how to use new APIs. To investigate the difficulties developers encounter when learning to use APIs, we conducted a programming study in which twenty participants completed two programming tasks using real-world APIs. Through a systematic analysis of the screen captured videos and the verbalizations of the participants, we isolated twenty different types of questions the programmers asked when learning to use APIs, and identified five of the twenty questions as the most difficult for the programmers to answer in the context of our study. Drawing from varied sources of evidence, such as the verbalizations and the navigation paths of the participants, we explain why the participants found certain questions hard to answer, and provide new insights to the cause of the difficulties. To facilitate the API learning process, we designed and evaluated two novel programming tools: API Explorer and Introspector. The API Explorer tool addresses the difficulty a developer faces when the API types or methods necessary to implement a task are not accessible from the type the developer is working with. API Explorer leverages the structural relationships between API elements to recommend relevant methods on other objects, and to identify API types relevant to the use of a method or class. The Introspector tool addresses the difficulty of formulating effective queries when searching for code examples relevant to implementing a task. Introspector combines the structural relationships between API types to recommend types that should be used together with a seed to search for code examples for a given task. Using the types recommended by Introspector as search query, a developer can search for code examples across two code repositories, and in return, will get a list of code examples ranked based on their relevance to the search query. We evaluated API Explorer through a programming study, and evaluated Introspector quantitatively using ten tasks from six different APIs. The results of the evaluations suggest that these programming tools provide effective support to programmers learning how to use APIs.Les interfaces de programmation (API) permettent aux développeurs de réutiliser du code, des bibliothèques, des cadres d'application ou des services sans avoir à réimplémenter des fonctionnalités importantes à partir de zéro. Les avantages de la réutilisation de code source ou de services par des APIs ont encouragé l'adoption des APIs comme composant essentiel des logiciels modernes. Cependant, pour tirer parti des avantages des APIs, les développeurs doivent fréquemment apprendre à utiliser des APIs inconnus, un processus rendu difficile par la taille grandissante des APIs et par l'augmentation du nombre d'APIs avec lesquels les développeurs doivent travailler. Dans cette dissertation, nous avons étudié les défis que les développeurs rencontrent quand ils travaillent avec des APIs inconnus et nous avons conçu et implémenté de nouveaux outils de programmation pour aider les développeurs à apprendre comment utiliser ces APIs. Pour étudier les difficultés que les développeurs rencontrent lorsqu'ils apprennent à utiliser les APIs, nous avons conduit une étude dans laquelle 20 participants ont complété deux exercices de programmation enutilisant des APIs populaires. Par une analyse détaillée des bandes vidéo enregistrées lors des exercices et des commentaires émis par les participants, nous avons isolé vingt différent types de questions que les programmeurs ont posées lorsqu'ils apprenaient à utiliser les APIs. Nous avons aussi identifié cinq questions sur les 20 comme étant les plus difficiles à répondre par les programmeurs dans le contexte denotre étude. Notre analyse fournit des éléments probants qui expliquent la cause des difficultés observées.Pour faciliter l'apprentissage des APIs, nous avons conçu et évalué deux outils de programmation: API Explorer et Introspector. API Explorer est un outil qui a pour but de diminuer la difficulté que les développeursrencontrent quand les types ou les méthodes nécessaire à l'accomplissement d'une tâche dans une API ne sont pas accessibles à partir du type avec lequel le développeur travaille. API Explorer tire parti des relations structurelles entre les éléments d'une API pour recommander des méthodes pertinentes sur d'autres objets et pour identifier les types d'une API pertinents pour l'utilisation d'une méthode ou d'une classe. Introspector est un outil qui a pour but de réduire la difficulté à formuler des requêtes efficaces lorsque les développeurs cherchent des exemples de code reliés à l'accomplissement d'une tâche de programmation. Introspector combine les relations structurelles entre les types d'une API pour recommander les types qui devraient être utilisés ensemble avec un germe pour chercher des exemples de code pour une tâche particulière. Un développeur peut ainsi chercher des exemples de code dans deux référentiels en utilisant les types recommandés par Introspector. En retour, l'utilisateur recevra une liste d'exemples de code triée en fonction de leur pertinence avec leur tâche courante. Nous avons évalué API Explorer grâce à une étude avec des utilisateurs et nous avons évalué quantitativement Introspector en analysant les résultats de dix tâches effectuées avec six APIs différents. Les résultats de notre évaluation suggèrent que les outils de programmation que nous proposons offrent un support efficace pour les programmeurs désirant apprendre à utiliser une API.
37
Dagenais, Barthélémy. "Analysis and recommendations for developer learning resources." Thesis, McGill University, 2012. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=110512.
Full textAbstract:
Developer documentation helps developers learn frameworks and libraries, yet developing and maintaining accurate documentation require considerable effort and resources. Contributors who work on developer documentation need to at least take into account the project's code and the support needs of users. Although related, the documentation, the code, and the support needs evolve and are not always synchronized: for example, new features in the code are not always documented and questions repeatedly asked by users on support channels such as mailing lists may not be addressed by the documentation. Our thesis is that by studying how the relationships between documentation, code, and users' support needs are created and maintained, we can identify documentation improvements and automatically recommend some of these improvements to contributors. In this dissertation, we (1) studied the perspective of documentation contributors by interviewing open source contributors and users, (2) developed a technique that automatically generates the model of documentation, code, and users' support needs, (3) devised a technique that recovers fine-grained traceability links between the learning resources and the code, (4) investigated strategies to infer high-level documentation structures based on the traceability links, and (5) devised a recommendation system that uses the traceability links and the high-level documentation structures to suggest adaptive changes to the documentation when the underlying code evolves.La documentation pour les développeurs aide ces derniers à apprendre à utiliser des bibliothèques de fonctions et des cadres d'applications. Pourtant, créer et maintenir cette documentation requiert des efforts et des ressources considérables. Les contributeurs qui travaillent sur la documentation pour les développeurs doivent tenir compte de l'évolution du code et des besoins potentiels des utilisateurs de la documentation. Même s'ils sont reliés, la documentation, le code et les besoins des utilisateurs ne sont pas toujours synchronisés: par exemple, les nouvelles fonctionnalités ajoutées au code ne sont pas toujours documentées et la documentation n'apporte pas nécessairement de réponse aux questions posées à répétition sur des forums de discussion. Notre thèse est qu'en étudiant comment les relations entre la documentation, le code, et les besoins des utilisateurs sont crées et maintenues, nous pouvons identifier des possibilités d'améliorations à la documentation et automatiquement recommander certaines de ces améliorations aux contributeurs de documentation. Dans cette dissertation, nous avons (1) étudié la perspective des contributeurs de documentation en interviewant des contributeurs de projets en code source libre, (2) développé une technique qui génère automatique un modèle de la documentation, du code, et des questions des utilisateurs, (3) développé une technique qui recouvre les liens de traçabilité entre les ressources d'apprentissage et le code, (4) examiné des stratégies pour inférer des structures abstraites de documentation à partir des liens de traçabilité et (5) développé un système de recommandation qui utilise les liens de traçabilités et les structures abstraites de documentation pour suggérer des changements adaptatifs quand le code sous-jacent évolue.
38
Giguère, Philippe. "Unsupervised learning for mobile robot terrain classification." Thesis, McGill University, 2010. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=95062.
Full textAbstract:
In this thesis, we consider the problem of having a mobile robot autonomously learn to perceive differences between terrains. The targeted application is for terrain identification. Robust terrain identification can be used to enhance the capabilities of mobile systems, both in terms of locomotion and navigation. For example, a legged amphibious robot that has learned to differentiate sand from water can automatically select its gait on a beach: walking for sand, and swimming for water. The same terrain information can also be used to guide a robot in order to avoid specific terrain types. The problem of autonomous terrain identification is decomposed into two sub-problems: a sensing sub-problem, and a learning sub-problem. In the sensing sub-problem, we look at extracting terrain information from existing sensors, and at the design of a new tactile probe. In particular, we show that inertial sensor measurements and actuator feedback information can be combined to enable terrain identification for a legged robot. In addition, we describe a novel tactile probe designed for improved terrain sensing. In the learning sub-problem, we discuss how temporal or spatial continuities can be exploited to perform the clustering of both time-series and images. Specifically, we present a new algorithm that can be used to train a number of classifiers in order to perform clustering when temporal or spatial dependencies between samples are present. We combine our sensing approach with this clustering technique, to obtain a computational architecture that can learn autonomously to differentiate terrains. This approach is validated experimentally using several different sensing modalities (proprioceptive and tactile) and with two different robotic platforms (on a legged robot named AQUA and a wheeled robot iRobot Create). Finally, we show that the same clustering technique, when combined with image information, can be used to define a new image segmentation algorithm.Au travers de cette thèse, nous examinons la problématique entourant la perception des différences entre divers terrains, pour un robot mobile autonome. L'application visée par les résultats de nos recherches est l'identification des types de terrains. Cette identification, faite de manière robuste, permet d'augmenter les capacités de systèmes mobiles, tant au niveau de la locomotion que de la navigation. Par exemple, un robot amphibie à pattes qui aurait apprit à distinguer le sable et la mer pourra choisir de lui-même la démarche appropriée : marcher sur le sable, et nager dans l'eau. Cette même information sur le type de terrain peut aussi être utile pour guider un robot, lui permettant d'éviter des types de terrains spécifiques. Nous abordons la problématique d'identification des terrains autour de deux axes principaux: un problème de capture d'information (sensoriel), et un problème d'apprentissage. Dans le problème de la capture d'information, la question traitée est celle d'extraire l'information pertinente à l'identification du type de sol à partir de capteurs sur un robot, ou à l'aide d'une sonde tactile. En particulier, nous démontrons qu'en combinant l'information provenant d'une centrale inertielle avec celle provenant des actionneurs d'un robot à pattes, il est possible d'identifier certains types de sols. De plus, nous présentons une nouvelle sonde tactile possédant des caractéristiques améliorant la capture d'informations relatives aux terrains. Pour le problème de l'apprentissage, nous analysons comment il est possible d'exploiter les continuités spatiales et temporelles afin de séparer des séries temporelles ou des images en leurs classes constituantes (clustering). Nous présentons un nouvel algorithme de clustering basé sur ce principe. En combinant l'approche sensorielle et ce nouvel algorithme, nous obtenons une architecture permettant l'apprentissage, de façon autonome, des terrains. Cette approche est
39
Dinculescu, Monica. "Learning approximate representations of partially observable systems." Thesis, McGill University, 2010. http://digitool.Library.McGill.CA:8881/R/?func=dbin-jump-full&object_id=92257.
Full text
40
Zhang, Yue. "Sparsity in Image Processing and Machine Learning: Modeling, Computation and Theory." Case Western Reserve University School of Graduate Studies / OhioLINK, 2018. http://rave.ohiolink.edu/etdc/view?acc_num=case1523017795312546.
Full text
41
Heine, Matthew Alan. "A constrained optimization model for partitioning students into cooperative learning groups." Thesis, Colorado State University, 2016. http://pqdtopen.proquest.com/#viewpdf?dispub=10138918.
Full textAbstract:
The problem of the constrained partitioning of a set using quantitative relationships amongst the elements is considered. An approach based on constrained integer programming is proposed that permits a group objective function to be optimized subject to group quality constraints. A motivation for this problem is the partitioning of students, e.g., in middle school, into groups that target educational objectives. The method is compared to another grouping algorithm in the literature on a data set collected in the Poudre School District.
42
Easdown, Alison. "Improving numerical properties of learning algorithms for neural networks." Thesis, University of Brighton, 1997. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.363854.
Full text
43
Modarres, Najafabadi Sayed Reza. "Prediction of stock market indices using machine learning." Thesis, McGill University, 2009. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=40795.
Full textAbstract:
Financial time series prediction is a very important economical problem but the data available is very noisy. In this thesis, we explain the use of statistical and machine learning methods for stock market prediction and we evaluate the performance of these methods on data from the S&P/TSX 60 stock index. We use both linear regression and support vector regression, a state-of-art machine learning method, which is usually robust to noise. The results are mixed, illustrating the difficulty of the problem. We discuss the utility of using different types of data pre-processing for this task as well.La prediction des series de donnees economiques est un probleme tres important, mais les donnees disponiblessont tres aleatoires. Dans cette these, nous expliquons l'utilisation des statistiques et des methodes d'apprentissage automatique en vue de prevoir la valuer prochaine du S&P/TSX60. Nous utilisons deux methodes: la regression lineaire et les machines a vecteur de support pour la regression, une methode d'apprentissagemoderne, qui est tres robuste. Les resultats sont mitiges, illustrant la difficulte du probleme. Nous discutons l'utilite des differents types de donnees et le pre-traitement necessaire pour cette tache.}
44
Wang, Jun. "Optimizing the time warp protocol with learning techniques." Thesis, McGill University, 2010. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=92203.
Full textAbstract:
The history of the Time Warp protocol, one of the most commonly used synchronization protocols in parallel and distributed simulation, has been a history of optimizations. Usually the optimization problems are solved by creating an analytical model for the simulation system through careful analysis of the behavior of Time Warp. The model is expressed as a closed-form function that maps system state variables to a control parameter. At run-time, the system state variables are measured and then utilized to derive the value for the control parameter. This approach makes the quality of the optimization heavily dependent on how closely the model actually reflects the simulation reality. Because of the simplifications that are necessary to make in the course of creating such models, it is not certain the control strategies are optimal. Furthermore, models based on a specific application cannot be readily adopted for other applications.In this thesis, as an alternative approach, we present a number of model-free direct-search algorithms based on techniques from system control, machine learning, and evolutionary computing, namely, learning automata, reinforcement learning, and genetic algorithms. What these methods have in common is the notion of learning. Unlike the traditional methods used in Time Warp optimization, these learning methods treat the Time Warp simulator as a black box. They start with a set of candidate solutions for the optimization parameter and try to find the best solution through a trial-and-error process: learning automata give a better solution a higher probability to be tried; reinforcement learning keeps a value for each candidate that reflects the candidate's quality; genetic algorithms have a dynamic set of candidates and improves the quality of the set by mimicking the evolutionary process. We describe how some optimization problems in Time Warp can be transformed into a search problem, and how the learning methods can be utilized to directly search for the optimal value for the system control parameter. Compared with the analytical model-based approach, these methods are more generic in nature. Since the search is based on actual run-time performance of different values for the control parameter, the learning methods also better reflect the simulation reality.
L'histoire du protocole Time Warp, l'un des protocoles de synchronisation le plus couramment utilise en matiere de simulation parallele et distribue, a ete une histoire de optimisations. Habituellement, la problemes d'optimisation sont resolus par creer un modele analytique pour le systeme de simulation par une analyse minutieuse de la comportement de Time Warp. Le modele est exprime comme une fonction de la forme ferme entre les variables d'etat du systeme et un parametre de controle. Au moment de l'execution, les variables d'etat du systeme sont mesures et servent ensuite à calculer la valeur du parametre de controle. Cette approche rend la qualite de l'optimisation dépend fortement sur la maniere de pres le modele reflete reellement la realite de simulation. En raison de la simplications qui sont necessaires de faire dans le courant de creer de tels modeles, il n'est pas certain que les strategies de controle sont optimale. En outre, les modeles bases sur une application specifique ne peut etre facilement adopte pour d'autres applications.
Dans cette these, comme une approche alternative, nous presentons un certain nombre de algorithmes de direct recherche sans modeles base sur des techniques de controle du systeme, l'apprentissage automatique et evolutive l'informatique, a savoir, l'apprentissage des automates, apprentissage par renforcement, et genetique algorithmes. Ce que ces methodes ont en commun est la notion d'apprentissage. Contrairement aux methodes traditionnelles utilisees dans d'optimisation de Time Warp, ces apprentissages méthodes traitent le simulateur Time Warp comme une boite noire. Ils commencent par une ensemble de solutions candidates pour le parametre d'optimisation et essayer de trouver la meilleure solution grace a un essai-erreur de processus: l'apprentissage d'automates donner un meilleur solution d'une plus grande probabilite d'etre juge; apprentissage par renforcement garde un valeur pour chaque candidat qui reflete la qualite du candidat; genetiques algorithmes ont un ensemble dynamique de candidats et ameliore la qualite de la mis en imitant le processus evolutif. Nous decrivons comment certains probl`emes d'optimisation dans Time Warp peut etre transforme en un probleme de recherche, et comment les methodes d'apprentissage peut etre utilise pour directement recherche de la valeur optimale pour le parametre de controle du systeme. En comparaison avec le modele analytique approche, ces methodes sont plus generiques dans la nature. Comme la recherche est basee sur l'ecoulement des performances en temps reel des differents valeurs pour le parametre de controle, les methodes d'apprentissage aussi de mieux refleter la realite de la simulation.
45
Ross, Stéphane. "Model-based Bayesian reinforcement learning in complex domains." Thesis, McGill University, 2008. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=21960.
Full textAbstract:
Reinforcement Learning has emerged as a useful framework for learning to perform a task optimally from experience in unknown systems. A major problem for such learning algorithms is how to balance optimally the exploration of the system, to gather knowledge, and the exploitation of current knowledge, to complete the task. Model-based Bayesian Reinforcement Learning (BRL) methods provide an optimal solution to this problem by formulating it as a planning problem under uncertainty. However, the complexity of these methods has so far limited their applicability to small and simple domains. To improve the applicability of model-based BRL, this thesis presents several extensions to more complex and realistic systems, such as partially observable and continuous domains. To improve learning efficiency in large systems, this thesis includes another extension to automatically learn and exploit the structure of the system. Approximate algorithms are proposed to efficiently solve the resulting inference and planning problems.L'apprentissage par renforcement a émergé comme une technique utile pour apprendre à accomplir une tâche de façon optimale à partir d'expérience dans les systèmes inconnus. L'un des problèmes majeurs de ces algorithmes d'apprentissage est comment balancer de façon optimale l'exploration du système, pour acquérir des connaissances, et l'exploitation des connaissances actuelles, pour compléter la tâche. L'apprentissage par renforcement bayésien avec modèle permet de résoudre ce problème de façon optimale en le formulant comme un problème de planification dans l'incertain. La complexité de telles méthodes a toutefois limité leur applicabilité à de petits domaines simples. Afin d'améliorer l'applicabilité de l'apprentissage par renforcement bayésian avec modèle, cette thèse presente plusieurs extensions de ces méthodes à des systèmes beaucoup plus complexes et réalistes, où le domaine est partiellement observable et/ou continu. Afin d'améliorer l'efficacité de l'apprentissage dans les gros systèmes, cette thèse inclue une autre extension qui permet d'apprendre automatiquement et d'exploiter la structure du système. Des algorithmes approximatifs sont proposés pour résoudre efficacement les problèmes d'inference et de planification résultants.
46
Guez, Arthur. "Adaptive control of epileptic seizures using reinforcement learning." Thesis, McGill University, 2010. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=95059.
Full textAbstract:
This thesis presents a new methodology for automatically learning an optimal neurostimulation strategy for the treatment of epilepsy. The technical challenge is to automatically modulate neurostimulation parameters, as a function of the observed field potential recording, so as to minimize the frequency and duration of seizures. The methodology leverages recent techniques from the machine learning literature, in particular the reinforcement learning paradigm, to formalize this optimization problem. We present an algorithm which is able to learn an adaptive neurostimulation strategy directly from labeled training data acquired from animal brain tissues. Our results suggest that this methodology can be used to automatically find a stimulation strategy which effectively reduces the incidence of seizures, while also minimizing the amount of stimulation applied. This work highlights the crucial role that modern machine learning techniques can play in the optimization of treatment strategies for patients with chronic disorders such as epilepsy.Cette thèse présente une nouvelle méthodologie pour apprendre, de façon automatique, une stratégie optimale de neurostimulation pour le traitement de l'épilepsie. Le défi technique est de moduler automatiquement les paramètres de stimulation, en fonction de l'enregistrement de potentiel de champ observé, afin de minimiser la fréquence et la durée des crises d'épilepsie. Cette méthodologie fait appel à des techniques récentes développées dans le domaine de l'apprentissage machine, en particulier le paradigme d'apprentissage par renforcement, pour formaliser ce problème d'optimisation. Nous présentons un algorithme qui est capable d'apprendre une stratégie adaptative de neurostimulation, et ce directement à partir de données d'apprentissage, étiquetées, acquises depuis des tissus de cerveaux d'animaux. Nos résultats suggèrent que cette méthodologie peut être utiliser pour trouver, automatiquement, une stratégie de stimulation qui réduit efficacement l'indicence des crises d'épilepsie tout en minimisant le nombre de stimulations appliquées. Ce travail met en évidence le rôle crucial que les techniques modernes d'apprentissage machine peuvent jouer dans l'optimisation de stratégies de traitements pour des patients souffrant de maladies chroniques telle l'épilepsie.
47
Gendron-Bellemare, Marc. "Learning prediction and abstraction in partially observable models." Thesis, McGill University, 2007. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=18471.
Full textAbstract:
Markov models have been a keystone in Artificial Intelligence for many decades. However, they remain unsatisfactory when the environment modeled is partially observable. There are pathological examples where no history of fixed length is sufficient for accurate deci- sion making. On the other hand, working with a hidden state (such as in POMDPs) has a high computational cost. In order to circumvent this problem, I suggest the use of a context- based model. My approach replaces strict transition probabilities by a linear approximation of probability distributions. The method proposed provides a trade-off between a fully and partially observable model. I also discuss improving the approximation by constructing history-based features. Simple examples are given in order to show that the linear approx- imation can behave like certain Markov models. Empirical results on feature construction are also given to illustrate the power of the approach.Depuis plusieurs déecennies, les modèeles de Markov forment l'une des bases de l'Intelligence Artificielle. Lorsque l'environnement modélisé n'est que partiellement observable, cepen- dant, ceux-ci demeurent insatisfaisants. Il est connu que la prise de décision optimale dans certains problèmes exige un historique infini. D'un autre côté, faire appel au con- cept d'état caché (tel qu'à travers l'utilisation de POMDPs) implique un coût computa- tionnel plus élevé. Afin de pallier à ce problème, je propose un modèle se servant une représentation concise de l'historique. Plutôt que de stocker un modèle parfait des prob- abilitités de transition, mon approche emploie d'une approximation linéaire des distribu- tions de probabilités. La méthode proposée est un compromis entre les modèles partielle- ment et complètement observables. Je traite aussi de la construction d'éléments en lien avec l'historique afin d'améliorer l'approximation linéaire. Des exemples restreints sont présentés afin de montrer qu'une approximation linéaire de certains modèles de Markov peut être atteinte. Des résultats empiriques au niveau de la construction d'éléments sont aussi présentés afin d'illustrer les bénéfices de mon approche.
48
Png, ShaoWei. "Bayesian reinforcement learning for POMDP-based dialogue systems." Thesis, McGill University, 2011. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=104830.
Full textAbstract:
Spoken dialogue systems are gaining popularity with improvements in speech recognition technologies. Dialogue systems have been modeled effectively using Partially observable Markov decision processes (POMDPs), achieving improvements in robustness. However, past research on POMDP-based dialogue systems usually assumes that the model parameters are known. This limitation can be addressed through model-based Bayesian reinforcement learning, which offers a rich framework for simultaneous learning and planning. However, due to the high complexity of the framework, a major challenge is to scale up these algorithms for complex dialogue systems. In this work, we show that by exploiting certain known components of the system, such as knowledge of symmetrical properties, and using an approximate on-line planning algorithm, we are able to apply Bayesian RL on several realistic spoken dialogue system domains. We consider several experimental domains. First, a small synthetic data case, where we illustrate several properties of the approach. Second, a small dialogue manager based on the SACTI1 corpus which contains 144 dialogues between 36 users and 12 experts. Third, a dialogue manager aimed at patients with dementia, to assist them with activities of daily living. Finally, we consider a large dialogue manager designed to help patients to operate a wheelchair.Les systèmes de dialogues sont de plus en plus populaires depuis l'amélioration des technologies de reconnaissance vocale. Ces systèmes de dialogues peuvent être modélisés efficacement à l'aide des processus de décision markoviens partiellement observables (POMDP). Toutefois, les recherches antérieures supposent généralement une connaissance des paramètres du modèle. L'apprentissage par renforcement basée sur un modèle bayéesien, qui offre un cadre riche pour l'apprentissage et la planification simultanéee, peut éeliminer la néecessitée de cette supposition à cause de la grande complexitée du cadre, le déeveloppement de ces algorithmes pour les systèmes de dialogues complexes repréesente un déefi majeur. Dans ce document, nous déemontrons qu'en exploitant certaines propriéetées connues du système, comme les syméetries, et en utilisant un algorithme de planification approximatif en ligne, nous sommes capables d'appliquer les techniques d'apprentissage par renforcement bayéesien dans le cadre de sur plusieurs domaines de dialogues réealistes. Nous considéerons quelques domaines expéerimentaux. Le premier comprend des donnéees synthéetiques qui servent à illustrer plusieurs propriéetées de notre approche. Le deuxième est un gestionnaire de dialogues basée sur le corpus SACTI1 qui contient 144 dialogues entre 36 utilisateurs et 12 experts. Le troisième gestionnaire aide les patients atteints de déemence à vivre au quotidien. Finalement, nous considéerons un grand gestionnaire de dialogue qui assise des patients à manoeuvrer une chaise roulante automatiséee.
49
Comanici, Gheorghe. "Optimal time scales for reinforcement learning behaviour strategies." Thesis, McGill University, 2010. http://digitool.Library.McGill.CA:8881/R/?func=dbin-jump-full&object_id=92340.
Full text
50
Carrera, Viñas Arnau. "Robot learning applied to autonomous underwater vehicles for intervention tasks." Doctoral thesis, Universitat de Girona, 2017. http://hdl.handle.net/10803/450868.
Full textAbstract:
The interest in performing underwater tasks using Autonomous Underwater
Vehicles (AUVs) has been growing over the past few years. In this thesis, a
flexible framework for underwater interventions using a Learning by
Demonstration algorithm as a core has been developed. This algorithm allows
to the robot's user to transfer a skill or knowledge to the I-AUV using a
natural and intuitive form.
The developed framework for interventions has been tailored to the GIRONA 500
AUV in order to enable it to perform an underwater valve turning task under
different conditions. The GIRONA 500 has been equipped with a 4 DOF
Manipulator and a custom end-effector.
Throughout this thesis, the experiments developed have been carried out in a
mock-up scenario of a sub-sea installation with a valve panel. The difficulty
of the task has been increased gradually in order to test the new
improvements and the robustness in the proposed frameworkDurant les últimes dècades ha augmentat l’interès en la utilització de Vehicles Autònoms Submarins (AUVs) per realitzar tasques submarines. En aquesta tesis s’ha desenvolupat un marc de treball (framework) per a realitzar intervencions submarines amb AUVs basat en un algorisme d’Aprenentatge per Demostració (LbD). Aquest algorisme permet a l’usuari del robot transferir el seu coneixement al vehicle d’intervenció d’una forma natural. El framework desenvolupat s’ha ajustat a les característiques del GIRONA 500 AUV, amb l’objectiu de que pugui girar vàlvules submarines en diverses condicions. El GIRONA 500 s’ha equipat amb un braç robòtic i un element terminal personalitzat. Al llarg de tota la tesis s’ha utilitzat com entorn de desenvolupament un tanc d’aigua amb una recreació d’un escenari d’intervenció subaquàtic on s’han de girar determinades vàlvules d’un panell. El grau de dificultat de la tasca s’ha incrementat de forma gradual, per tal de poder provar les noves millores