Dissertations / Theses on the topic 'Machine Learning Model Robustness'
To see the other types of publications on this topic, follow the link: Machine Learning Model Robustness.
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 'Machine Learning Model Robustness.'
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
Adams, William A. "Analysis of Robustness in Lane Detection using Machine Learning Models." Ohio University / OhioLINK, 2015. http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1449167611.
Full text
2
Lundström, Linnea. "Formally Verifying the Robustness of Machine Learning Models : A Comparative Study." Thesis, Linköpings universitet, Programvara och system, 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-167504.
Full textAbstract:
Machine learning models have become increasingly popular in recent years, and not without reason. They enable software to become more powerful, and with less human involvement. As a consequence however, the actions of the software are hard for a human to understand and anticipate. This prohibits the use of machine learning in systems where safety has to be assured, typically using formal proofs of relevant properties. This thesis is focused on robustness - one of many properties that can impact the safety of a system. There are several tools available that enable formal robustness verification of machine learning models, and a goal of this thesis is to evaluate their performance. A variety of machine learning models are also assessed according to how robust they can be proved to be. A digit recognition problem was used in order to evaluate how sensitive different model types are to perturbations of pixels in an image, and also to assess the performance of applicable verification tools. On this particular problem, we discovered that a Support Vector Machine demonstrates the highest degree of robustness, which could be verified with short enough time using the tool SAVer. In addition, machine learning models were trained on a data set consisting of Android applications that are labelled either as malware or benign. In this verification problem, we check whether adding permission requests to an application that is malware can make it become labelled as benign. For this problem, a Gradient Boosting Machine proved to be the most robust with a very short verification time using the tool VoTE. Although not the most robust, Neural Networks were proved to be relatively robust on both problems using the tool ERAN, whereas Random Forests performed the worst, in terms of robustness.
3
MAURI, LARA. "DATA PARTITIONING AND COMPENSATION TECHNIQUES FOR SECURE TRAINING OF MACHINE LEARNING MODELS." Doctoral thesis, Università degli Studi di Milano, 2022. http://hdl.handle.net/2434/932387.
Full textAbstract:
Advances in Machine Learning (ML), coupled with increased availability of huge amounts of data collected from diverse sources and improvements in computing power, have led to a widespread adoption of ML-based solutions in critical application scenarios. However, ML models intrinsically introduce new security vulnerabilities within the systems into which they are integrated, thereby expanding their attack surface. The security of ML-based systems hinges on the robustness of the ML model employed. By interfering with any of the phases of the learning process, an adversary can manipulate data and prevent the model from learning the correct correlations or mislead it into taking potentially harmful actions. Adversarial ML is a recent research field that addresses two specific research topics. One of them concerns the identification of security issues related to the use of ML models, and the other concerns the design of defense mechanisms to prevent or mitigate the detrimental effects of attacks.
In this dissertation, we investigate how to improve the resilience of ML models against training-time attacks under black-box knowledge assumption on both the attacker and the defender. The main contribution of this work is a novel defense mechanism which combines ensemble models (an approach traditionally used only for increasing the generalization capabilities of the model) and security risk analysis. Specifically, the results from the risk analysis in the input data space are used to guide the partitioning of the training data via an unsupervised technique. Then, we employ an ensemble of models, each trained on a different partition, and combine their output based on a majority voting mechanism to obtain the final prediction. Experiments are carried out on a publicly available dataset to assess the effectiveness of the proposed method. This novel defence technique is complemented by two other contributions, which respectively support using a Distributed Ledger to make training data tampering less convenient for attackers, and using a quantitative index to compute ML models’ performance degradation before and after the deployment of the defense. Taken together, this set of techniques provides a framework to improve the robustness of the ML lifecycle.
4
Rado, Omesaad A. M. "Contributions to evaluation of machine learning models. Applicability domain of classification models." Thesis, University of Bradford, 2019. http://hdl.handle.net/10454/18447.
Full textAbstract:
Artificial intelligence (AI) and machine learning (ML) present some application opportunities and
challenges that can be framed as learning problems. The performance of machine learning models
depends on algorithms and the data. Moreover, learning algorithms create a model of reality through
learning and testing with data processes, and their performance shows an agreement degree of their
assumed model with reality. ML algorithms have been successfully used in numerous classification
problems. With the developing popularity of using ML models for many purposes in different domains,
the validation of such predictive models is currently required more formally. Traditionally, there are
many studies related to model evaluation, robustness, reliability, and the quality of the data and the
data-driven models. However, those studies do not consider the concept of the applicability domain
(AD) yet. The issue is that the AD is not often well defined, or it is not defined at all in many fields. This
work investigates the robustness of ML classification models from the applicability domain
perspective. A standard definition of applicability domain regards the spaces in which the model
provides results with specific reliability.
The main aim of this study is to investigate the connection between the applicability domain approach
and the classification model performance. We are examining the usefulness of assessing the AD for
the classification model, i.e. reliability, reuse, robustness of classifiers. The work is implemented using
three approaches, and these approaches are conducted in three various attempts: firstly, assessing
the applicability domain for the classification model; secondly, investigating the robustness of the
classification model based on the applicability domain approach; thirdly, selecting an optimal model
using Pareto optimality. The experiments in this work are illustrated by considering different machine
learning algorithms for binary and multi-class classifications for healthcare datasets from public
benchmark data repositories. In the first approach, the decision trees algorithm (DT) is used for the
classification of data in the classification stage. The feature selection method is applied to choose
features for classification. The obtained classifiers are used in the third approach for selection of
models using Pareto optimality. The second approach is implemented using three steps; namely,
building classification model; generating synthetic data; and evaluating the obtained results.
The results obtained from the study provide an understanding of how the proposed approach can help
to define the model’s robustness and the applicability domain, for providing reliable outputs. These
approaches open opportunities for classification data and model management. The proposed
algorithms are implemented through a set of experiments on classification accuracy of instances,
which fall in the domain of the model. For the first approach, by considering all the features, the
highest accuracy obtained is 0.98, with thresholds average of 0.34 for Breast cancer dataset. After
applying recursive feature elimination (RFE) method, the accuracy is 0.96% with 0.27 thresholds
average. For the robustness of the classification model based on the applicability domain approach,
the minimum accuracy is 0.62% for Indian Liver Patient data at r=0.10, and the maximum accuracy is
0.99% for Thyroid dataset at r=0.10. For the selection of an optimal model using Pareto optimality,
the optimally selected classifier gives the accuracy of 0.94% with 0.35 thresholds average.
This research investigates critical aspects of the applicability domain as related to the robustness of
classification ML algorithms. However, the performance of machine learning techniques depends on
the degree of reliable predictions of the model. In the literature, the robustness of the ML model can
be defined as the ability of the model to provide the testing error close to the training error. Moreover,
the properties can describe the stability of the model performance when being tested on the new
datasets. Concluding, this thesis introduced the concept of applicability domain for classifiers and
tested the use of this concept with some case studies on health-related public benchmark datasets.Ministry of Higher Education in Libya
5
Cherief-Abdellatif, Badr-Eddine. "Contributions to the theoretical study of variational inference and robustness." Electronic Thesis or Diss., Institut polytechnique de Paris, 2020. http://www.theses.fr/2020IPPAG001.
Full textAbstract:
Cette thèse de doctorat traite de l'inférence variationnelle et de la robustesse en statistique et en machine learning. Plus précisément, elle se concentre sur les propriétés statistiques des approximations variationnelles et sur la conception d'algorithmes efficaces pour les calculer de manière séquentielle, et étudie les estimateurs basés sur le Maximum Mean Discrepancy comme règles d'apprentissage qui sont robustes à la mauvaise spécification du modèle.Ces dernières années, l'inférence variationnelle a été largement étudiée du point de vue computationnel, cependant, la littérature n'a accordé que peu d'attention à ses propriétés théoriques jusqu'à très récemment. Dans cette thèse, nous étudions la consistence des approximations variationnelles dans divers modèles statistiques et les conditions qui assurent leur consistence. En particulier, nous abordons le cas des modèles de mélange et des réseaux de neurones profonds. Nous justifions également d'un point de vue théorique l'utilisation de la stratégie de maximisation de l'ELBO, un critère numérique qui est largement utilisé dans la communauté VB pour la sélection de modèle et dont l'efficacité a déjà été confirmée en pratique. En outre, l'inférence Bayésienne offre un cadre d'apprentissage en ligne attrayant pour analyser des données séquentielles, et offre des garanties de généralisation qui restent valables même en cas de mauvaise spécification des modèles et en présence d'adversaires. Malheureusement, l'inférence Bayésienne exacte est rarement tractable en pratique et des méthodes d'approximation sont généralement employées, mais ces méthodes préservent-elles les propriétés de généralisation de l'inférence Bayésienne ? Dans cette thèse, nous montrons que c'est effectivement le cas pour certains algorithmes d'inférence variationnelle (VI). Nous proposons de nouveaux algorithmes tempérés en ligne et nous en déduisons des bornes de généralisation. Notre résultat théorique repose sur la convexité de l'objectif variationnel, mais nous soutenons que notre résultat devrait être plus général et présentons des preuves empiriques à l'appui. Notre travail donne des justifications théoriques en faveur des algorithmes en ligne qui s'appuient sur des méthodes Bayésiennes approchées.Une autre question d'intérêt majeur en statistique qui est abordée dans cette thèse est la conception d'une procédure d'estimation universelle. Cette question est d'un intérêt majeur, notamment parce qu'elle conduit à des estimateurs robustes, un thème d'actualité en statistique et en machine learning. Nous abordons le problème de l'estimation universelle en utilisant un estimateur de minimisation de distance basé sur la Maximum Mean Discrepancy. Nous montrons que l'estimateur est robuste à la fois à la dépendance et à la présence de valeurs aberrantes dans le jeu de données. Nous mettons également en évidence les liens qui peuvent exister avec les estimateurs de minimisation de distance utilisant la distance L2. Enfin, nous présentons une étude théorique de l'algorithme de descente de gradient stochastique utilisé pour calculer l'estimateur, et nous étayons nos conclusions par des simulations numériques. Nous proposons également une version Bayésienne de notre estimateur, que nous étudions à la fois d'un point de vue théorique et d'un point de vue computationnelThis PhD thesis deals with variational inference and robustness. More precisely, it focuses on the statistical properties of variational approximations and the design of efficient algorithms for computing them in an online fashion, and investigates Maximum Mean Discrepancy based estimators as learning rules that are robust to model misspecification.In recent years, variational inference has been extensively studied from the computational viewpoint, but only little attention has been put in the literature towards theoretical properties of variational approximations until very recently. In this thesis, we investigate the consistency of variational approximations in various statistical models and the conditions that ensure the consistency of variational approximations. In particular, we tackle the special case of mixture models and deep neural networks. We also justify in theory the use of the ELBO maximization strategy, a model selection criterion that is widely used in the Variational Bayes community and is known to work well in practice.Moreover, Bayesian inference provides an attractive online-learning framework to analyze sequential data, and offers generalization guarantees which hold even under model mismatch and with adversaries. Unfortunately, exact Bayesian inference is rarely feasible in practice and approximation methods are usually employed, but do such methods preserve the generalization properties of Bayesian inference? In this thesis, we show that this is indeed the case for some variational inference algorithms. We propose new online, tempered variational algorithms and derive their generalization bounds. Our theoretical result relies on the convexity of the variational objective, but we argue that our result should hold more generally and present empirical evidence in support of this. Our work presents theoretical justifications in favor of online algorithms that rely on approximate Bayesian methods. Another point that is addressed in this thesis is the design of a universal estimation procedure. This question is of major interest, in particular because it leads to robust estimators, a very hot topic in statistics and machine learning. We tackle the problem of universal estimation using a minimum distance estimator based on the Maximum Mean Discrepancy. We show that the estimator is robust to both dependence and to the presence of outliers in the dataset. We also highlight the connections that may exist with minimum distance estimators using L2-distance. Finally, we provide a theoretical study of the stochastic gradient descent algorithm used to compute the estimator, and we support our findings with numerical simulations. We also propose a Bayesian version of our estimator, that we study from both a theoretical and a computational points of view
6
Ilyas, Andrew. "On practical robustness of machine learning systems." Thesis, Massachusetts Institute of Technology, 2018. https://hdl.handle.net/1721.1/122911.
Full textAbstract:
This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.Thesis: M. Eng., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2018
Cataloged from student-submitted PDF version of thesis.
Includes bibliographical references (pages 71-79).
We consider the importance of robustness in evaluating machine learning systems, an in particular systems involving deep learning. We consider these systems' vulnerability to adversarial examples--subtle, crafted perturbations to inputs which induce large change in output. We show that these adversarial examples are not only theoretical concern, by desigining the first 3D adversarial objects, and by demonstrating that these examples can be constructed even when malicious actors have little power. We suggest a potential avenue for building robust deep learning models by leveraging generative models.
by Andrew Ilyas.
M. Eng.
M.Eng. Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science
7
Ishii, Shotaro, and David Ljunggren. "A Comparative Analysis of Robustness to Noise in Machine Learning Classifiers." Thesis, KTH, Skolan för elektroteknik och datavetenskap (EECS), 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-302532.
Full textAbstract:
Data that stems from real measurements often to some degree contain distortions. Such distortions are generally referred to as noise in machine learning terminology, and can lead to decreased classification accuracy and poor prediction results. In this study, three machine learning classifiers were compared by their performance and robustness in the presence of noise. More specifically, random forests, support vector machines and artificial neural networks were trained and compared on four different data sets with varying levels of noise artificially added to them. In summary, the random forest classifier performed the best and was the most robust classifier at eight out of ten of noise levels, closely followed by the artificial neural network classifier. At the two remaining noise levels, the support vector machine classifier with a linear kernel performed the best and was the most robust classifier.Data som härstammar från verkliga mätningar innehåller ofta förvrängningar i viss utsträckning. Sådana förvrängningar kan i vissa fall leda till försämrad klassificeringsnoggrannhet. I den här studien jämförs tre klassificeringsalgoritmer med avseende på hur pass robusta de är när den data de presenteras innehåller syntetiska förvrängningar. Mer specifikt så tränades och jämfördes slumpskogar, stödvektormaskiner och artificiella neuronnät på fyra olika mängder data med varierande nivåer av syntetiska förvrängningar. Sammanfattningsvis så presterade slumpskogen bäst, och var den mest robusta klassificeringsalgoritmen på åtta av tio förvrängningsnivåer, tätt följt av det artificiella neuronnätet. På de två återstående förvrängningsnivåerna presterade stödvektormaskinen med linjär kärna bäst och var den mest robusta klassificeringsalgoritmen.
8
Ebrahimi, Javid. "Robustness of Neural Networks for Discrete Input: An Adversarial Perspective." Thesis, University of Oregon, 2019. http://hdl.handle.net/1794/24535.
Full textAbstract:
In the past few years, evaluating on adversarial examples has become a standard
procedure to measure robustness of deep learning models. Literature on adversarial
examples for neural nets has largely focused on image data, which are represented as
points in continuous space. However, a vast proportion of machine learning models
operate on discrete input, and thus demand a similar rigor in understanding their
vulnerabilities and robustness. We study robustness of neural network architectures
for textual and graph inputs, through the lens of adversarial input perturbations.
We will cover methods for both attacks and defense; we will focus on 1) addressing
challenges in optimization for creating adversarial perturbations for discrete data;
2) evaluating and contrasting white-box and black-box adversarial examples; and 3)
proposing efficient methods to make the models robust against adversarial attacks.
9
Fagogenis, Georgios. "Increasing the robustness of autonomous systems to hardware degradation using machine learning." Thesis, Heriot-Watt University, 2016. http://hdl.handle.net/10399/3378.
Full textAbstract:
Autonomous systems perform predetermined tasks (missions) with minimum supervision. In most applications, the state of the world changes with time. Sensors are employed to measure part or whole of the world's state. However, sensors often fail amidst operation; feeding as such decision-making with wrong information about the world. Moreover, hardware degradation may alter dynamic behaviour, and subsequently the capabilities, of an autonomous system; rendering the original mission infeasible. This thesis applies machine learning to yield powerful and robust tools that can facilitate autonomy in modern systems. Incremental kernel regression is used for dynamic modelling. Algorithms of this sort are easy to train and are highly adaptive. Adaptivity allows for model adjustments, whenever the environment of operation changes. Bayesian reasoning provides a rigorous framework for addressing uncertainty. Moreover, using Bayesian Networks, complex inference regarding hardware degradation can be answered. Specifically, adaptive modelling is combined with Bayesian reasoning to yield recursive estimation algorithms that are robust to sensor failures. Two solutions are presented by extending existing recursive estimation algorithms from the robotics literature. The algorithms are deployed on an underwater vehicle and the performance is assessed in real-world experiments. A comparison against standard filters is also provided. Next, the previous algorithms are extended to consider sensor and actuator failures jointly. An algorithm that can detect thruster failures in an Autonomous Underwater Vehicle has been developed. Moreover, the algorithm adapts the dynamic model online to compensate for the detected fault. The performance of this algorithm was also tested in a real-world application. One step further than hardware fault detection, prognostics predict how much longer can a particular hardware component operate normally. Ubiquitous sensors in modern systems render data-driven prognostics a viable solution. However, training is based on skewed datasets; datasets where the samples from the faulty region of operation are much fewer than the ones from the healthy region of operation. This thesis presents a prognostic algorithm that tackles the problem of imbalanced (skewed) datasets.
10
Haussamer, Nicolai Haussamer. "Model Calibration with Machine Learning." Master's thesis, University of Cape Town, 2018. http://hdl.handle.net/11427/29451.
Full textAbstract:
This dissertation focuses on the application of neural networks to financial model calibration. It provides an introduction to the mathematics of basic neural networks and training algorithms. Two simplified experiments based on the Black-Scholes and constant elasticity of variance models are used to demonstrate the potential usefulness of neural networks in calibration. In addition, the main experiment features the calibration of the Heston model using model-generated data. In the experiment, we show that the calibrated model parameters reprice a set of options to a mean relative implied volatility error of less than one per cent. The limitations and shortcomings of neural networks in model calibration are also investigated and discussed.
11
Zhao, Yajing. "Chaotic Model Prediction with Machine Learning." BYU ScholarsArchive, 2020. https://scholarsarchive.byu.edu/etd/8419.
Full textAbstract:
Chaos theory is a branch of modern mathematics concerning the non-linear dynamic systems that are highly sensitive to their initial states. It has extensive real-world applications, such as weather forecasting and stock market prediction. The Lorenz system, defined by three ordinary differential equations (ODEs), is one of the simplest and most popular chaotic models. Historically research has focused on understanding the Lorenz system's mathematical characteristics and dynamical evolution including the inherent chaotic features it possesses. In this thesis, we take a data-driven approach and propose the task of predicting future states of the chaotic system from limited observations. We explore two directions, answering two distinct fundamental questions of the system based on how informed we are about the underlying model. When we know the data is generated by the Lorenz System with unknown parameters, our task becomes parameter estimation (a white-box problem), or the ``inverse'' problem. When we know nothing about the underlying model (a black-box problem), our task becomes sequence prediction. We propose two algorithms for the white-box problem: Markov-Chain-Monte-Carlo (MCMC) and a Multi-Layer-Perceptron (MLP). Specially, we propose to use the Metropolis-Hastings (MH) algorithm with an additional random walk to avoid the sampler being trapped into local energy wells. The MH algorithm achieves moderate success in predicting the $\rho$ value from the data, but fails at the other two parameters. Our simple MLP model is able to attain high accuracy in terms of the $l_2$ distance between the prediction and ground truth for $\rho$ as well, but also fails to converge satisfactorily for the remaining parameters. We use a Recurrent Neural Network (RNN) to tackle the black-box problem. We implement and experiment with several RNN architectures including Elman RNN, LSTM, and GRU and demonstrate the relative strengths and weaknesses of each of these methods. Our results demonstrate the promising role of machine learning and modern statistical data science methods in the study of chaotic dynamic systems. The code for all of our experiments can be found on \url{https://github.com/Yajing-Zhao/}
12
Nitesh, Varma Rudraraju Nitesh, and Boyanapally Varun Varun. "Data Quality Model for Machine Learning." Thesis, Blekinge Tekniska Högskola, Institutionen för programvaruteknik, 2019. http://urn.kb.se/resolve?urn=urn:nbn:se:bth-18498.
Full textAbstract:
Context: - Machine learning is a part of artificial intelligence, this area is now continuously growing day by day. Most internet related services such as Social media service, Email Spam, E-commerce sites, Search engines are now using machine learning. The Quality of machine learning output relies on the input data, so the input data is crucial for machine learning and good quality of input data can give a better outcome to the machine learning system. In order to achieve quality data, a data scientist can use a data quality model on data of machine learning. Data quality model can help data scientists to monitor and control the input data of machine learning. But there is no considerable amount of research done on data quality attributes and data quality model for machine learning. Objectives: - The primary objectives of this paper are to find and understand the state-of-art and state-of-practice on data quality attributes for machine learning, and to develop a data quality model for machine learning in collaboration with data scientists. Methods: - This paper mainly consists of two studies: - 1) Conducted a literature review in the different database in order to identify literature on data quality attributes and data quality model for machine learning. 2) An in-depth interview study was conducted to allow a better understanding and verifying of data quality attributes that we identified from our literature review study, this process is carried out with the collaboration of data scientists from multiple locations. Totally of 15 interviews were performed and based on the results we proposed a data quality model based on these interviewees perspective. Result: - We identified 16 data quality attributes as important from our study which is based on the perspective of experienced data scientists who were interviewed in this study. With these selected data quality attributes, we proposed a data quality model with which quality of data for machine learning can be monitored and improved by data scientists, and effects of these data quality attributes on machine learning have also been stated. Conclusion: - This study signifies the importance of quality of data, for which we proposed a data quality model for machine learning based on the industrial experiences of a data scientist. This research gap is a benefit to all machine learning practitioners and data scientists who intended to identify quality data for machine learning. In order to prove that data quality attributes in the data quality model are important, a further experiment can be conducted, which is proposed in future work.
13
Harte, Thomas James. "Discrete-time model-based Iterative Learning Control : stability, monotonicity and robustness." Thesis, University of Sheffield, 2007. http://etheses.whiterose.ac.uk/3624/.
Full textAbstract:
In this thesis a new robustness analysis for model-based Iterative Learning Control (ILC) is presented. ILC is a method of control for systems that are required to track a reference signal in a repetitive manner. The repetitive nature of such a system allows for the use of past information such that the control system iteratively learns control signals that give high levels of tracking. ILC algorithms that learn in a monotonic fashion are desirable as it implies that tracking performance is improved at each iteration. A number of model-based ILC algorithms are known to result in a monotonically converging tracking error signal. However clear and meaningful robustness conditions for monotonic convergence in spite of model uncertainty are lacking. This thesis gives new robustness conditions for monotonically converging tracking error for two-model based ILC algorithms: the inverse and adjoint algorithms. It is found that the two algorithms can always guarantee robust monotone convergence to zero error if the multiplicative plant uncertainty matrix satisfies a matrix positivity requirement. This result is extended to the frequency domain using a simple graphical test. The analysis further extends to a Parameter Optimal control setting where optimisation is applied to the inverse and adjoint algorithms. The results show an increased degree in robust monotone convergence upon a previous attempt to apply optimisation to the two algorithms. This thesis further considers the case where the multiplicative plant uncertainty fails to satisfy the positivity requirement. Robustness analysis shows that use of an appropriately designed filter with the inverse or adjoint algorithm allows a filtered error signal to monotonically converge to zero.
14
Ferdowsi, Khosrowshahi Aidin. "Distributed Machine Learning for Autonomous and Secure Cyber-physical Systems." Diss., Virginia Tech, 2020. http://hdl.handle.net/10919/99466.
Full textAbstract:
Autonomous cyber-physical systems (CPSs) such as autonomous connected vehicles (ACVs), unmanned aerial vehicles (UAVs), critical infrastructure (CI), and the Internet of Things (IoT) will be essential to the functioning of our modern economies and societies. Therefore, maintaining the autonomy of CPSs as well as their stability, robustness, and security (SRS) in face of exogenous and disruptive events is a critical challenge. In particular, it is crucial for CPSs to be able to not only operate optimally in the vicinity of a normal state but to also be robust and secure so as to withstand potential failures, malfunctions, and intentional attacks. However, to evaluate and improve the SRS of CPSs one must overcome many technical challenges such as the unpredictable behavior of a CPS's cyber-physical environment, the vulnerability to various disruptive events, and the interdependency between CPSs. The primary goal of this dissertation is, thus, to develop novel foundational analytical tools, that weave together notions from machine learning, game theory, and control theory, in order to study, analyze, and optimize SRS of autonomous CPSs. Towards achieving this overarching goal, this dissertation led to several major contributions. First, a comprehensive control and learning framework was proposed to thwart cyber and physical attacks on ACV networks. This framework brings together new ideas from optimal control and reinforcement learning (RL) to derive a new optimal safe controller for ACVs in order to maximize the street traffic flow while minimizing the risk of accidents. Simulation results show that the proposed optimal safe controller outperforms the current state of the art controllers by maximizing the robustness of ACVs to physical attacks. Furthermore, using techniques from convex optimization and deep RL a joint trajectory and scheduling policy is proposed in UAV-assisted networks that aims at maintaining the freshness of ground node data at the UAV. The analytical and simulation results show that the proposed policy can outperform policies such discretized state RL and value-based methods in terms of maximizing the freshness of data.
Second, in the IoT domain, a novel watermarking algorithm, based on long short term memory cells, is proposed for dynamic authentication of IoT signals. The proposed watermarking algorithm is coupled with a game-theoretic framework so as to enable efficient authentication in massive IoT systems. Simulation results show that using our approach, IoT messages can be transmitted from IoT devices with an almost 100% reliability.
Next, a brainstorming generative adversarial network (BGAN) framework is proposed. It is shown that this framework can learn to generate real-looking data in a distributed fashion while preserving the privacy of agents (e.g. IoT devices, ACVs, etc). The analytical and simulation results show that the proposed BGAN architecture allows heterogeneous neural network designs for agents, works without reliance on a central controller, and has a lower communication over head compared to other state-of-the-art distributed architectures.
Last, but not least, the SRS challenges of interdependent CI (ICI) are addressed. Novel game-theoretic frameworks are proposed that allow the ICI administrator to assign different protection levels on ICI components to maximizing the expected ICI security. The mixed-strategy Nash of the games are derived analytically. Simulation results coupled with theoretical analysis show that, using the proposed games, the administrator can maximize the security level in ICI components. In summary, this dissertation provided major contributions across the areas of CPSs, machine learning, game theory, and control theory with the goal of ensuring SRS across various domains such as autonomous vehicle networks, IoT systems, and ICIs. The proposed approaches provide the necessary fundamentals that can lay the foundations of SRS in CPSs and pave the way toward the practical deployment of autonomous CPSs and applications.Doctor of Philosophy
In order to deliver innovative technological services to their residents, smart cities will rely on autonomous cyber-physical systems (CPSs) such as cars, drones, sensors, power grids, and other networks of digital devices. Maintaining stability, robustness, and security (SRS) of those smart city CPSs is essential for the functioning of our modern economies and societies. SRS can be defined as the ability of a CPS, such as an autonomous vehicular system, to operate without disruption in its quality of service. In order to guarantee SRS of CPSs one must overcome many technical challenges such as CPSs' vulnerability to various disruptive events such as natural disasters or cyber attacks, limited resources, scale, and interdependency. Such challenges must be considered for CPSs in order to design vehicles that are controlled autonomously and whose motion is robust against unpredictable events in their trajectory, to implement stable Internet of digital devices that work with a minimum communication delay, or to secure critical infrastructure to provide services such as electricity, gas, and water systems. The primary goal of this dissertation is, thus, to develop novel foundational analytical tools, that weave together notions from machine learning, game theory, and control theory, in order to study, analyze, and optimize SRS of autonomous CPSs which eventually will improve the quality of service provided by smart cities. To this end, various frameworks and effective algorithms are proposed in order to enhance the SRS of CPSs and pave the way toward the practical deployment of autonomous CPSs and applications. The results show that the developed solutions can enable a CPS to operate efficiently while maintaining its SRS. As such, the outcomes of this research can be used as a building block for the large deployment of smart city technologies that can be of immense benefit to tomorrow's societies.
15
Tigreat, Philippe. "Sparsity, redundancy and robustness in artificial neural networks for learning and memory." Thesis, Ecole nationale supérieure Mines-Télécom Atlantique Bretagne Pays de la Loire, 2017. http://www.theses.fr/2017IMTA0046/document.
Full textAbstract:
L'objectif de la recherche en Intelligence Artificielle (IA) est de répliquer les capacités cognitives humaines au moyen des ordinateurs modernes. Les résultats de ces dernières années semblent annoncer une révolution technologique qui pourrait changer profondément la société. Nous focalisons notre intérêt sur deux aspects cognitifs fondamentaux, l'apprentissage et la mémoire. Les mémoires associatives offrent la possibilité de stocker des éléments d'information et de les récupérer à partir d'une partie de leur contenu, et imitent ainsi la mémoire cérébrale. L'apprentissage profond permet de passer d'une perception analogique du monde extérieur à une représentation parcimonieuse et plus compacte. Dans le chapitre 2, nous présentons une mémoire associative inspirée des réseaux de Willshaw, avec une connectivité contrainte. Cela augmente la performance de récupération des messages et l'efficacité du stockage de l'information.Dans le chapitre 3, une architecture convolutive a été appliquée sur une tâche de lecture de mots partiellement affichés dans des conditions similaires à une étude de psychologie sur des sujets humains. Cette expérimentation montre la similarité de comportement du réseau avec les sujets humains concernant différentes caractéristiques de l'affichage des mots.Le chapitre 4 introduit une méthode de représentation des catégories par des assemblées de neurones dans les réseaux profonds. Pour les problèmes à grand nombre de classes, cela permet de réduire significativement les dimensions d'un réseau.Le chapitre 5 décrit une méthode d'interfaçage des réseaux de neurones profonds non supervisés avec les mémoires associatives à cliquesThe objective of research in Artificial Intelligence (AI) is to reproduce human cognitive abilities by means of modern computers. The results of the last few years seem to announce a technological revolution that could profoundly change society. We focus our interest on two fundamental cognitive aspects, learning and memory. Associative memories offer the possibility to store information elements and to retrieve them using a sub-part of their content, thus mimicking human memory. Deep Learning allows to transition from an analog perception of the outside world to a sparse and more compact representation.In Chapter 2, we present a neural associative memory model inspired by Willshaw networks, with constrained connectivity. This brings an performance improvement in message retrieval and a more efficient storage of information.In Chapter 3, a convolutional architecture was applied on a task of reading partially displayed words under similar conditions as in a former psychology study on human subjects. This experiment put inevidence the similarities in behavior of the network with the human subjects regarding various properties of the display of words.Chapter 4 introduces a new method for representing categories usingneuron assemblies in deep networks. For problems with a large number of classes, this allows to reduce significantly the dimensions of a network.Chapter 5 describes a method for interfacing deep unsupervised networks with clique-based associative memories
16
Menke, Joshua Ephraim. "Improving Machine Learning Through Oracle Learning." BYU ScholarsArchive, 2007. https://scholarsarchive.byu.edu/etd/843.
Full textAbstract:
The following dissertation presents a new paradigm for improving the training of machine learning algorithms, oracle learning. The main idea in oracle learning is that instead of training directly on a set of data, a learning model is trained to approximate a given oracle's behavior on a set of data. This can be beneficial in situations where it is easier to obtain an oracle than it is to use it at application time. It is shown that oracle learning can be applied to more effectively reduce the size of artificial neural networks, to more efficiently take advantage of domain experts by approximating them, and to adapt a problem more effectively to a machine learning algorithm.
17
Wang, Gang. "Solution path algorithms : an efficient model selection approach /." View abstract or full-text, 2007. http://library.ust.hk/cgi/db/thesis.pl?CSED%202007%20WANGG.
Full text
18
Wang, Jiahao. "Vehicular Traffic Flow Prediction Model Using Machine Learning-Based Model." Thesis, Université d'Ottawa / University of Ottawa, 2021. http://hdl.handle.net/10393/42288.
Full textAbstract:
Intelligent Transportation Systems (ITS) have attracted an increasing amount of attention in recent years. Thanks to the fast development of vehicular computing hardware, vehicular sensors and citywide infrastructures, many impressive applications have been proposed under the topic of ITS, such as Vehicular Cloud (VC), intelligent traffic controls, etc. These applications can bring us a safer, more efficient, and also more enjoyable transportation environment. However, an accurate and efficient traffic flow prediction system is needed to achieve these applications, which creates an opportunity for applications under ITS to deal with the possible road situation in advance. To achieve better traffic flow prediction performance, many prediction methods have been proposed, such as mathematical modeling methods, parametric methods, and non-parametric methods. It is always one of the hot topics about how to implement an efficient, robust and accurate vehicular traffic prediction system.
With the help of Machine Learning-based (ML) methods, especially Deep Learning-based (DL) methods, the accuracy of the prediction model is increased. However, we also noticed that there are still many open challenges under ML-based vehicular traffic prediction model real-world implementation. Firstly, the time consumption for DL model training is relatively huge compared to parametric models, such as ARIMA, SARIMA, etc. Second, it is still a hot topic for the road traffic prediction that how to capture the special relationship between road detectors, which is affected by the geographic correlation, as well as the time change. The last but not the least, it is important for us to implement the prediction system in the real world; meanwhile, we should find a way to make use of the advanced technology applied in ITS to improve the prediction system itself.
In our work, we focus on improving the features of the prediction model, which can be helpful for implementing the model in the real word. Firstly, we introduced an optimization strategy for ML-based models' training process, in order to reduce the time cost in this process. Secondly, We provide a new hybrid deep learning model by using GCN and the deep aggregation structure (i.e., the sequence to sequence structure) of the GRU. Meanwhile, in order to solve the real-world prediction problem, i.e., the online prediction task, we provide a new online prediction strategy by using refinement learning. In order to further improve the model's accuracy and efficiency when applied to ITS, we provide a parallel training strategy by using the benefits of the vehicular cloud structure.
19
Ferreira, E. (Eija). "Model selection in time series machine learning applications." Doctoral thesis, Oulun yliopisto, 2015. http://urn.fi/urn:isbn:9789526209012.
Full textAbstract:
Abstract Model selection is a necessary step for any practical modeling task. Since the true model behind a real-world process cannot be known, the goal of model selection is to find the best approximation among a set of candidate models. In this thesis, we discuss model selection in the context of time series machine learning applications. We cover four steps of the commonly followed machine learning process: data preparation, algorithm choice, feature selection and validation. We consider how the characteristics and the amount of data available should guide the selection of algorithms to be used, and how the data set at hand should be divided for model training, selection and validation to optimize the generalizability and future performance of the model. We also consider what are the special restrictions and requirements that need to be taken into account when applying regular machine learning algorithms to time series data. We especially aim to bring forth problems relating model over-fitting and over-selection that might occur due to careless or uninformed application of model selection methods. We present our results in three different time series machine learning application areas: resistance spot welding, exercise energy expenditure estimation and cognitive load modeling. Based on our findings in these studies, we draw general guidelines on which points to consider when starting to solve a new machine learning problem from the point of view of data characteristics, amount of data, computational resources and possible time series nature of the problem. We also discuss how the practical aspects and requirements set by the environment where the final model will be implemented affect the choice of algorithms to useTiivistelmä Mallinvalinta on oleellinen osa minkä tahansa käytännön mallinnusongelman ratkaisua. Koska mallinnettavan ilmiön toiminnan taustalla olevaa todellista mallia ei voida tietää, on mallinvalinnan tarkoituksena valita malliehdokkaiden joukosta sitä lähimpänä oleva malli. Tässä väitöskirjassa käsitellään mallinvalintaa aikasarjamuotoista dataa sisältävissä sovelluksissa neljän koneoppimisprosessissa yleisesti noudatetun askeleen kautta: aineiston esikäsittely, algoritmin valinta, piirteiden valinta ja validointi. Väitöskirjassa tutkitaan, kuinka käytettävissä olevan aineiston ominaisuudet ja määrä tulisi ottaa huomioon algoritmin valinnassa, ja kuinka aineisto tulisi jakaa mallin opetusta, testausta ja validointia varten mallin yleistettävyyden ja tulevan suorituskyvyn optimoimiseksi. Myös erityisiä rajoitteita ja vaatimuksia tavanomaisten koneoppimismenetelmien soveltamiselle aikasarjadataan käsitellään. Työn tavoitteena on erityisesti tuoda esille mallin ylioppimiseen ja ylivalintaan liittyviä ongelmia, jotka voivat seurata mallinvalin- tamenetelmien huolimattomasta tai osaamattomasta käytöstä. Työn käytännön tulokset perustuvat koneoppimismenetelmien soveltamiseen aikasar- jadatan mallinnukseen kolmella eri tutkimusalueella: pistehitsaus, fyysisen harjoittelun aikasen energiankulutuksen arviointi sekä kognitiivisen kuormituksen mallintaminen. Väitöskirja tarjoaa näihin tuloksiin pohjautuen yleisiä suuntaviivoja, joita voidaan käyttää apuna lähdettäessä ratkaisemaan uutta koneoppimisongelmaa erityisesti aineiston ominaisuuksien ja määrän, laskennallisten resurssien sekä ongelman mahdollisen aikasar- jaluonteen näkökulmasta. Työssä pohditaan myös mallin lopullisen toimintaympäristön asettamien käytännön näkökohtien ja rajoitteiden vaikutusta algoritmin valintaan
20
Uziela, Karolis. "Protein Model Quality Assessment : A Machine Learning Approach." Doctoral thesis, Stockholms universitet, Institutionen för biokemi och biofysik, 2017. http://urn.kb.se/resolve?urn=urn:nbn:se:su:diva-137695.
Full textAbstract:
Many protein structure prediction programs exist and they can efficiently generate a number of protein models of a varying quality. One of the problems is that it is difficult to know which model is the best one for a given target sequence. Selecting the best model is one of the major tasks of Model Quality Assessment Programs (MQAPs). These programs are able to predict model accuracy before the native structure is determined. The accuracy estimation can be divided into two parts: global (the whole model accuracy) and local (the accuracy of each residue). ProQ2 is one of the most successful MQAPs for prediction of both local and global model accuracy and is based on a Machine Learning approach. In this thesis, I present my own contribution to Model Quality Assessment (MQA) and the newest developments of ProQ program series. Firstly, I describe a new ProQ2 implementation in the protein modelling software package Rosetta. This new implementation allows use of ProQ2 as a scoring function for conformational sampling inside Rosetta, which was not possible before. Moreover, I present two new methods, ProQ3 and ProQ3D that both outperform their predecessor. ProQ3 introduces new training features that are calculated from Rosetta energy functions and ProQ3D introduces a new machine learning approach based on deep learning. ProQ3 program participated in the 12th Community Wide Experiment on the Critical Assessment of Techniques for Protein Structure Prediction (CASP12) and was one of the best methods in the MQA category. Finally, an important issue in model quality assessment is how to select a target function that the predictor is trying to learn. In the fourth manuscript, I show that MQA results can be improved by selecting a contact-based target function instead of more conventional superposition based functions.At the time of the doctoral defense, the following paper was unpublished and had a status as follows: Paper 3: Manuscript.
21
de, la Rúa Martínez Javier. "Scalable Architecture for Automating Machine Learning Model Monitoring." Thesis, KTH, Skolan för elektroteknik och datavetenskap (EECS), 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-280345.
Full textAbstract:
Last years, due to the advent of more sophisticated tools for exploratory data analysis, data management, Machine Learning (ML) model training and model serving into production, the concept of MLOps has gained more popularity. As an effort to bring DevOps processes to the ML lifecycle, MLOps aims at more automation in the execution of diverse and repetitive tasks along the cycle and at smoother interoperability between teams and tools involved. In this context, the main cloud providers have built their own ML platforms [4, 34, 61], offered as services in their cloud solutions. Moreover, multiple frameworks have emerged to solve concrete problems such as data testing, data labelling, distributed training or prediction interpretability, and new monitoring approaches have been proposed [32, 33, 65]. Among all the stages in the ML lifecycle, one of the most commonly overlooked although relevant is model monitoring. Recently, cloud providers have presented their own tools to use within their platforms [4, 61] while work is ongoing to integrate existent frameworks [72] into open-source model serving solutions [38]. Most of these frameworks are either built as an extension of an existent platform (i.e lack portability), follow a scheduled batch processing approach at a minimum rate of hours, or present limitations for certain outliers and drift algorithms due to the platform architecture design in which they are integrated. In this work, a scalable automated cloudnative architecture is designed and evaluated for ML model monitoring in a streaming approach. An experimentation conducted on a 7-node cluster with 250.000 requests at different concurrency rates shows maximum latencies of 5.9, 29.92 and 30.86 seconds after request time for 75% of distance-based outliers detection, windowed statistics and distribution-based data drift detection, respectively, using windows of 15 seconds length and 6 seconds of watermark delay.Under de senaste åren har konceptet MLOps blivit alltmer populärt på grund av tillkomsten av mer sofistikerade verktyg för explorativ dataanalys, datahantering, modell-träning och model serving som tjänstgör i produktion. Som ett försök att föra DevOps processer till Machine Learning (ML)-livscykeln, siktar MLOps på mer automatisering i utförandet av mångfaldiga och repetitiva uppgifter längs cykeln samt på smidigare interoperabilitet mellan team och verktyg inblandade. I det här sammanhanget har de största molnleverantörerna byggt sina egna ML-plattformar [4, 34, 61], vilka erbjuds som tjänster i deras molnlösningar. Dessutom har flera ramar tagits fram för att lösa konkreta problem såsom datatestning, datamärkning, distribuerad träning eller tolkning av förutsägelse, och nya övervakningsmetoder har föreslagits [32, 33, 65]. Av alla stadier i ML-livscykeln förbises ofta modellövervakning trots att det är relevant. På senare tid har molnleverantörer presenterat sina egna verktyg att kunna användas inom sina plattformar [4, 61] medan arbetet pågår för att integrera befintliga ramverk [72] med lösningar för modellplatformer med öppen källkod [38]. De flesta av dessa ramverk är antingen byggda som ett tillägg till en befintlig plattform (dvs. saknar portabilitet), följer en schemalagd batchbearbetningsmetod med en lägsta hastighet av ett antal timmar, eller innebär begränsningar för vissa extremvärden och drivalgoritmer på grund av plattformsarkitekturens design där de är integrerade. I det här arbetet utformas och utvärderas en skalbar automatiserad molnbaserad arkitektur för MLmodellövervakning i en streaming-metod. Ett experiment som utförts på ett 7nodskluster med 250.000 förfrågningar vid olika samtidigheter visar maximala latenser på 5,9, 29,92 respektive 30,86 sekunder efter tid för förfrågningen för 75% av avståndsbaserad detektering av extremvärden, windowed statistics och distributionsbaserad datadriftdetektering, med hjälp av windows med 15 sekunders längd och 6 sekunders fördröjning av vattenstämpel.
22
Kothawade, Rohan Dilip. "Wine quality prediction model using machine learning techniques." Thesis, Högskolan i Skövde, Institutionen för informationsteknologi, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:his:diva-20009.
Full textAbstract:
The quality of a wine is important for the consumers as well as the wine industry. The traditional (expert) way of measuring wine quality is time-consuming. Nowadays, machine learning models are important tools to replace human tasks. In this case, there are several features to predict the wine quality but the entire features will not be relevant for better prediction. So, our thesis work is focusing on what wine features are important to get the promising result. For the purposeof classification model and evaluation of the relevant features, we used three algorithms namely support vector machine (SVM), naïve Bayes (NB), and artificial neural network (ANN). In this study, we used two wine quality datasets red wine and white wine. To evaluate the feature importance we used the Pearson coefficient correlation and performance measurement matrices such as accuracy, recall, precision, and f1 score for comparison of the machine learning algorithm. A grid search algorithm was applied to improve the model accuracy. Finally, we achieved the artificial neural network (ANN) algorithm has better prediction results than the Support Vector Machine (SVM) algorithm and the Naïve Bayes (NB) algorithm for both red wine and white wine datasets.
23
Chida, Anjum A. "Protein Tertiary Model Assessment Using Granular Machine Learning Techniques." Digital Archive @ GSU, 2012. http://digitalarchive.gsu.edu/cs_diss/65.
Full textAbstract:
The automatic prediction of protein three dimensional structures from its amino acid sequence has become one of the most important and researched fields in bioinformatics. As models are not experimental structures determined with known accuracy but rather with prediction it’s vital to determine estimates of models quality. We attempt to solve this problem using machine learning techniques and information from both the sequence and structure of the protein. The goal is to generate a machine that understands structures from PDB and when given a new model, predicts whether it belongs to the same class as the PDB structures (correct or incorrect protein models). Different subsets of PDB (protein data bank) are considered for evaluating the prediction potential of the machine learning methods. Here we show two such machines, one using SVM (support vector machines) and another using fuzzy decision trees (FDT). First using a preliminary encoding style SVM could get around 70% in protein model quality assessment accuracy, and improved Fuzzy Decision Tree (IFDT) could reach above 80% accuracy. For the purpose of reducing computational overhead multiprocessor environment and basic feature selection method is used in machine learning algorithm using SVM.
Next an enhanced scheme is introduced using new encoding style. In the new style, information like amino acid substitution matrix, polarity, secondary structure information and relative distance between alpha carbon atoms etc is collected through spatial traversing of the 3D structure to form training vectors. This guarantees that the properties of alpha carbon atoms that are close together in 3D space and thus interacting are used in vector formation. With the use of fuzzy decision tree, we obtained a training accuracy around 90%. There is significant improvement compared to previous encoding technique in prediction accuracy and execution time. This outcome motivates to continue to explore effective machine learning algorithms for accurate protein model quality assessment.
Finally these machines are tested using CASP8 and CASP9 templates and compared with other CASP competitors, with promising results. We further discuss the importance of model quality assessment and other information from proteins that could be considered for the same.
24
Lee, Wei-En. "Visualizations for model tracking and predictions in machine learning." Thesis, Massachusetts Institute of Technology, 2017. http://hdl.handle.net/1721.1/113133.
Full textAbstract:
Thesis: M. Eng., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2017.This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.
Cataloged from student-submitted PDF version of thesis.
Includes bibliographical references (pages 82-84).
Building machine learning models is often an exploratory and iterative process. A data scientist frequently builds and trains hundreds of models with different parameters and feature sets in order to find one that meets the desired criteria. However, it can be difficult to keep track of all the parameters and metadata that are associated with the models. ModelDB, an end-to-end system for managing machine learning models, is a tool that solves this problem of model management. In this thesis, we present a graphical user interface for ModelDB, along with an extension for visualizing model predictions. The core user interface for model management augments the ModelDB system, which previously consisted only of native client libraries and a backend. The interface provides new ways of exploring, visualizing, and analyzing model data through a web application. The prediction visualizations extend the core user interface by providing a novel prediction matrix that displays classifier outputs in order to convey model performance at the example level. We present the design and implementation of both the core user interface and the prediction visualizations, discussing at each step the motivations behind key features. We evaluate the prediction visualizations through a pilot user study, which produces preliminary feedback on the practicality and utility of the interface. The overall goal of this research is to provide a powerful, user-friendly interface that leverages the data stored in ModelDB to generate effective visualizations for analyzing and improving models.
by Wei-En Lee.
M. Eng.
25
Bagheri, Rajeoni Alireza. "ANALOG CIRCUIT SIZING USING MACHINE LEARNING BASED TRANSISTORCIRCUIT MODEL." University of Akron / OhioLINK, 2021. http://rave.ohiolink.edu/etdc/view?acc_num=akron1609428170125214.
Full text
26
Sharma, Sagar. "Towards Data and Model Confidentiality in Outsourced Machine Learning." Wright State University / OhioLINK, 2019. http://rave.ohiolink.edu/etdc/view?acc_num=wright1567529092809275.
Full text
27
Lanctot, J. Kevin (Joseph Kevin) Carleton University Dissertation Mathematics. "Discrete estimator algorithms: a mathematical model of machine learning." Ottawa, 1989.
Find full text
28
Kokkonen, H. (Henna). "Effects of data cleaning on machine learning model performance." Bachelor's thesis, University of Oulu, 2019. http://jultika.oulu.fi/Record/nbnfioulu-201911133081.
Full textAbstract:
Abstract. This thesis is focused on the preprocessing and challenges of a university student data set and how different levels of data preprocessing affect the performance of a prediction model both in general and in selected groups of interest. The data set comprises the students at the University of Oulu who were admitted to the Faculty of Information Technology and Electrical Engineering during years 2006–2015. This data set was cleaned at three different levels, which resulted in three differently processed data sets: one set is the original data set with only basic cleaning, the second has been cleaned out of the most obvious anomalies and the third has been systematically cleaned out of possible anomalies. Each of these data sets was used to build a Gradient Boosting Machine model that predicted the cumulative number of ECTS the students would achieve by the end of their second-year studies based on their first-year studies and the Matriculation Examination results. The effects of the cleaning on the model performance were examined by comparing the prediction accuracy and the information the models gave of the factors that might indicate a slow ECTS accumulation. The results showed that the prediction accuracy improved after each cleaning stage and the influences of the features altered significantly, becoming more reasonable.Datan siivouksen vaikutukset koneoppimismallin suorituskykyyn. Tiivistelmä. Tässä tutkielmassa keskitytään opiskelijadatan esikäsittelyyn ja haasteisiin sekä siihen, kuinka eritasoinen esikäsittely vaikuttaa ennustemallin suorituskykyyn sekä yleisesti että tietyissä kiinnostuksen kohteena olevissa ryhmissä. Opiskelijadata koostuu Oulun yliopiston Tieto- ja sähkötekniikan tiedekuntaan vuosina 2006–2015 valituista opiskelijoista. Tätä opiskelijadataa käsiteltiin kolmella eri tasolla, jolloin saatiin kolme eritasoisesti siivottua versiota alkuperäisestä datajoukosta. Ensimmäinen versio on alkuperäinen datajoukko, jolle on tehty vain perussiivous, toisessa versiossa datasta on poistettu vain ilmeisimmät poikkeavuudet ja kolmannessa versiossa datasta on systemaattisesti poistettu mahdolliset poikkeavuudet. Jokaisella datajoukolla opetettiin Gradient Boosting Machine koneoppismismalli ennustamaan opiskelijoiden opintopistekertymää toisen vuoden loppuun mennessä perustuen heidän ensimmäisen vuoden opintoihinsa ja ylioppilaskirjoitustensa tuloksiin. Datan eritasoisen siivouksen vaikutuksia mallin suorituskykyyn tutkittiin vertailemalla mallien ennustetarkkuutta sekä tietoa, jota mallit antoivat niistä tekijöistä, jotka voivat ennakoida hitaampaa opintopistekertymää. Tulokset osoittivat mallin ennustetarkkuuden parantuneen jokaisen käsittelytason jälkeen sekä mallin ennustajien vaikutusten muuttuneen järjellisemmiksi.
29
Badayos, Noah Garcia. "Machine Learning-Based Parameter Validation." Diss., Virginia Tech, 2014. http://hdl.handle.net/10919/47675.
Full textAbstract:
As power system grids continue to grow in order to support an increasing energy demand, the system's behavior accordingly evolves, continuing to challenge designs for maintaining security. It has become apparent in the past few years that, as much as discovering vulnerabilities in the power network, accurate simulations are very critical. This study explores a classification method for validating simulation models, using disturbance measurements from phasor measurement units (PMU). The technique used employs the Random Forest learning algorithm to find a correlation between specific model parameter changes, and the variations in the dynamic response. Also, the measurements used for building and evaluating the classifiers were characterized using Prony decomposition. The generator model, consisting of an exciter, governor, and its standard parameters have been validated using short circuit faults. Single-error classifiers were first tested, where the accuracies of the classifiers built using positive, negative, and zero sequence measurements were compared. The negative sequence measurements have consistently produced the best classifiers, with majority of the parameter classes attaining F-measure accuracies greater than 90%. A multiple-parameter error technique for validation has also been developed and tested on standard generator parameters. Only a few target parameter classes had good accuracies in the presence of multiple parameter errors, but the results were enough to permit a sequential process of validation, where elimination of a highly detectable error can improve the accuracy of suspect errors dependent on the former's removal, and continuing the procedure until all corrections are covered.Ph. D.
30
Abdurahiman, Vakulathil. "Towards inducing a simulation model description." Thesis, Brunel University, 1994. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.239138.
Full text
31
Geras, Krzysztof Jerzy. "Exploiting diversity for efficient machine learning." Thesis, University of Edinburgh, 2018. http://hdl.handle.net/1842/28839.
Full textAbstract:
A common practice for solving machine learning problems is currently to consider each problem in isolation, starting from scratch every time a new learning problem is encountered or a new model is proposed. This is a perfectly feasible solution when the problems are sufficiently easy or, if the problem is hard when a large amount of resources, both in terms of the training data and computation, are available. Although this naive approach has been the main focus of research in machine learning for a few decades and had a lot of success, it becomes infeasible if the problem is too hard in proportion to the available resources. When using a complex model in this naive approach, it is necessary to collect large data sets (if possible at all) to avoid overfitting and hence it is also necessary to use large computational resources to handle the increased amount of data, first during training to process a large data set and then also at test time to execute a complex model. An alternative to this strategy of treating each learning problem independently is to leverage related data sets and computation encapsulated in previously trained models. By doing that we can decrease the amount of data necessary to reach a satisfactory level of performance and, consequently, improve the accuracy achievable and decrease training time. Our attack on this problem is to exploit diversity - in the structure of the data set, in the features learnt and in the inductive biases of different neural network architectures. In the setting of learning from multiple sources we introduce multiple-source cross-validation, which gives an unbiased estimator of the test error when the data set is composed of data coming from multiple sources and the data at test time are coming from a new unseen source. We also propose new estimators of variance of the standard k-fold cross-validation and multiple-source cross-validation, which have lower bias than previously known ones. To improve unsupervised learning we introduce scheduled denoising autoencoders, which learn a more diverse set of features than the standard denoising auto-encoder. This is thanks to their training procedure, which starts with a high level of noise, when the network is learning coarse features and then the noise is lowered gradually, which allows the network to learn some more local features. A connection between this training procedure and curriculum learning is also drawn. We develop further the idea of learning a diverse representation by explicitly incorporating the goal of obtaining a diverse representation into the training objective. The proposed model, the composite denoising autoencoder, learns multiple subsets of features focused on modelling variations in the data set at different levels of granularity. Finally, we introduce the idea of model blending, a variant of model compression, in which the two models, the teacher and the student, are both strong models, but different in their inductive biases. As an example, we train convolutional networks using the guidance of bidirectional long short-term memory (LSTM) networks. This allows to train the convolutional neural network to be more accurate than the LSTM network at no extra cost at test time.
32
Stroulia, Eleni. "Failure-driven learning as model-based self-redesign." Diss., Georgia Institute of Technology, 1994. http://hdl.handle.net/1853/8291.
Full text
33
Caceres, Carlos Antonio. "Machine Learning Techniques for Gesture Recognition." Thesis, Virginia Tech, 2014. http://hdl.handle.net/10919/52556.
Full textAbstract:
Classification of human movement is a large field of interest to Human-Machine Interface researchers. The reason for this lies in the large emphasis humans place on gestures while communicating with each other and while interacting with machines. Such gestures can be digitized in a number of ways, including both passive methods, such as cameras, and active methods, such as wearable sensors. While passive methods might be the ideal, they are not always feasible, especially when dealing in unstructured environments. Instead, wearable sensors have gained interest as a method of gesture classification, especially in the upper limbs. Lower arm movements are made up of a combination of multiple electrical signals known as Motor Unit Action Potentials (MUAPs). These signals can be recorded from surface electrodes placed on the surface of the skin, and used for prosthetic control, sign language recognition, human machine interface, and a myriad of other applications.
In order to move a step closer to these goal applications, this thesis compares three different machine learning tools, which include Hidden Markov Models (HMMs), Support Vector Machines (SVMs), and Dynamic Time Warping (DTW), to recognize a number of different gestures classes. It further contrasts the applicability of these tools to noisy data in the form of the Ninapro dataset, a benchmarking tool put forth by a conglomerate of universities. Using this dataset as a basis, this work paves a path for the analysis required to optimize each of the three classifiers. Ultimately, care is taken to compare the three classifiers for their utility against noisy data, and a comparison is made against classification results put forth by other researchers in the field.
The outcome of this work is 90+ % recognition of individual gestures from the Ninapro dataset whilst using two of the three distinct classifiers. Comparison against previous works by other researchers shows these results to outperform all other thus far. Through further work with these tools, an end user might control a robotic or prosthetic arm, or translate sign language, or perhaps simply interact with a computer.Master of Science
34
Follett, Stephen James. "A computational model of learning in Go." Thesis, University of South Wales, 2001. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.343412.
Full text
35
Nath, Gourabmoy. "A Model of Situation Learning in Design." Thesis, The University of Sydney, 1999. https://hdl.handle.net/2123/25096.
Full textAbstract:
This research is a theoretical and technical exploration into the machine learning of a form of design process knowledge viz. heuristic associations between patterns of the design context and preferences on design decisions such that the learned knowledge could be profitably reused for the synthesis of appropriate designs. Appropriate
designs mean designs that satisfy all feasibility and most desirability constraints or criteria as laid down in design specifications. A design decision means an unit design transformation process. A design context, as defined in this thesis refers to a combination of some initial design information that is modeled under a representation, some new information that is generated by various transformation and inference processes that operate on the representation as well as some information about the process that operates on the representation. By a context pattern it is meant that there exists some combination of some of the variables that represents the above information, such that the possible values of some or all of the variables in the pattern are constrained by relationships with other variables in that pattern. It is such a context pattern to which is referred to, as a situation. Situation learning is the process of associating a situation with a preference on a design decision. A preference indicates whether a design decision is to be preferred over others, or worse compared to others or should be totally rejected.
36
Murray-Smith, Roderick. "A local model network approach to nonlinear modelling." Thesis, University of Strathclyde, 1994. http://oleg.lib.strath.ac.uk:80/R/?func=dbin-jump-full&object_id=27067.
Full textAbstract:
This thesis describes practical learning systems able to model unknown nonlinear dynamic processes from their observed input-output behaviour. Local Model Networks use a number of simple, locally accurate models to represent a globally complex process, and provide a powerful, flexible framework for the integration of different model structures and learning algorithms. A major difficulty with Local Model Nets is the optimisation of the model structure. A novel Multi-Resolution Constructive (MRC) structure identification algorithm for local model networks is developed. The algorithm gradually adds to the model structure by searching for 'complexity' at ever decreasing scales of 'locality'. Reliable error estimates are useful during development and use of models. New methods are described which use the local basis function structure to provide interpolated state-dependent estimates of model accuracy. Active learning methods which automatically construct a training set for a given Local Model structure are developed, letting the training set grow in step with the model structure - the learning system 'explores' its data set looking for useful information. Local Learning methods developed in this work are explicitly linked to the local nature of the basis functions and provide a more computationally efficient method, more interpretable models and, due to the poor conditioning of the parameter estimation problem, often lead to an improvement in generalisation, compared to global optimisation methods. Important side-effects of normalisation of the basis functions are examined. A new hierarchical extension of Local Model Nets is presented: the Learning Hierarchy of Models (LHM), where local models can be sub-networks, leading to a tree-like hierarchy of softly interpolated local models. Constructive model structure identification algorithms are described, and the advantages of hierarchical 'divide-and-conquer' methods for modelling, especially in high dimensional spaces are discussed. The structures and algorithms are illustrated using several synthetic examples of nonlinear multivariable systems (dynamic and static), and applied to real world examples. Two nonlinear dynamic applications are described: predicting the strip thickness in an aluminium rolling mill from observed process data, and modelling robot actuator nonlinearities from measured data. The Local Model Nets reliably constructed models which provided the best results to date on the Rolling Mill application.
37
Abdullah, Siti Norbaiti binti. "Machine learning approach for crude oil price prediction." Thesis, University of Manchester, 2014. https://www.research.manchester.ac.uk/portal/en/theses/machine-learning-approach-for-crude-oil-price-prediction(949fa2d5-1a4d-416a-8e7c-dd66da95398e).html.
Full textAbstract:
Crude oil prices impact the world economy and are thus of interest to economic experts and politicians. Oil price’s volatile behaviour, which has moulded today’s world economy, society and politics, has motivated and continues to excite researchers for further study. This volatile behaviour is predicted to prompt more new and interesting research challenges. In the present research, machine learning and computational intelligence utilising historical quantitative data, with the linguistic element of online news services, are used to predict crude oil prices via five different models: (1) the Hierarchical Conceptual (HC) model; (2) the Artificial Neural Network-Quantitative (ANN-Q) model; (3) the Linguistic model; (4) the Rule-based Expert model; and, finally, (5) the Hybridisation of Linguistic and Quantitative (LQ) model. First, to understand the behaviour of the crude oil price market, the HC model functions as a platform to retrieve information that explains the behaviour of the market. This is retrieved from Google News articles using the keyword “Crude oil price”. Through a systematic approach, price data are classified into categories that explain the crude oil price’s level of impact on the market. The price data classification distinguishes crucial behaviour information contained in the articles. These distinguished data features ranked hierarchically according to the level of impact and used as reference to discover the numeric data implemented in model (2). Model (2) is developed to validate the features retrieved in model (1). It introduces the Back Propagation Neural Network (BPNN) technique as an alternative to conventional techniques used for forecasting the crude oil market. The BPNN technique is proven in model (2) to have produced more accurate and competitive results. Likewise, the features retrieved from model (1) are also validated and proven to cause market volatility. In model (3), a more systematic approach is introduced to extract the features from the news corpus. This approach applies a content utilisation technique to news articles and mines news sentiments by applying a fuzzy grammar fragment extraction. To extract the features from the news articles systematically, a domain-customised ‘dictionary’ containing grammar definitions is built beforehand. These retrieved features are used as the linguistic data to predict the market’s behaviour with crude oil price. A decision tree is also produced from this model which hierarchically delineates the events (i.e., the market’s rules) that made the market volatile, and later resulted in the production of model (4). Then, model (5) is built to complement the linguistic character performed in model (3) from the numeric prediction model made in model (2). To conclude, the hybridisation of these two models and the integration of models (1) to (5) in this research imitates the execution of crude oil market’s regulators in calculating their risk of actions before executing a price hedge in the market, wherein risk calculation is based on the ‘facts’ (quantitative data) and ‘rumours’ (linguistic data) collected. The hybridisation of quantitative and linguistic data in this study has shown promising accuracy outcomes, evidenced by the optimum value of directional accuracy and the minimum value of errors obtained.
38
Viswanathan, Srinidhi. "ModelDB : tools for machine learning model management and prediction storage." Thesis, Massachusetts Institute of Technology, 2017. http://hdl.handle.net/1721.1/113540.
Full textAbstract:
Thesis: M. Eng., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2017.Cataloged from PDF version of thesis.
Includes bibliographical references (pages 99-100).
Building a machine learning model is often an iterative process. Data scientists train hundreds of models before finding a model that meets acceptable criteria. But tracking these models and remembering the insights obtained from them is an arduous task. In this thesis, we present two main systems for facilitating better tracking, analysis, and querying of scikit-learn machine learning models. First, we introduce our scikit-learn client for ModelDB, a novel end-to-end system for managing machine learning models. The client allows data scientists to easily track diverse scikit-learn workflows with minimal changes to their code. Then, we describe our extension to ModelDB, PredictionStore. While the ModelDB client enables users to track the different models they have run, PredictionStore creates a prediction matrix to tackle the remaining piece in the puzzle: facilitating better exploration and analysis of model performance. We implement a query API to assist in analyzing predictions and answering nuanced questions about models. We also implement a variety of algorithms to recommend particular models to ensemble utilizing the prediction matrix. We evaluate ModelDB and PredictionStore on different datasets and determine ModelDB successfully tracks scikit-learn models, and most complex model queries can be executed in a matter of seconds using our query API. In addition, the workflows demonstrate significant improvement in accuracy using the ensemble algorithms. The overall goal of this research is to provide a flexible framework for training scikit-learn models, storing their predictions/ models, and efficiently exploring and analyzing the results.
by Srinidhi Viswanathan.
M. Eng.
39
Adhikari, Bhisma. "Intelligent Simulink Modeling Assistance via Model Clones and Machine Learning." Miami University / OhioLINK, 2021. http://rave.ohiolink.edu/etdc/view?acc_num=miami1627040347560589.
Full text
40
Anam, Md Tahseen. "Evaluate Machine Learning Model to Better Understand Cutting in Wood." Thesis, Uppsala universitet, Institutionen för informationsteknologi, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-448713.
Full textAbstract:
Wood cutting properties for the chains of chainsaw is measured in the lab by analyzing the force, torque, consumed power and other aspects of the chain as it cuts through the wood log. One of the essential properties of the chains is the cutting efficiency which is the measured cutting surface per the power used for cutting per the time unit. These data are not available beforehand and therefore, cutting efficiency cannot be measured before performing the cut. Cutting efficiency is related to the relativehardness of the wood which means that it is affected by the existence of knots (hardstructure areas) and cracks (no material areas). The actual situation is that all the cuts with knots and cracks are eliminated and just the clean cuts are used, therefore estimating the relative wood hardness by identifying the knots and cracks beforehand can significantly help to automate the process of testing the chain properties, saving time and material and give a better understanding of cutting wood logs to improve chains quality.Many studies have been done to develop methods to analyze and measure different features of an end face. This thesis work is carried out to evaluate a machinelearning model to detect knots and cracks on end faces and to understand their impact on the average cutting efficiency. Mask R-CNN is widely used for instance segmentation and in this thesis work, Mask R-CNN is evaluated to detect and segment knots and cracks on an end face. Methods are also developed to estimatepith’s vertical position from the wood image and generate average cutting efficiency graph based on knot’s and crack’s percentage at each vertical position of wood image.
41
Zhou, Wei. "Analysing the Robustness of Semantic Segmentation for Autonomous Vehicles." Thesis, University of Sydney, 2020. https://hdl.handle.net/2123/22699.
Full textAbstract:
Intelligent systems require the capability to perceive and interact with the surrounding environment. Semantic segmentation, as a pixel-level classification task, is at the frontier of providing a human-like understanding to intelligent systems enabling them to view and understand the world as we do. Deep learning based semantic segmentation algorithms have shown considerable success for certain tasks in recent years. However, in real-world safety critical applications such as autonomous vehicles, there are still many complexities that restrict the use of this technology. My research topic is to bridge the gap between the theoretical development of semantic segmentation and real-world applications by addressing some of the key challenges. Firstly, it is important for autonomous vehicles to have a 360° situational awareness of its surroundings. Most existing semantic segmentation datasets focus only on forward-facing cameras with narrow field of view, and under limited environmental conditions. Models trained using these datasets do not adapt well to changes in the camera perspective, and variation in the environment. This problem can be solved by creating more hand-labelled data, though this is impractical given the cost and requirement to consider all possible environmental conditions. My research addresses this problem by exploring data augmentation techniques from a practical standpoint. By applying carefully selected data augmentation methods, the existing model performance can be greatly improved under varying illumination conditions and with different camera perspectives. Guaranteeing safety when operating an autonomous vehicle requires large validation datasets that cover all the expected operating conditions. A contribution of my research is the publication of a large-scale driving dataset covering over one year of collected weekly trips around the University of Sydney campus. This dataset covers 70 weeks driving data and includes significant variation in the seasons, illumination, traffic volumes and scene structures. It is a valuable contribution to the intelligent transportation community for training, testing and validating algorithms for autonomous vehicles. Since safety is the primary concern for intelligent vehicles, the third challenge addressed by my research is how to evaluate the robustness and reliability of semantic models used in intelligent systems. It is essential to determine when those semantic models are operating correctly, especially given significant changes in illumination, weather, and sensor mounting position. Semantic segmentation can have catastrophic failures if the model fails. Therefore, efficiently validating the performance of semantic models under various conditions is crucial. To address this problem, my research has incorporated a robust sensor modality, lidar, to automatically generate ground-truth labels for different driving scenarios. Then, semantic segmentation performance can be validated against these labels. Algorithms after this validation will be remarkably informative for vehicles to make driving decisions. After evaluating and quantifying the robustness of semantic segmentation, the future work could be to use previously automatically generated ground-truth data to further train semantic models. It will allow them to be continuously validated and improved for different scenarios without any hand labelling. This validate-improve pipeline is not restricted to autonomous vehicles but can be applied to any intelligent systems to eventually obtain a robust understanding of their surrounding environment.
42
Pouilly-Cathelain, Maxime. "Synthèse de correcteurs s’adaptant à des critères multiples de haut niveau par la commande prédictive et les réseaux de neurones." Electronic Thesis or Diss., université Paris-Saclay, 2020. http://www.theses.fr/2020UPASG019.
Full textAbstract:
Cette thèse porte sur la commande des systèmes non linéaires soumis à des contraintes non différentiables ou non convexes. L'objectif est de pouvoir réaliser une commande permettant de considérer tout type de contraintes évaluables en temps réel.Pour répondre à cet objectif, la commande prédictive a été utilisée en ajoutant des fonctions barrières à la fonction de coût. Un algorithme d'optimisation sans gradient a permis de résoudre ce problème d'optimisation. De plus, une formulation permettant de garantir la stabilité et la robustesse vis-à-vis de perturbations a été proposée dans le cadre des systèmes linéaires. La démonstration de la stabilité repose sur les ensembles invariants et la théorie de Lyapunov.Dans le cas des systèmes non linéaires, les réseaux de neurones dynamiques ont été utilisés comme modèle de prédiction pour la commande prédictive. L'apprentissage de ces réseaux ainsi que les observateurs non linéaires nécessaires à leur utilisation ont été étudiés. Enfin, notre étude s'est portée sur l'amélioration de la prédiction par réseaux de neurones en présence de perturbations.La méthode de synthèse de correcteurs présentée dans ces travaux a été appliquée à l’évitement d’obstacles par un véhicule autonomeThis PhD thesis deals with the control of nonlinear systems subject to nondifferentiable or nonconvex constraints. The objective is to design a control law considering any type of constraints that can be online evaluated.To achieve this goal, model predictive control has been used in addition to barrier functions included in the cost function. A gradient-free optimization algorithm has been used to solve this optimization problem. Besides, a cost function formulation has been proposed to ensure stability and robustness against disturbances for linear systems. The proof of stability is based on invariant sets and the Lyapunov theory.In the case of nonlinear systems, dynamic neural networks have been used as a predictor for model predictive control. Machine learning algorithms and the nonlinear observers required for the use of neural networks have been studied. Finally, our study has focused on improving neural network prediction in the presence of disturbances.The synthesis method presented in this work has been applied to obstacle avoidance by an autonomous vehicle
43
Darwiche, Aiman A. "Machine Learning Methods for Septic Shock Prediction." Diss., NSUWorks, 2018. https://nsuworks.nova.edu/gscis_etd/1051.
Full textAbstract:
Sepsis is an organ dysfunction life-threatening disease that is caused by a dysregulated body response to infection. Sepsis is difficult to detect at an early stage, and when not detected early, is difficult to treat and results in high mortality rates. Developing improved methods for identifying patients in high risk of suffering septic shock has been the focus of much research in recent years. Building on this body of literature, this dissertation develops an improved method for septic shock prediction. Using the data from the MMIC-III database, an ensemble classifier is trained to identify high-risk patients. A robust prediction model is built by obtaining a risk score from fitting the Cox Hazard model on multiple input features. The score is added to the list of features and the Random Forest ensemble classifier is trained to produce the model. The Cox Enhanced Random Forest (CERF) proposed method is evaluated by comparing its predictive accuracy to those of extant methods.
44
Chapala, Usha Kiran, and Sridhar Peteti. "Continuous Video Quality of Experience Modelling using Machine Learning Model Trees." Thesis, Blekinge Tekniska Högskola, Institutionen för datavetenskap, 1996. http://urn.kb.se/resolve?urn=urn:nbn:se:bth-17814.
Full textAbstract:
Adaptive video streaming is perpetually influenced by unpredictable network conditions, whichcauses playback interruptions like stalling, rebuffering and video bit rate fluctuations. Thisleads to potential degradation of end-user Quality of Experience (QoE) and may make userchurn from the service. Video QoE modelling that precisely predicts the end users QoE underthese unstable conditions is taken into consideration quickly. The root cause analysis for thesedegradations is required for the service provider. These sudden changes in trend are not visiblefrom monitoring the data from the underlying network service. Thus, this is challenging toknow this change and model the instantaneous QoE. For this modelling continuous time, QoEratings are taken into consideration rather than the overall end QoE rating per video. To reducethe user risk of churning the network providers should give the best quality to the users. In this thesis, we proposed the QoE modelling to analyze the user reactions change over timeusing machine learning models. The machine learning models are used to predict the QoEratings and change patterns in ratings. We test the model on video Quality dataset availablepublicly which contains the user subjective QoE ratings for the network distortions. M5P modeltree algorithm is used for the prediction of user ratings over time. M5P model gives themathematical equations and leads to more insights by given equations. Results of the algorithmshow that model tree is a good approach for the prediction of the continuous QoE and to detectchange points of ratings. It is shown that to which extent these algorithms are used to estimatechanges. The analysis of model provides valuable insights by analyzing exponential transitionsbetween different level of predicted ratings. The outcome provided by the analysis explains theuser behavior when the quality decreases the user ratings decrease faster than the increase inquality with time. The earlier work on the exponential transitions of instantaneous QoE overtime is supported by the model tree to the user reaction to sudden changes such as video freezes.
45
Wu, Michael (Michael Q. ). "The synthetic student : a machine learning model to simulate MOOC data." Thesis, Massachusetts Institute of Technology, 2015. http://hdl.handle.net/1721.1/100681.
Full textAbstract:
Thesis: M. Eng., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2015.Cataloged from PDF version of thesis.
Includes bibliographical references (page 103).
It's now possible to take all of your favorite courses online. With growing popularity, Massive Open Online Courses (MOOCs) offer a learning opportunity to anyone with a computer - as well as an opportunity for researchers to investigate student learning through the accumulation of data about student-course interactions. Unfortunately, efforts to mine student data for information are currently limited by privacy concerns over how the data can be distributed. In this thesis, we present a generative model that learns from student data at the click-by-click level. When fully trained, this model is able to generate synthetic student data at the click-by-click level that can be released to the public. To develop a model at such granularity, we had to learn problem submission tendencies, characterize time spent viewing webpages and problem submission grades, and analyze how student activity transitions from week to week. We further developed a novel multi-level time-series model that goes beyond the classic Markov model and HMM methods used by most state-of-the art ML methods for weblogs, and showed that our model performs better than these methods. After training our model on a 6.002x course on edX, we generated synthetic data and found that a classifier that predicts student dropout is 93% as effective (by AUC) when trained on the simulated data as when trained on the real data. Lastly, we found that using features learned by our model improves dropout prediction performance by 9.5%.
by Michael Wu.
M. Eng.
46
Shen, Yingzhen. "Forecasting Twitter topic popularity using bass diffusion model and machine learning." Thesis, Massachusetts Institute of Technology, 2015. http://hdl.handle.net/1721.1/99575.
Full textAbstract:
Thesis: S.M. in Transportation, Massachusetts Institute of Technology, Department of Civil and Environmental Engineering, 2015.Cataloged from PDF version of thesis.
Includes bibliographical references (pages 91-93).
Today social network websites like Twitter are important information sources for a company's marketing, logistics and supply chain. Sometimes a topic about a product will "explode" at a "peak day," suddenly being talked about by a large number of users. Predicting the diffusion process of a Twitter topic is meaningful for a company to forecast demand, and plan ahead to dispatch its products. In this study, we collected Twitter data on 220 topics, covering a wide range of fields. And we created 12 features for each topic at each time stage, e.g. number of tweets mentioning this topic per hour, number of followers of users already mentioning this topic, and percentage of root tweets among all tweets. The task in this study is to predict the total mention count within the whole time horizon, 180 days, as early and accurately as possible. To complete this task, we applied two models - fitting the curve denoting topic popularity (mention count curve) by Bass diffusion model; and using machine learning models including K-nearest-neighbor, linear regression, bagged tree, and ensemble to learn the topic popularity as a function of the features we created. The results of this study reveal that the Basic Bass model captures the underlying mechanism of the Twitter topic development process. And we can analogue Twitter topics' adoption to a new product's diffusion. Using only mention count, over the whole time horizon, the Bass model has much better predictive accuracy, compared to machine learning models with extra features. However, even with the best model (the Bass model) and focusing on the subset of topics with better predictability, predictive accuracy is still not good enough before the "explosion day." This is because "explosion" is usually triggered by news outside Twitter, and therefore is hard to predict without information outside Twitter.
by Yingzhen Shen.
S.M. in Transportation
47
Essaidi, Moez. "Model-Driven Data Warehouse and its Automation Using Machine Learning Techniques." Paris 13, 2013. http://scbd-sto.univ-paris13.fr/secure/edgalilee_th_2013_essaidi.pdf.
Full textAbstract:
L'objectif de ce travail de thèse est de proposer une approche permettant l'automatisation complète du processus de transformation de modèles pour le développement d'entrepôts de données. L'idée principale est de réduire au mieux l'intervention des experts humains en utilisant les traces de transformations réalisées sur des projets similaires. L'objectif est d'utiliser des techniques d'apprentissage supervisées pour traiter les définitions de concepts avec le même niveau d'expression que les données manipulées. La nature des données manipulées nous a conduits à choisir les langages relationnels pour la description des exemples et des hypothèses. Ces langages ont l'avantage d'être expressifs en donnant la possibilité d'exprimer les relations entres les objets manipulés mais présente l'inconvénient majeur de ne pas disposer d'algorithmes permettant le passage à l'échelle pour des applications industrielles. Pour résoudre ce problème, nous avons proposé une architecture permettant d'exploiter au mieux les connaissances issues des invariants de transformations entre modèles et métamodèles. Cette manière de procéder a mis en lumière des dépendances entre les concepts à apprendre et nous a conduits à proposer un paradigme d'apprentissage dit de concepts-dépendants. Enfin, cette thèse présente plusieurs aspects qui peuvent influencer la prochaine génération de plates-formes décisionnelles. Elle propose, en particulier, une architecture de déploiement pour la business intelligence en tant que service basée sur les normes industrielles et les technologies les plus récentes et les plus prometteusesThis thesis aims at proposing an end-to-end approach which allows the automation of the process of model transformations for the development of data warehousing components. The main idea is to reduce as much as possible the intervention of human experts by using once again the traces of transformations produced on similar projects. The goal is to use supervised learning techniques to handle concept definitions with the same expressive level as manipulated data. The nature of the manipulated data leads us to choose relational languages for the description of examples and hypothesises. These languages have the advantage of being expressive by giving the possibility to express relationships between the manipulated objects, but they have the major disadvantage of not having algorithms allowing the application on large scales of industrial applications. To solve this problem, we have proposed an architecture that allows the perfect exploitation of the knowledge obtained from transformations' invariants between models and metamodels. This way of proceeding has highlighted the dependencies between the concepts to learn and has led us to propose a learning paradigm, called dependent-concept learning. Finally, this thesis presents various aspects that may inuence the next generation of data warehousing platforms. The latter suggests, in particular, an architecture for business intelligence-as-a-service based on the most recent and promising industrial standards and technologies
48
Oskarsson, Emma. "Machine Learning Model for Predicting the Repayment Rate of Loan Takers." Thesis, Umeå universitet, Institutionen för fysik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:umu:diva-184154.
Full textAbstract:
Machine Learning (ML) uses statistics to find patterns in high dimensional data. The Swedish Board of Student Finance (CSN) wants to improve the way they classify new loan takers. Using Machine Learning (ML) on data from previous loan takers can establish patterns to use on new loan takers. The aim of this study is to investigate if CSN can improve the way they classify loan takers by their ability to pay back their loan. In this study, different ML models are applied to a data set from CSN, their performance are compared and investigated by the most related factors affecting an individuals repayment rate. A data set of a total of 2032095 individuals were analysed and used in the different models. Using Random Forest (RF) for binary classification produced the best result with a sensitivity of 0.9695 and a specificity of 0.8058.
49
Qader, Aso, and William Shiver. "Developing an Advanced Internal Ratings-Based Model by Applying Machine Learning." Thesis, KTH, Matematisk statistik, 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-273418.
Full textAbstract:
Since the regulatory framework Basel II was implemented in 2007, banks have been allowed to develop internal risk models for quantifying the capital requirement. By using data on retail non-performing loans from Hoist Finance, the thesis assesses the Advanced Internal Ratings-Based approach. In particular, it focuses on how banks active in the non-performing loan industry, can risk-classify their loans despite limited data availability of the debtors. Moreover, the thesis analyses the effect of the maximum-recovery period on the capital requirement. In short, a comparison of five different mathematical models based on prior research in the field, revealed that the loans may be modelled by a two-step tree model with binary logistic regression and zero-inflated beta-regression, resulting in a maximum-recovery period of eight years. Still it is necessary to recognize the difficulty in distinguishing between low- and high-risk customers by primarily assessing rudimentary data about the borrowers. Recommended future amendments to the analysis in further research would be to include macroeconomic variables to better capture the effect of economic downturns.Sedan det regulatoriska ramverket Basel II implementerades 2007, har banker tillåtits utveckla interna riskmodeller för att beräkna kapitalkravet. Genom att använda data på fallerade konsumentlån från Hoist Finance, utvärderar uppsatsen den avancerade interna riskklassificeringsmodellen. I synnerhet fokuserar arbetet på hur banker aktiva inom sektorn för fallerade lån, kan riskklassificera sina lån trots begränsad datatillgång om låntagarna. Dessutom analyseras effekten av maximala inkassoperioden på kapitalkravet. I sammandrag visade en jämförelse av fem modeller, baserade på tidigare forskning inom området, att lånen kan modelleras genom en tvåstegs trädmodell med logistisk regression samt s.k. zero-inflated beta regression, resulterande i en maximal inkassoperiod om åtta år. Samtidigt är det värt att notera svårigheten i att skilja mellan låg- och högriskslåntagare genom att huvudsakligen analysera elementär data om låntagarna. Rekommenderade tillägg till analysen i fortsatt forskning är att inkludera makroekonomiska variabler för att bättre inkorporera effekten av ekonomiska nedgångar.
50
Ferrer, Martínez Claudia. "Machine Learning for Solar Energy Prediction." Thesis, Högskolan i Gävle, Avdelningen för elektronik, matematik och naturvetenskap, 2018. http://urn.kb.se/resolve?urn=urn:nbn:se:hig:diva-27423.
Full textAbstract:
This thesis consists of the study of different Machine Learning models used to predict solar power data in photovoltaic plants. The process of implement a model of Machine Learning will be reviewed step by step: to collect the data, to pre-process the data in order to make it able to use as input for the model, to divide the data into training data and testing data, to train the Machine Learning algorithm with the training data, to evaluate the algorithm with the testing data, and to make the necessary changes to achieve the best results. The thesis will start with a brief introduction to solar energy in one part, and an introduction to Machine Learning in another part. The theory of different models and algorithms of supervised learning will be reviewed, such as Decision Trees, Naïve Bayer Classification, Support Vector Machines (SVM), K-Nearest Neighbor (KNN), Linear Regression, Logistic Regression, Artificial Neural Network (ANN). Then, the methods Linear Regression, SVM Regression and Artificial Neural Network will be implemented using MATLAB in order to predict solar energy from historical data of photovoltaic plants. The data used to train and test the models is extracted from the National Renewable Energy Laboratory (NREL), that provides a dataset called “Solar Power Data for Integration Studies” intended for use by Project developers and university researchers. The dataset consist of 1 year of hourly power data for approximately 6000 simulated PV plants throughout the United States. Finally, once the different models have been implemented, the results show that the technique which provide the best results is Linear Regression.