Rozprawy doktorskie na temat „Online learning”

The research that constitutes this thesis was driven by the two related goals in mind. The first one was to develop new efficient online learning algorithms and to study their properties and theoretical guarantees. The second one was to study real-world data and find algorithms appropriate for the particular real-world problems. This thesis studies online prediction with few assumptions about the nature of the data. This is important for real-world applications of machine learning as complex assumptions about the data are rarely justified. We consider two frameworks: conformal prediction, which is based on the randomness assumption, and prediction with expert advice, where no assumptions about the data are made at all. Conformal predictors are set predictors, that is a set of possible labels is issued by Learner at each trial. After the prediction is made the real label is revealed and Learner's prediction is evaluated. 10 case of classification the label space is finite so Learner makes an error if the true label is not in the set produced by Learner. Conformal prediction was originally developed for the supervised learning task and was proved to be valid in the sense of making errors with a prespecified probability. We will study possible ways of extending this approach to the semi-supervised case and build a valid algorithm for this t ask. Also, we will apply conformal prediction technique to the problem of diagnosing tuberculosis in cattle. Whereas conformal prediction relies on just the randomness assumption, prediction with expert advice drops this one as well. One may wonder whether it is possible to make good predictions under these circumstances. However Learner is provided with predictions of a certain class of experts (or prediction strategies) and may base his prediction on them. The goal then is to perform not much worse than the best strategy in the class. This is achieved by carefully mixing (aggregating) predictions of the base experts. However, often the nature of data changes over time, such that there is a region where one expert is good, followed by a region where another is good and so on. This leads to the algorithms which we call adaptive: they take into account this structure of the data. We explore the possibilities offered by the framework of specialist experts to build adaptive algorithms. This line of thought allows us then to provide an intuitive explanation for the mysterious Mixing Past Posteriors algorithm and build a new algorithm with sharp bounds for Online Multitask Learning.

As technology continues to advance, many instructors are incorporating online activities into their courses. While online learning has several benefits, there is still debate on how instructors can best develop and utilize these resources in their classroom. This study is split into two smaller projects that both aim to provide further insights on how to develop online activities that target undergraduate biology students. The first project revealed that elaborative feedback in a phylogenetic activity was more useful for students who had some exposure to phylogenetics prior to completing the activity. The results of the second project revealed that the appearance of two simulations’ user interfaces does not have a significant effect on learning outcomes. However, many students responded that these simulations did increase their understanding of the concepts, indicating simulations can play an important role in the biology classroom.

Online learning algorithms have several key advantages compared to their batch learning algorithm counterparts: they are generally more memory efficient, and computationally mor efficient; they are simpler to implement; and they are able to adapt to changes where the learning model is time varying. Online algorithms because of their simplicity are very appealing to practitioners. his thesis investigates several online learning algorithms and their application. The thesis has an underlying theme of the idea of combining several simple algorithms to give better performance. In this thesis we investigate: combining weights, combining hypothesis, and (sort of) hierarchical combining.¶ Firstly, we propose a new online variant of the Bayes point machine (BPM), called the online Bayes point machine (OBPM). We study the theoretical and empirical performance of the OBPm algorithm. We show that the empirical performance of the OBPM algorithm is comparable with other large margin classifier methods such as the approximately large margin algorithm (ALMA) and methods which maximise the margin explicitly, like the support vector machine (SVM). The OBPM algorithm when used with a parallel architecture offers potential computational savings compared to ALMA. We compare the test error performance of the OBPM algorithm with other online algorithms: the Perceptron, the voted-Perceptron, and Bagging. We demonstrate that the combinationof the voted-Perceptron algorithm and the OBPM algorithm, called voted-OBPM algorithm has better test error performance than the voted-Perceptron and Bagging algorithms. We investigate the use of various online voting methods against the problem of ranking, and the problem of collaborative filtering of instances. We look at the application of online Bagging and OBPM algorithms to the telecommunications problem of channel equalization. We show that both online methods were successful at reducing the effect on the test error of label flipping and additive noise.¶ Secondly, we introduce a new mixture of experts algorithm, the fixed-share hierarchy (FSH) algorithm. The FSH algorithm is able to track the mixture of experts when the switching rate between the best experts may not be constant. We study the theoretical aspects of the FSH and the practical application of it to adaptive equalization. Using simulations we show that the FSH algorithm is able to track the best expert, or mixture of experts, in both the case where the switching rate is constant and the case where the switching rate is time varying.

In this thesis, I aim to advance the theories of online non-linear subspace learning through the development of strategies which are both efficient and robust. The use of subspace learning methods is very popular in computer vision and they have been employed to numerous tasks. With the increasing need for real-time applications, the formulation of online (i.e. incremental and real-time) learning methods is a vibrant research field and has received much attention from the research community. A major advantage of incremental systems is that they update the hypothesis during execution, thus allowing for the incorporation of the real data seen in the testing phase. Tracking acts as an attractive and popular evaluation tool for incremental systems, and thus, the connection between online learning and adaptive tracking is seen commonly in the literature. The proposed system in this thesis facilitates learning from noisy input data, e.g. caused by occlusions, casted shadows and pose variations, that are challenging problems in general tracking frameworks. First, a fast and robust alternative to standard L2-norm principal component analysis (PCA) is introduced, which I coin Euler PCA (e-PCA). The formulation of e-PCA is based on robust, non-linear kernel PCA (KPCA) with a cosine-based kernel function that is expressed via an explicit feature space. When applied to tracking, face reconstruction and background modeling, promising results are achieved. In the second part, the problem of matching vectors of 3D rotations is explicitly targeted. A novel distance which is robust for 3D rotations is introduced, and formulated as a kernel function. The kernel leads to a new representation of 3D rotations, the full-angle quaternion (FAQ) representation. Finally, I propose 3D object recognition from point clouds, and object tracking with color values using FAQs. A domain-specific kernel function designed for visual data is then presented. KPCA with Krein space kernels is introduced, as this kernel is indefinite, and an exact incremental learning framework for the new kernel is developed. In a tracker framework, the presented online learning outperforms the competitors in nine popular and challenging video sequences. In the final part, the generalized eigenvalue problem is studied. Specifically, incremental slow feature analysis (SFA) with indefinite kernels is proposed, and applied to temporal video segmentation and tracking with change detection. As online SFA allows for drift detection, further improvements are achieved in the evaluation of the tracking task.

Thesis: S.M., Massachusetts Institute of Technology, School of Architecture and Planning, Program in Media Arts and Sciences, 2014.
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 71-72).
This thesis explores how we can design online learning communities to better support connections to the people and resources beginners need when learning to program. I describe and analyze the design and implementation of the Scripts Workshop, a learning environment that supports members of the Scratch online community who are stuck on a programming problem in a Scratch project. The Scripts Workshop considers the People, Activities and Spaces needed to support these users in getting un-stuck. I conclude by describing a set of design principles for building learning webs within online communities, derived from the Scripts Workshop experiment.
by Champika Fernando.
S.M.

In tele-operated robotics applications, the primary information channel from the robot to its human operator is a video stream. For autonomous robotic systems however, a much larger selection of sensors is employed, although the most relevant information for the operation of the robot is still available in a single video stream. The issue lies in autonomously interpreting the visual data and extracting the relevant information, something humans and animals perform strikingly well. On the other hand, humans have great diculty expressing what they are actually looking for on a low level, suitable for direct implementation on a machine. For instance objects tend to be already detected when the visual information reaches the conscious mind, with almost no clues remaining regarding how the object was identied in the rst place. This became apparent already when Seymour Papert gathered a group of summer workers to solve the computer vision problem 48 years ago [35]. Articial learning systems can overcome this gap between the level of human visual reasoning and low-level machine vision processing. If a human teacher can provide examples of what to be extracted and if the learning system is able to extract the gist of these examples, the gap is bridged. There are however some special demands on a learning system for it to perform successfully in a visual context. First, low level visual input is often of high dimensionality such that the learning system needs to handle large inputs. Second, visual information is often ambiguous such that the learning system needs to be able to handle multi modal outputs, i.e. multiple hypotheses. Typically, the relations to be learned  are non-linear and there is an advantage if data can be processed at video rate, even after presenting many examples to the learning system. In general, there seems to be a lack of such methods. This thesis presents systems for learning perception-action mappings for robotic systems with visual input. A range of problems are discussed, such as vision based autonomous driving, inverse kinematics of a robotic manipulator and controlling a dynamical system. Operational systems demonstrating solutions to these problems are presented. Two dierent approaches for providing training data are explored, learning from demonstration (supervised learning) and explorative learning (self-supervised learning). A novel learning method fullling the stated demands is presented. The method, qHebb, is based on associative Hebbian learning on data in channel representation. Properties of the method are demonstrated on a vision-based autonomously driving vehicle, where the system learns to directly map low-level image features to control signals. After an initial training period, the system seamlessly continues autonomously. In a quantitative evaluation, the proposed online learning method performed comparably with state of the art batch learning methods.

Locally weighted regression is a non-parametric technique of regression that is capable of coping with non-stationarity of the input distribution. Online algorithms like Receptive FieldWeighted Regression and Locally Weighted Projection Regression use a sparse representation of the locally weighted model to approximate a target function, resulting in an efficient learning algorithm. However, these algorithms are fairly sensitive to parameter initializations and have multiple open learning parameters that are usually set using some insights of the problem and local heuristics. In this thesis, we attempt to alleviate these problems by using a probabilistic formulation of locally weighted regression followed by a principled Bayesian inference of the parameters. In the Randomly Varying Coefficient (RVC) model developed in this thesis, locally weighted regression is set up as an ensemble of regression experts that provide a local linear approximation to the target function. We train the individual experts independently and then combine their predictions using a Product of Experts formalism. Independent training of experts allows us to adapt the complexity of the regression model dynamically while learning in an online fashion. The local experts themselves are modeled using a hierarchical Bayesian probability distribution with Variational Bayesian Expectation Maximization steps to learn the posterior distributions over the parameters. The Bayesian modeling of the local experts leads to an inference procedure that is fairly insensitive to parameter initializations and avoids problems like overfitting. We further exploit the Bayesian inference procedure to derive efficient online update rules for the parameters. Learning in the regression setting is also extended to handle a classification task by making use of a logistic regression to model discrete class labels. The main contribution of the thesis is a spatially localised online learning algorithm set up in a probabilistic framework with principled Bayesian inference rule for the parameters of the model that learns local models completely independent of each other, uses only local information and adapts the local model complexity in a data driven fashion. This thesis, for the first time, brings together the computational efficiency and the adaptability of ‘non-competitive’ locally weighted learning schemes and the modelling guarantees of the Bayesian formulation.

The establishment of online discussion forums and their application to higher education have encouraged the use of online discussion within tertiary teaching. Recent studies related to online discussions have provided different ways of understanding the effect of online discussions on teaching and learning. This study investigates how personal learning is facilitated through various ways of engagement in an online discussion environment. The rationale behind this effort has been the concern that online discussions may be being used only because of the availability and technological opportunities the method provides. Personal learning is generally viewed in the literature as an individual's cognitive and knowledge construction and endeavour to make meaning through involvement and interaction in a community and context. There are, however, great variations in the way individuals engaged in their own learning within a community of learners. Motivation and strategies are also seen as factors that influence to individual level of engagement in online discussions. The findings reveal different types of interactions and highlight different levels of individual participation and engagement in the online discussions. From the findings, the Types of Online Interaction Model is developed to show the different roles that individual might adopt in the online discussion environment. The adopted roles are the individual approaches and actions that contribute to personal learning during the online discussion. The roles are flexible and individuals are likely to move from one role to another when there are reasons to do so. This study also shows the importance of the interactions that enable learning within the community. Two case studies discussed in this thesis illustrate the individual strategies of a provocateur and an eventual participant, which show how different ways of engaging in an online discussion community of learners contribute to individual learning.

This research set out to discover why some Higher Education (HE) students adapted very quickly to online environments and showed excellent learning behaviours and outcomes, while others found many barriers to the same activity. Given the rapid spread of virtual learning environments (VLEs) in HE Institutions, HE teachers need clear ideas about how to prepare and support learners in these environments. If individual differences among students could be identified, which affected “readiness” for learning online, then this information could be used to develop appropriate support and prevent such differences working to disadvantage groups of students. The project explored the perspectives of a group of HE teachers who could speak from experience as 'early adopters' of VLEs for pedagogic purposes, in order to discuss the 'readiness' of students for learning in an online context. Research questions focussed on how teachers could manage transition and integration of online technologies within HE, and how they could identify variations in students' approaches to the technologies and mediate the less successful ones. A grounded analysis method was applied to transcripts of interviews with HE teachers with experience and enthusiasm for integrating online and face-to-face teaching and learning. The 'constant comparative' method was used to fragment the data and develop categories of ideas in relation to the research questions. The findings confirmed differences between traditional and online teaching and learning, affecting the approach of both teacher and student, but gave no support to the concept of 'readiness'. Conclusions focussed on the process of preparing students for learning with online technologies. Further outcomes related to the changing teacher's role and the impact of teachers' beliefs on the design and integration of online technologies. Detailed suggestions were produced for appropriate learner induction to enable a more positive engagement with online technologies. The potential plasticity of the online learning space is shown to offer opportunities for supporting diverse learning approaches.

This thesis is dedicated to the study of online learning algorithms. In particular, after reviewing fundamental concepts in the theory of convex and online linear optimization, we provide a refined analysis of Implicit updates in the framework of Online Mirror Descent. We design a new adaptive algorithm based on it and carefully study its regret bound in the static case, linking it to the variability of the sequence of loss functions. Furthermore, we extend its application to the dynamic setting, studying its dynamic regret. In particular, we show that it achieves the optimal dynamic regret bound, when the quantities of interest are observable or known beforehand. On the other hand, in order to have a fully adaptive algorithm we show how to combine strongly adaptive algorithms with a simple greedy strategy. Finally, we focus on the well known problem of learning with expert advice. We review existing algorithm and describe an existing open problem. We provide some recent results and partial progress on how this problem could be addressed.

This dissertation explores how faculty development for online teaching in higher education might facilitate transformative learning and the transfer of instructional practices across teaching modalities. The first manuscript examines how the essential constructs of transformative learning are promoted in online faculty development and which elements of faculty development help to foster transformative learning. The second manuscript describes a case study that emerged from a university faculty development seminar to prepare instructors to teach online. The purpose of this study was to examine how, if at all, the Online Faculty Development Seminar changed five participants’ perspectives of teaching. This study found written reflection activities, combined with dialogue with colleagues, and having experienced instructors come in to tour their courses and discuss lessons learned contributed to perspective transformation. The third manuscript examines whether instructional practices introduced in the seminar would transfer to instructors’ in-person teaching and how faculty development and the experience of teaching online may have facilitated that transfer. The study found participants experienced perspective transformations that affected how they perceived their role as instructors, and they transferred some online course design and instructional practices to their in-person teaching. These practices included incorporating more digital tools to in-person courses, communicating clearly and transparently, designing courses with intentionality, and paying forward the lessons they learned to assist colleagues transitioning to teaching remotely in Spring 2020. Findings suggest that a structured course design process, self-reflection activities, opportunities to dialogue with colleagues, and course tours from colleagues aided in transfer of practices across modalities.

Thesis (Ed.D.)--Boston University
PLEASE NOTE: Boston University Libraries did not receive an Authorization To Manage form for this thesis or dissertation. It is therefore not openly accessible, though it may be available by request. If you are the author or principal advisor of this work and would like to request open access for it, please contact us at open-help@bu.edu. Thank you.
The purpose of this study was to determine whether a match between a participant's learning style and type of online instruction improved learner performance on tests measuring comprehension and retention. Learning style was measured by the Self-Directed Leamer Readiness Scale (SDLRS) and the Grasha-Riechmann Student Learning Style Scale (GRSLSS) and online instruction varied among online courses, recorded online courses, and computer-based tutorials. The setting for the study was a high tech machine vision company in Massachusetts and online users of its products were the participants. Three groups of learners participated in the study: employees, high school students, and customers. All three groups were comprised of engineers or engineering students. All 106 participants completed a survey that measured their preference for self-directed and collaborative learning style with the standard instruments SDLRS and GRSLSS. Participants completed 323 pre- and post-tests for 46 live online courses, recorded online courses, and computer-based tutorials during the data collection phase of the study. Those participants learning in their preferred learning style had the highest mean improvement from pre- to post-tests. Those participants with average or below average scores for self-directed and collaborative learning style showed the least improvement. The results of this study supported the hypothesis that matching the type of activity, collaborative or self-directed, to the learner's preferred learning style improved performance. The study included ten research questions.
2031-01-01

With the adoption of social technologies in mainstream society, post-secondary educators have been adopting social technologies as alternatives to traditional learning management systems, perceiving them to be more open, participatory, student-centered, and reflective of socio-constructivist approaches to learning. At the same time, as we open up our boundaries of learning, researchers have suggested that these spaces can be uncanny, unsettling or troublesome as they challenge traditional, hierarchical learning models and their more familiar and comfortable references, roles and norms of the academy. How we incorporate these networked learning principles into the design of open online learning spaces, and how these spaces then are enacted as learning spaces is the focus of this project. A virtual ethnographic case study of an open boundary course was conducted to investigate how the available learning spaces are perceived and used by both teachers and learners, particularly as they intersect formal and informal contexts. To answer the overall research question "What effects do open online learning spaces have on the development of a learning culture in networked learning environments?" a two-tiered analytic framework was developed. The first tier examined the everyday practices within the course, including interactions between material and social spaces, through examination of the structures, communications and resulting practices. The second stage used a spatial lens, based on Boys' (2011) adaptation of Lefebvre's spatial triad (1991), to explore the tensions between how space is perceived (daily practices), conceived (designed), and lived (enacted) by participants. One finding is that participants all valued direct pathways for their learning experiences and felt that too many resources and routes lead to confusion and disorientation. Finding and maintaining coherence was a challenge for both instructors and participants, with each wrangling with the principles of openness, autonomy and social dialogue to meet their own needs and create different learning spaces. For the instructors this meant providing wayfinding and mooring points through the signalling of pathways, active participation and a repurposing/remixing of the different tools and structures available to them. The designed environment was inscribed with the familiar indicators of formal educational spaces (timeframes, structured activities, roles, active facilitation, educational metaphors, familiar asynchronous and synchronous communication) and provided a "homely" feel (Knox, 2014b). The participants chose different pathways depending on their expectations and learning needs (assessed/non- assessed), made visible their struggles with technology, and stuck to the course "home" space where visibility, recognition and meaningful connections were more likely to be encountered. This allowed for the development of a small cohort of engaged, active learners who developed an open and supportive learning culture where they could take risks in their own learning processes. The spatial analysis highlighted that there is a constant shifting and renegotiating within the learning spaces we try to create, both as designers and as learners. In this case tensions related to visibility/anonymity, assessment, flexibility (pathways, time), resources, conceptions of openness, and complexity of the learning environment, all had an impact on how the learning spaces were perceived or enacted. The hierarchically defined spaces created through digital tools, even those created by social technologies that many consider inherently more open and participatory, are only permeable and accessible in certain ways, and to certain types of practices. As the results of this research highlight, these underlying structures, with their own set of rules, ownership, and hierarchical ordering affects the resulting spaces, dictating how learners and teachers can shape and interact with them. Those considering designing learning experiences with more open, permeable boundaries will need to ask critical questions about how resulting tensions may create different types of enclosures or barriers, and flow this impacts on the spaces for learning.

The quality of the student learning experience in an online learning course has raised many debates in educational studies. Evidence found in current literature indicates that individualised learning and interactive learning do contribute to the student learning experience in online learning courses. However, there is little evidence of any major studies that have tried to explore the impact of both individualised learning and interactive learning on the students' experience.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2006.
Includes bibliographical references (p. 99-102).
Many practical problems such as forecasting, real-time decision making, streaming data applications, and resource-constrained learning, can be modeled as learning with online constraints. This thesis is concerned with analyzing and designing algorithms for learning under the following online constraints: i) The algorithm has only sequential, or one-at-time, access to data. ii) The time and space complexity of the algorithm must not scale with the number of observations. We analyze learning with online constraints in a variety of settings, including active learning. The active learning model is applicable to any domain in which unlabeled data is easy to come by and there exists a (potentially difficult or expensive) mechanism by which to attain labels. First, we analyze a supervised learning framework in which no statistical assumptions are made about the sequence of observations, and algorithms are evaluated based on their regret, i.e. their relative prediction loss with respect to the hindsight-optimal algorithm in a comparator class. We derive a, lower bound on regret for a class of online learning algorithms designed to track shifting concepts in this framework. We apply an algorithm we provided in previous work, that avoids this lower bound, to an energy-management problem in wireless networks, and demonstrate this application in a network simulation.
(cont.) Second, we analyze a supervised learning framework in which the observations are assumed to be iid, and algorithms are compared by the number of prediction mistakes made in reaching a target generalization error. We provide a lower bound on mistakes for Perceptron, a standard online learning algorithm, for this framework. We introduce a modification to Perceptron and show that it avoids this lower bound, and in fact attains the optimal mistake-complexity for this setting. Third, we motivate and analyze an online active learning framework. The observations are assumed to be iid, and algorithms are judged by the number of label queries to reach a target generalization error. Our lower bound applies to the active learning setting as well, as a lower bound on labels for Perceptron paired with any active learning rule. We provide a new online active learning algorithm that avoids the lower bound, and we upper bound its label-complexity. The upper bound is optimal and also bounds the algorithm's total errors (labeled and unlabeled). We analyze the algorithm further, yielding a label-complexity bound under relaxed assumptions. Using optical character recognition data, we empirically compare the new algorithm to an online active learning algorithm with data-dependent performance guarantees, as well as to the combined variants of these two algorithms.
by Claire E. Monteleoni.
Ph.D.

Students in the United States are obtaining more college degrees than ever before. In 1975, 21.9% of Americans held bachelor’s degrees, and in 2012, 33.5% of Americans held bachelor’s degrees (Rampell, 2013). A study in 2011 indicated that Americans possessing a bachelor’s degree earn approximately $2.27 million, those with master’s degrees earn $2.67 million and those with doctoral degrees earn $3.65 million over their adult lifetime, dwarfing those with some college, who earn $1.55 million, or no college, who earn $1.30 million (Burnsed, 2011).

Unfortunately, the increase in college degree attainment in the United States does not include all Americans. Among low-income students, degree attainment has been fairly flat for several decades (Mortenson, 2016). Although education can be a great equalizer and opportunity generator, among lower income students it is often times an insurmountable challenge to obtain a bachelor’s or post baccalaureate degree. College students can have challenges in obtaining learning opportunities due to factors beyond their control, such as geography and access to quality instruction.

In order to provide equity and opportunity for nontraditional students who either working, have family responsibilities, or are low income or first generation college attenders, it is important to make every effort to connect these students with meaningful and attainable opportunities to obtain a college degree. One such delivery model of curriculum is online learning. Online learning in higher education—in which students are obtaining bachelors, masters, or doctoral degrees—takes place either partially or fully in a virtual environment accessible from e-learning devices such as laptops, tablets, or smartphones.

The goal of this study was a greater understanding the best practices in adult online education. The participants in the study were asked to help identify both the challenges and successes experienced in their online learning environments. While success in both teaching and learning is subjective, the data revealed a number of common themes, which indicated similar elements that lead to success in an online environment in areas of curriculum design, classroom management, and use of technology.

Online grooming is a major problem in today's society where more and more time is spent online. To become friends and establish a relationship with their young victims in online communities, groomers often pretend to be children. In this paper we describe an approach that can be used to detect if an adult is pretending to be a child in a chat room conversation. The approach involves a two step process wherein authors are first classified as being a children or adults, and then each child is being examined and false children distinguished from genuine children. Our results shows that even if it is hard to separate ordinary adults from children in chat logs it is possible to distinguish real children from adults pretending to be children with a high accuracy. In this report the accuracy of the methods proposed is discussed, as well as the features that were important in their success. We believe that this work is an important step towards automated analysis of chat room conversation to detect possible attempts of grooming. Our approach where we use text analysis to distinguish adults who are pretending to be children from actual children could be used to inform children about the true age of the person that they are communicating. This would be a step towards making the Internet more secure for young children and eliminate grooming.

With an ever increasing demand on large scale data, difficulties exist in terms of processing and utilising the information available. In particular, making decisions based upon sequentially acquired data where only limited information is initially known, is an important problem. Often the input data in such problems have a complex combinatorial structure, for example consider an internet advertising system that manages advertisement placement over a network of websites. The ways of placing m different advertisements on n websites with replacement, is an exponential number of mⁿ possible combinations that scales badly with large n. As a combinatorial problem, the data can be manipulated within a frequently occurring computational object called graph, allowing the structure to be exploited for intelligent automatic processing. Traditionally, machine learning techniques require a separate initial training phase before predictions can occur on unseen data. However, the sequential nature of some problems necessitate real-time prediction, thereby making many existing techniques unsuitable. Online learning is a field of machine learning that has an ensemble of algorithms that learn from sequential streaming data, where the learner cannot control or in influence the data collection procedure. Although these existing online methods have theoretical guarantees on performance, in the context of combinatorial complexity of graphical structures they are not yet fully matured. In this thesis, a series of algorithms that attempt to overcome the shortcomings of existing online algorithms are presented. The discrete graphical model, called the Ising model, is explored to develop online approximation algorithms for label prediction. A deterministic approximation algorithm with sequential guarantee is developed, by capturing the persistent structures of maximum flows and minimum cuts in the network and an efficient enumeration of all label consistent minimum cuts. Novel mistake bounds are provided that improve and match previous performance bounds in the literature. Additionally, a variational approximation technique using mean field approximation is built for online prediction of multi-class labelling on the Ising model. An online sequential action selection algorithm for the limited feedback setting (bandit feedback) and side information is developed with a linear programming relaxation of the classic maximal flow problem. Finally, the multiple objective optimization problem with conflicting objectives and full feedback is studied and an online algorithm is built that outperforms the traditional approaches under similar assumptions.

Thesis (S.M.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2003.
Includes bibliographical references (p. 47-48).
by Claire E. Monteleoni.
S.M.

My thesis is going to focus on the development of a standalone, web based, machine learning educational platform. This platform will have a specific focus on neural networks. This tool will have the primary intention to provide a theoretical background to the mathematics of neural networks and thereafter to allow users to train their own networks on regression problems of their own creation. This is so as to provide the user with both theoretical, and first-hand, experience in the applications and functions of artificial intelligence. The primary success metric of this project will be how informative it is to the user. The key deliverable will be a fully functional prototype in additional to a written piece inclusive of a literature review and any other relevant findings and conclusions.

Thesis: Ph. D., Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, September, 2020
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 213-219).
With the rapid advancement of information technology and accelerated development of data science, the importance of integrating data into decision-making has never been stronger. In this thesis, we propose data-driven algorithms to incorporate learning from data in three operations problems, concerning both online learning and offline learning settings. First, we study a single product pricing problem with demand censoring in an offline data-driven setting. In this problem, a retailer is given a finite level of inventory, and faces a random demand that is price sensitive in a linear fashion with unknown parameters and distribution. Any unsatisfied demand is lost and unobservable. The retailer's objective is to use offline censored demand data to find an optimal price, maximizing her expected revenue with finite inventories.
We characterize an exact condition for the identifiability of near-optimal algorithms, and propose a data-driven algorithm that guarantees near-optimality in the identifiable case and approaches best-achievable optimality gap in the unidentifiable case. Next, we study the classic multi-period joint pricing and inventory control problem in an offline data-driven setting. We assume the demand functions and noise distributions are unknown, and propose a data-driven approximation algorithm, which uses offline demand data to solve the joint pricing and inventory control problem. We establish a polynomial sample complexity bound, the number of data samples needed to guarantee a near-optimal profit. A simulation study suggests that the data-driven algorithm solves the dynamic program effectively. Finally, we study an online learning problem for product selection in urban warehouses managed by fast-delivery retailers. We distill the problem into a semi-bandit model with linear generalization.
There are n products, each with a feature vector of dimension T. In each of the T periods, a retailer selects K products to offer, where T is much greater than T or b. We propose an online learning algorithm that iteratively shrinks the upper confidence bounds within each period. Compared to the standard UCB algorithm, we prove the new algorithm reduces the most dominant regret term by a factor of d, and experiments on datasets from Alibaba Group suggest it lowers the total regret by at least 10%..
by Li Wang.
Ph. D.
Ph.D. Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center

The aim of this thesis is to develop online kernel based algorithms for learning clas sification functions over a graph. An important question in machine learning is: how to learn functions in a high dimension One of the benefits of using a graphical representation of data is that it can provide a dimensionality reduction of the data to the number of nodes plus edges in the graph. Graphs are useful discrete repre sentations of data that have already been used successfully to incorporate structural information in data to aid in semi-supervised learning techniques. In this thesis, an online learning framework is used to provide guarantees on performance of the algo rithms developed. The first step in developing these algorithms required motivating the idea of a "natural" kernel defined on a graph. This natural kernel turns out to be the Laplacian operator associated with the graph. The next step was to look at a well known online algorithm - the perceptron algorithm - with the associated bound, and formulate it for online learning with this kernel. This was a matter of using the Laplacian kernel with the kernel perceptron algorithm. For a binary classification problem, the bound on the performance of this algorithm can be interpreted in terms of natural properties of the graph, such as the graph diameter. Further algorithms were developed, motivated by the idea of a series of alternate projections, which also share this bound interpretation. The minimum norm interpolation algorithm was developed in batch mode and then transformed into an online algorithm. These al gorithms were tested and compared with other proposed algorithms on toy and real data sets. The main comparison algorithm used was k-nearest neighbour along the graph. Once the kernel has been calculated, the new algorithms perform well and offer some advantages over other approaches in terms of computational complexity.

At a local community college in Texas, student retention remained a concern as enrollment was increasing while online student retention was decreasing. The purpose of this study was to examine student retention in online courses at the college. The conceptual framework that guided the project study was Tinto’s integration model, which provided insight as to why students choose to leave or continue their educational journey. The overarching question that guided the study queried the factors influencing students’ decisions to take online courses at the higher education level. A qualitative case study was used to capture information on 10 students regarding their perceptions of online learning and retention issues within the programs. Interviews were used to collect the data, along with research notes from each 40 minute interview. All information was transcribed and member checked, the data and research notes were uploaded in Nvivo 11. Once analyzed the following themes emerged, personal, academic, and institutional. The results also revealed that student participation and belonging are key indicators of student performance online and seem to be the most significant reason for failure or withdrawal from online courses. To address the reasons, a professional development plan was developed for the local community college to increase student, faculty, and staff awareness, interaction, and to assist in creating a welcoming, learning, and supportive environment. The implications for social change include presenting the professional development to the local community college to increase student retention and success rates for online courses by understanding the student population and their needs to be successful, resulting in an increase for graduation.

Cette thèse se concentre principalement sur les problèmes d'appariement en ligne, où des ensembles de ressources sont alloués séquentiellement à des flux de demandes. Nous les traitons à la fois du point de vue de l'apprentissage en ligne et de l'analyse compétitive, toujours lorsqueEn ce qui concerne l'apprentissage en ligne, nous étudions comment la structure spécifique de l'appariement influence l'apprentissage dans la première partie, puis comment les effets de report dans le système affectent ses performances.En ce qui concerne l'analyse compétitive, nous étudions le problème de l'appariement en ligne dans des classes spécifiques de graphes aléatoires, dans un effort pour s'éloigner de l'analyse du pire cas.Enfin, nous explorons la manière dont l'apprentissage peut être exploité dans le problème d'ordonnancement des machines
This thesis focuses mainly on online matching problems, where sets of resources are sequentially allocated to demand streams. We treat them both from an online learning and a competitive analysis perspective, always in the case when the input is stochastic.On the online learning side, we study how the specific matching structure influences learning in the first part, then how carry-over effects in the system affect its performance.On the competitive analysis side, we study the online matching problem in specific classes of random graphs, in an effort to move away from worst-case analysis.Finally, we explore how learning can be leveraged in the scheduling problem

Despite the continued growth of K-12 online learning, there remains a need for additional research addressing roles of online facilitators and how they can improve the sense of community at K-12 online schools. The first article of this dissertation presents a case study illustrating how online facilitators can provide the same level of support for their students that on-site facilitators provide students in blended environments. Data was gathered from teachers at Mountain Heights Academy (MHA), a fully online high school. MHA implemented a "Shepherding Program" to provide student with online facilitators. Each teacher, or shepherd, was responsible for 20 to 25 students. Teacher focus groups and one-on-one interviews were used to examine the perceived effects of a shepherding program on shepherd-student relationships. Additionally, the teacher roles in the shepherding program were compared to the roles of on-site facilitators. Teachers were largely satisfied with the perceived impact of the shepherding program on their relationships with their students. Findings also highlighted strong similarities between the support the shepherding program provided online students and the support on-site facilitators provide blended learning students. The second article was a continuation of the case study from the first article. A key addition to the case study for the second article was the inclusion of student interviews. This article examined how teachers and students perceived that the shepherding program influenced instructor-student relationships. The analysis exposing similarities and differences between teacher and student perspectives of the shepherding program was conducted based on the four dimensions of Rovai's online sense of community: spirit, trust, interaction, and learning. Findings illustrated shepherd-student relationships consisting of all four elements of community in some degree.

Thesis (Ph.D.)--Indiana University, Dept. of Instructional Systems Technology of the School of Education, 2006.
Source: Dissertation Abstracts International, Volume: 67-01, Section: A, page: 0152. Adviser: Thomas M. Duffy. "Title from dissertation home page (viewed Jan. 8, 2007)."

Big Data is the certain result of our knowledge-intensive world when practically everything is being monitored and measured. According to the 2014 report of the International Data Corporation (IDC), the amount of information created, captured or replicated had exceeded available storage for the first time in 2007. The digital universe is doubling in size every two years and will multiply 10-fold between 2013 and 2020 – from 4.4 trillion gigabytes to 44 trillion gigabytes. Besides the massive volume, the velocity of data is another concern as in our information society, very often data come in the form of streams which continuously and rapidly grow over time. Examples of such data can be easily seen in many real-world applications like network traffic, sensor networks, web searches, stock market systems, social media and others. Mining big data can bring back big values and benefit humans in every aspect of life such as social communication, business management, and scientific research. They present a new world of opportunities as well as challenges that human beings need to deal with in a responsible way, maintaining adaptability, scalability and efficiency adequately. From the perspective of machine learning, the main approach is based on algorithmic improvement. Traditional offline machine learning techniques suffer from many restrictions such as the limitation of computational storage for saving the whole training set, and the impossibility of handling real-time data and responding instantly. To overcome those challenges, the advent of online methods offers the essential ability of predictive models which can be trained on-the-fly after the arrival of every new data point and be ready to give predictions at any time if requested, by making use of a single/set of observations and then discarding them permanently before the next observations are used. This typical one-pass-throw-away streamed learning requires online methods to preserve as much information extracted from the past instances as possible and at the same time must learn current instances effectively. They are also expected to better handle dynamically evolving environments with concept drifts (the change in data distribution), which are widely encountered in stream contexts. Furthermore, potentially extremely skewed classes in a large number of daily applications, such as accident diagnosis of real-time traffic surveillance, fault detection in online banking and intrusion monitoring, can significantly hinder the classification performance of online methods. There is a need for building more effective online frameworks, which offer rich information and high flexibility in adaption to concept changes, class imbalances and other advanced tasks (if needed) coming from real-life data stream mining. To address the research challenges mentioned above, we first proposed novel online supervised Bayesian classifiers based on variational inference (VI) for multivariate Gaussians (Minibatch-VIGO and VIGO), which outperformed recent and well-known methods on a wide range of datasets (reducing at least 5.5% mistake rate compared to the other benchmarks). From the same theoretical background, a lossless online classifier (OVIG) is presented, guaranteeing to produce the same prediction model as its offline counterpart regardless of the incremental training order. Through the application to movie genre classification, two strategies for dealing with highdimensional data were suggested including random projection based and stacking based ensembles. Exploiting the flexibility of a variational inference mechanism, we also developed advanced techniques to effectively tackle dynamic streaming data with concept drifts. They include VIGOd (online VI with a built-in concept drift detector for multivariate Gaussians) and VIGOw (online VI weighted for multivariate Gaussians), which are almost 20 times faster than the most accurate adaptive benchmark method. Next, we introduced OCSB (online cost-sensitive learning and sampling for Bayesian classifiers), a new imbalanced learning strategy that combines cost-sensitive learning and intermediate sampling. When applying OCSB to Minibatch-VIGO, it helps to increase the recognition of rare instances and the overall accuracy significantly (by over 25% and 16%, respectively). Furthermore, we generalize our online multi-class classifiers (each class has 1 label) to online multi-label classifiers (each class can have many labels) and our online VI-based multivariate Gaussians (VIGO) (data of each class described by a multivariate Gaussian) to online VI-based mixtures of Gaussians (VIMGO) (data of each class described by a number of Gaussians). For the former (multi-label learning), one more concept drift adaptation technique using a decay factor to weight the importance of instances according to their age is suggested and tested. At the same time, a new dynamic intermediate sampling technique (DIS) is developed to handle new challenges of online imbalanced learning for the multi-label scenarios. Although having to approximate a much bigger number of Gaussians in mixture models, the latter (VIMGO) and its adaptive version VIMGOw are optimized to run effectively. VIMGOw is combined with a simplified version of OCSB to obtain iVIMGOw. Applications of iVIMGOw in network intrusion and credit card fraud detection through recent UNSW-NB15 network data and real-world credit card data respectively show that it can accurately recognize attacks with a wide range of frequency. Finally, it is worth mentioning that all our proposed classifiers are second-order generative methods with rich information. They not only explore the underlying structure of data effectively but can also be used as a base framework for further advanced tasks of stream learning like cost-sensitive learning, active learning and semisupervised learning.
Thesis (PhD Doctorate)
Doctor of Philosophy (PhD)
School of Info & Comm Tech
Science, Environment, Engineering and Technology
Full Text

This project examines online learning tools and software that are applicable to K-12 and post-secondary distance learning environments. Powerpoint, Webquest and Inspiration 7 have been utilized to develop a lesson plan and storyboard that incorporate the constructivist theory of learning. An accompanying Web site, "Learning Tools for the On-Line Student," serves as a resource for instructional technology educators and includes information on designing lesson plans and evaluating students. The site also provides links to online tools and software that are useful in online learning environments.

The desire to translate constructivist and sociocultural approaches to learning in specific learning activities is evident in most forms of training at current, not least in online education. Teachers worldwide are struggling with questions of how to create conditions in this fairly new realm of education for learners to contribute to the development of a good quality in their own and others' learning. Collaboration in forms of text talk in asynchronous, text based forums (ADF) is often used so students can participate at the location and time that suits them best given the other aspects of their life situation. But previous research show how collaboration in forms of text talk do not always evolve in expected quality, and how participation sometimes can be so low that no discussions at all take place. Perhaps it is time to move on and make use of the variety of user-friendly audio-visible technologies that offers conditions for collaboration similar to those in the physical environment? Is there any point to use ADF for collaboration, beyond the flexible opportunity for participation it allows? If so, why, how and under what conditions are it worthwhile to use ADF for tasks meant to be worked collaboratively on? These questions were the starting point of the studies in this thesis that was researched through two case studies involving different techniques and data samples of various natures, with the aim to understand more about collaborative text talk. The research approach differs from the vast majority of studies in the research field of Computer Supported Collaborative Learning (CSCL) where many studies currently are conducted by analysis of quantifiable data. The first case study was conducted in the context of non-formal learning in Swedish Liberal Adult Education online, and the second in the context of higher education online in Sweden. The studies in the thesis were made on basis of socio-cultural theory and empirical studies. Empirical data was collected from questionnaires, interviews and texts created by students participating in tasks that they jointly resolved through text talk. Some results were brought back to the students for further explanation of the results. Findings from data analysis were triangulated with other results and with sociocultural theory. The results indicate that students can create knowledge relevant to their studies through text talk, but can feel restrained or dismiss the activity as irrelevant if important conditions are lacking.  Collaboration through text talk makes individual resources accessible in a specific place where it can be observed and its validity for the purpose of the task evaluated by others. Students with good insight in what they are supposed to accomplish seem be able to consult relevant guidance for this evaluation, from teachers, textbooks, scientific articles and other valid experiences important to their studies, and thereby contribute to learning of the quality they studies are meant to produce. Text talk also increases teachers’ possibilities to identify what the guidance the study group needs when evaluating the gathered resources and through their own active participation provide support in the students “zone of proximal development”. Contributions offered to the CSCL research field is the identifications of important mechanisms related to learning collaboratively through text talk, and the use of case study methodology as inspiration for others to try also these kinds of strategies to capture online learning.

Els grans avenços en el camp de l’aprenentatge automàtic han resultat en el disseny de màquines competents que són capaces d’aprendre i d’extreure informació útil i original de l’experiència. Recentment, algunes d’aquestes tècniques d’aprenentatge s’han aplicat amb èxit per resoldre problemes del món real en àmbits tecnològics, mèdics, científics i industrials, els quals no es podien tractar amb tècniques convencionals d’anàlisi ja sigui per la seva complexitat o pel gran volum de dades a processar. Donat aquest èxit inicial, actualment els sistemes d’aprenentatge s’enfronten a problemes de complexitat més elevada, el que ha resultat en un augment de l’activitat investigadora entorn sistemes capaços d’afrontar nous problemes del món real eficientment i de manera escalable. Una de les famílies d’algorismes més prometedores en l’aprenentatge automàtic són els sistemes classificadors basats en algorismes genetics (LCSs), el funcionament dels quals s’inspira en la natura. Els LCSs intenten representar les polítiques d’actuació d’experts humans amb un conjunt de regles que s’empren per escollir les millors accions a realitzar en tot moment. Així doncs, aquests sistemes aprenen polítiques d’actuació de manera incremental a mida que van adquirint experiència a través de la informació nova que se’ls va presentant durant el temps. Els LCSs s’han aplicat, amb èxit, a camps tan diversos com la predicció de càncer de pròstata o el suport a la inversió en borsa, entre altres. A més en alguns casos s’ha demostrat que els LCSs realitzen tasques superant la precisió dels éssers humans. El propòsit d’aquesta tesi és explorar la naturalesa de l’aprenentatge online dels LCSs d’estil Michigan per a la mineria de grans quantitats de dades en forma de fluxos d’informació continus a alta velocitat i canviants en el temps. Molt sovint, l’extracció de coneixement a partir d’aquestes fonts de dades és clau per tal d’obtenir una millor comprensió dels processos que les dades estan descrivint. Així, aprendre d’aquestes dades planteja nous reptes a les tècniques tradicionals d’aprenentatge automàtic, les quals no estan dissenyades per tractar fluxos de dades continus i on els conceptes i els nivells de soroll poden variar amb el temps de forma arbitrària. La contribució de la present tesi pren l’eXtended Classifier System (XCS), el LCS d’estil Michigan més estudiat i un dels algoritmes d’aprenentatge automàtic més competents, com el punt de partida. D’aquesta manera els reptes abordats en aquesta tesi són dos: el primer desafiament és la construcció d’un sistema supervisat competent sobre el framework dels LCSs d’estil Michigan que aprèn dels fluxos de dades amb una capacitat de reacció ràpida als canvis de concepte i entrades amb soroll. Com moltes aplicacions científiques i industrials generen grans quantitats de dades sense etiquetar, el segon repte és aplicar les lliçons apreses per continuar amb el disseny de LCSs d’estil Michigan capaços de solucionar problemes online sense assumir una estructura a priori en els dades d’entrada.
Los grandes avances en el campo del aprendizaje automático han resultado en el diseño de máquinas capaces de aprender y de extraer información útil y original de la experiencia. Recientemente alguna de estas técnicas de aprendizaje se han aplicado con éxito para resolver problemas del mundo real en ámbitos tecnológicos, médicos, científicos e industriales, los cuales no se podían tratar con técnicas convencionales de análisis ya sea por su complejidad o por el gran volumen de datos a procesar. Dado este éxito inicial, los sistemas de aprendizaje automático se enfrentan actualmente a problemas de complejidad cada vez m ́as elevada, lo que ha resultado en un aumento de la actividad investigadora en sistemas capaces de afrontar nuevos problemas del mundo real de manera eficiente y escalable. Una de las familias más prometedoras dentro del aprendizaje automático son los sistemas clasificadores basados en algoritmos genéticos (LCSs), el funcionamiento de los cuales se inspira en la naturaleza. Los LCSs intentan representar las políticas de actuación de expertos humanos usando conjuntos de reglas que se emplean para escoger las mejores acciones a realizar en todo momento. Así pues estos sistemas aprenden políticas de actuación de manera incremental mientras van adquiriendo experiencia a través de la nueva información que se les va presentando. Los LCSs se han aplicado con éxito en campos tan diversos como en la predicción de cáncer de próstata o en sistemas de soporte de bolsa, entre otros. Además en algunos casos se ha demostrado que los LCSs realizan tareas superando la precisión de expertos humanos. El propósito de la presente tesis es explorar la naturaleza online del aprendizaje empleado por los LCSs de estilo Michigan para la minería de grandes cantidades de datos en forma de flujos continuos de información a alta velocidad y cambiantes en el tiempo. La extracción del conocimiento a partir de estas fuentes de datos es clave para obtener una mejor comprensión de los procesos que se describen. Así, aprender de estos datos plantea nuevos retos a las técnicas tradicionales, las cuales no están diseñadas para tratar flujos de datos continuos y donde los conceptos y los niveles de ruido pueden variar en el tiempo de forma arbitraria. La contribución del la presente tesis toma el eXtended Classifier System (XCS), el LCS de tipo Michigan más estudiado y uno de los sistemas de aprendizaje automático más competentes, como punto de partida. De esta forma los retos abordados en esta tesis son dos: el primer desafío es la construcción de un sistema supervisado competente sobre el framework de los LCSs de estilo Michigan que aprende de flujos de datos con una capacidad de reacción rápida a los cambios de concepto y al ruido. Como muchas aplicaciones científicas e industriales generan grandes volúmenes de datos sin etiquetar, el segundo reto es aplicar las lecciones aprendidas para continuar con el diseño de nuevos LCSs de tipo Michigan capaces de solucionar problemas online sin asumir una estructura a priori en los datos de entrada.
Last advances in machine learning have fostered the design of competent algorithms that are able to learn and extract novel and useful information from data. Recently, some of these techniques have been successfully applied to solve real-­‐world problems in distinct technological, scientific and industrial areas; problems that were not possible to handle by the traditional engineering methodology of analysis either for their inherent complexity or by the huge volumes of data involved. Due to the initial success of these pioneers, current machine learning systems are facing problems with higher difficulties that hamper the learning process of such algorithms, promoting the interest of practitioners for designing systems that are able to scalably and efficiently tackle real-­‐world problems. One of the most appealing machine learning paradigms are Learning Classifier Systems (LCSs), and more specifically Michigan-­‐style LCSs, an open framework that combines an apportionment of credit mechanism with a knowledge discovery technique inspired by biological processes to evolve their internal knowledge. In this regard, LCSs mimic human experts by making use of rule lists to choose the best action to a given problem situation, acquiring their knowledge through the experience. LCSs have been applied with relative success to a wide set of real-­‐ world problems such as cancer prediction or business support systems, among many others. Furthermore, on some of these areas LCSs have demonstrated learning capacities that exceed those of human experts for that particular task. The purpose of this thesis is to explore the online learning nature of Michigan-­‐style LCSs for mining large amounts of data in the form of continuous, high speed and time-­‐changing streams of information. Most often, extracting knowledge from these data is key, in order to gain a better understanding of the processes that the data are describing. Learning from these data poses new challenges to traditional machine learning techniques, which are not typically designed to deal with data in which concepts and noise levels may vary over time. The contribution of this thesis takes the extended classifier system (XCS), the most studied Michigan-­‐style LCS and one of the most competent machine learning algorithms, as the starting point. Thus, the challenges addressed in this thesis are twofold: the first challenge is building a competent supervised system based on the guidance of Michigan-­‐style LCSs that learns from data streams with a fast reaction capacity to changes in concept and noisy inputs. As many scientific and industrial applications generate vast amounts of unlabelled data, the second challenge is to apply the lessons learned in the previous issue to continue with the design of unsupervised Michigan-­‐style LCSs that handle online problems without assuming any a priori structure in input data.

Nous étudions le problème de suivi de cible dans une séquence vidéo sans aucune connaissance préalable autre qu'une référence annotée dans la première image. Pour résoudre ce problème, nous proposons une nouvelle méthode de suivi temps-réel se basant sur à la fois une représentation originale de l’objet à suivre (descripteur) et sur un algorithme adaptatif capable de suivre la cible même dans les conditions les plus difficiles comme le cas où la cible disparaît et réapparait dans le scène (ré-identification). Tout d'abord, pour la représentation d’une région de l’image à suivre dans le temps, nous proposons des améliorations au descripteur de covariance. Ce nouveau descripteur est capable d’extraire des caractéristiques spécifiques à la cible, tout en ayant la capacité à s’adapter aux variations de l’apparence de la cible. Ensuite, l’étape algorithmique consiste à mettre en cascade des modèles génératifs et des modèles discriminatoires afin d’exploiter conjointement leurs capacités à distinguer la cible des autres objets présents dans la scène. Les modèles génératifs sont déployés dans les premières couches afin d’éliminer les candidats les plus faciles alors que les modèles discriminatoires sont déployés dans les couches suivantes afin de distinguer la cibles des autres objets qui lui sont très similaires. L’analyse discriminante des moindres carrés partiels (AD-MCP) est employée pour la construction des modèles discriminatoires. Enfin, un nouvel algorithme d'apprentissage en ligne AD-MCP a été proposé pour la mise à jour incrémentale des modèles discriminatoires
We study the challenging problem of tracking an arbitrary object in video sequences with no prior knowledge other than a template annotated in the first frame. To tackle this problem, we build a robust tracking system consisting of the following components. First, for image region representation, we propose some improvements to the region covariance descriptor. Characteristics of a specific object are taken into consideration, before constructing the covariance descriptor. Second, for building the object appearance model, we propose to combine the merits of both generative models and discriminative models by organizing them in a detection cascade. Specifically, generative models are deployed in the early layers for eliminating most easy candidates whereas discriminative models are in the later layers for distinguishing the object from a few similar "distracters". The Partial Least Squares Discriminant Analysis (PLS-DA) is employed for building the discriminative object appearance models. Third, for updating the generative models, we propose a weakly-supervised model updating method, which is based on cluster analysis using the mean-shift gradient density estimation procedure. Fourth, a novel online PLS-DA learning algorithm is developed for incrementally updating the discriminative models. The final tracking system that integrates all these building blocks exhibits good robustness for most challenges in visual tracking. Comparing results conducted in challenging video sequences showed that the proposed tracking system performs favorably with respect to a number of state-of-the-art methods

Thesis (Ph.D.)--Indiana University, Dept. of Instructional Systems Technology, 2006.
"Title from dissertation home page (viewed June 26, 2007)." Source: Dissertation Abstracts International, Volume: 67-06, Section: A, page: 2121. Adviser: Charles Reigeluth.

The purpose of this dissertation research was to examine collaborative learning processes during a project-based small group activity in a graduate online course. The specific research questions were: (1) How can group collaboration be assessed quantitatively? (2) What factors hinder or facilitate small group collaboration? (3) Which participation behaviours in whole group discussions before entering small groups are associated with small group collaboration? I developed an analytical framework, the Small Group Collaborative Learning Model (SGCLM), for assessing small group collaboration during project-based activity by modifying the Community of Inquiry model (Garrison, Anderson, & Archer, 2000) in combination with the online interaction learning model (Benbunan-Fich, Hiltz, & Harasim, 2005) which used the input-process-output (IPO) framework (McGrath, 1964, 1984; McGrath, Arrow, & Berdahl, 2000). Based on the SGCLM, I analyzed 2,029 messages (732 messages from small group forums and 1297 messages from the whole group discussions by twenty four students enrolled for 13 weeks). The data were coded into three communication categories (cognitive, social, and managerial) as well as communication directions (sender and receiver). For the data analysis, multiple methodological approaches (content analysis, social network analysis, and qualitative analysis) were employed. Collaboration in six small groups was assessed by three quantitative indices in terms of a group’s communication quantity, group members’ participation equality, and a group’s information sharedness. Following the quantitative assessment, a qualitative examination of the collaboration processes was conducted to identify the specific problems indicated by the quantitative indices. Finally, statistical analyses were performed on students’ participation behaviours before entering the small groups to discover whether these behaviours were related to more/less collaboration in the context of the small groups. I conclude that the three indices can be helpful for researchers, instructors, and course designers who aim at assessing and facilitating project-based small group collaborations in terms of more active communication, more democratic contributions, and more open communication. The collaboration indices can be a useful rubric for instructors to capture potential problems during small group activities and to provide support for the groups. Limitations and suggestions for future research are discussed.

In this thesis we consider the problem of online learning with labelled graphs, in particular designing algorithms that can perform this problem quickly and with low memory requirements. We consider the tasks of Classification (in which we are asked to predict the labels of vertices) and Similarity Prediction (in which we are asked to predict whether two given vertices have the same label). The first half of the thesis considers non- probabilistic online learning, where there is no probability distribution on the labelling and we bound the number of mistakes of an algorithm by a function of the labelling's complexity (i.e. its "naturalness"), often the cut- size. The second half of the thesis considers probabilistic machine learning in which we have a known probability distribution on the labelling. Before considering probabilistic online learning we first analyse the junction tree algorithm, on which we base our online algorithms, and design a new ver- sion of it, superior to the otherwise current state of the art. Explicitly, the novel contributions of this thesis are as follows: • A new algorithm for online prediction of the labelling of a graph which has better performance than previous algorithms on certain graph and labelling families. • Two algorithms for online similarity prediction on a graph (a novel problem solved in this thesis). One performs very well whilst the other not so well but which runs exponentially faster. • A new (better than before, in terms of time and space complexity) state of the art junction tree algorithm, as well as an application of it to the problem of online learning in an Ising model. • An algorithm that, in linear time, finds the optimal junction tree for online inference in tree-structured Ising models, the resulting online junction tree algorithm being far superior to the previous state of the art. All claims in this thesis are supported by mathematical proofs.

L'apprentissage est un processus fondamental et sous-jacent au comportement du consommateur. Ceci est particulièrement vrai lorsque des (nouveaux) produits ou services sont achetés et utilisés pour la première fois. La recherche actuelle en psychologie et marketing a mis l'accent sur des principes pédagogiques pour expliquer comment les consommateurs apprennent à utiliser des produits hors ligne. Cette recherche vise à élargir ce concept en explorant les processus d'apprentissage dans des contextes en ligne et en se basant sur une approche andragogique, à savoir l’apprentissage des adultes, et d’un point de vue cognitif. En utilisant une approche multi-méthodes basée sur une approche qualitative, à savoir des entretiens semi-structurés et des observations non-participantes, et une approche quantitative, à savoir une enquête et des expérimentations, nos résultats contribuent à la compréhension du "consumer e-learning". Tout d'abord, nous identifions comment et par quels processus les consommateurs adultes apprennent dans un environnement en ligne. Deuxièmement, nous déterminons les facteurs andragogiques et online qui aident à réduire les efforts cognitifs des consommateurs dans l’apprentissage d’un nouveau produit, et par conséquent, améliorent leur appropriation d'un produit. Dans un contexte où l'inefficacité des modes d’emploi traditionnels conduit à un manque d’apprentissage et l’utilisation moindre d’un produit, cette recherche contribue au niveau théorique au champ de l'apprentissage des consommateurs en abordant la question de l'apprentissage des consommateurs d’un point de vue andragogique et cognitif, et en abordant des questions critiques telles que le désapprentissage de pratiques antérieures
Learning is a fundamental process underlying consumer behavior. This is especially true when (new) products or services are purchased and used for the first time. Existing research in psychology and marketing has focused on pedagogical principles to explain how consumers learn to use products in offline settings. This research aims to broaden this scope by exploring learning processes in online contexts and by drawing on an andragogical, i.e. adult learning, and cognitive perspective. Using a multi-method approach based on a qualitative study including semi-structured interviews and non-participant observations, and a quantitative part involving a survey and experiments, our results contribute to the understanding of consumer e-learning. First, we identify how and by which processes adult consumers learn in an online environment. Second, we determine andragogical and online factors that help reducing consumers’ cognitive effort in new product learning, and consequently improve their appropriation of the product usage. In a context in which the ineffectiveness of traditional step-by-step instructions leads to reduced insight-based learning and product usage intention, this research contributes theoretically to the field of consumer learning by investigating consumer learning from an andragogical and cognitive perspective, and addressing critical issues such as product unlearning