Implicit online learning

Online learning algorithms have recently risen to prominence due to their strong theoretical guarantees and an increasing number of practical applications for large-scale data analysis problems. In this paper, we analyze a class of online learning algorithms based on fixed potentials and nonlinearized losses, which yields algorithms with implicit update rules. We show how to efficiently compute these updates, and we prove regret bounds for the algorithms. We apply our formulation to several special cases where our approach has benefits over existing online learning methods. In particular, we provide improved algorithms and bounds for the online metric learning problem, and show improved robustness for online linear prediction problems. Results over a variety of data sets demonstrate the advantages of our framework.

Learning the Kernel Matrix with Semi-Definite Programming

Kernel-based learning algorithms work by embedding the data into a Euclidean space, and then searching for linear relations among the embedded data points. The embedding is performed implicitly, by specifying the inner products between each pair of points in the embedding space. This information is contained in the so-called kernel matrix, a symmetric and positive definite matrix that encodes the relative positions of all points. Specifying this matrix amounts to specifying the geometry of the embedding space and inducing a notion of similarity in the input space -- classical model selection problems in machine learning. In this paper we show how the kernel matrix can be learned from data via semi-definite programming (SDP) techniques. When applied to a kernel matrix associated with both training and test data this gives a powerful transductive algorithm -- using the labelled part of the data one can learn an embedding also for the unlabelled part. The similarity between test points is inferred from training points and their labels. Importantly, these learning problems are convex, so we obtain a method for learning both the model class and the function without local minima. Furthermore, this approach leads directly to a convex method to learn the 2-norm soft margin parameter in support vector machines, solving another important open problem. Finally, the novel approach presented in the paper is supported by positive empirical results.

Near knowledge : inductive learning systems in law

Induction is an interesting model of legal reasoning, since it provides a method of capturing initial states of legal principles and rules, and adjusting these principles and rules over time as the law changes. In this article I explain how Artificial Intelligence-based inductive learning algorithms work, and show how they have been used in law to model legal domains. I identify some problems with implementations undertaken in law to date, and create a taxonomy of appropriate cases to use in legal inductive inferencing systems. I suggest that inductive learning algorithms have potential in modeling law, but that the artificial intelligence implementations to date are problematic. I argue that induction should be further investigated, since it has the potential to be an extremely useful mechanism for understanding legal domains.

Brain decoding based on functional magnetic resonance imaging using machine learning : a comparative study

Brain decoding of functional Magnetic Resonance Imaging data is a pattern analysis task that links brain activity patterns to the experimental conditions. Classifiers predict the neural states from the spatial and temporal pattern of brain activity extracted from multiple voxels in the functional images in a certain period of time. The prediction results offer insight into the nature of neural representations and cognitive mechanisms and the classification accuracy determines our confidence in understanding the relationship between brain activity and stimuli. In this paper, we compared the efficacy of three machine learning algorithms: neural network, support vector machines, and conditional random field to decode the visual stimuli or neural cognitive states from functional Magnetic Resonance data. Leave-one-out cross validation was performed to quantify the generalization accuracy of each algorithm on unseen data. The results indicated support vector machine and conditional random field have comparable performance and the potential of the latter is worthy of further investigation.

Detection of rain in acoustic recordings of the environment using machine learning techniques

This thesis is concerned with the detection and prediction of rain in environmental recordings using different machine learning algorithms. The results obtained in this research will help ecologists to efficiently analyse environmental data and monitor biodiversity.

External knowledge and query strategies in active learning: A study in clinical information extraction

This paper presents a new active learning query strategy for information extraction, called Domain Knowledge Informativeness (DKI). Active learning is often used to reduce the amount of annotation effort required to obtain training data for machine learning algorithms. A key component of an active learning approach is the query strategy, which is used to iteratively select samples for annotation. Knowledge resources have been used in information extraction as a means to derive additional features for sample representation. DKI is, however, the first query strategy that exploits such resources to inform sample selection. To evaluate the merits of DKI, in particular with respect to the reduction in annotation effort that the new query strategy allows to achieve, we conduct a comprehensive empirical comparison of active learning query strategies for information extraction within the clinical domain. The clinical domain was chosen for this work because of the availability of extensive structured knowledge resources which have often been exploited for feature generation. In addition, the clinical domain offers a compelling use case for active learning because of the necessary high costs and hurdles associated with obtaining annotations in this domain. Our experimental findings demonstrated that 1) amongst existing query strategies, the ones based on the classification model’s confidence are a better choice for clinical data as they perform equally well with a much lighter computational load, and 2) significant reductions in annotation effort are achievable by exploiting knowledge resources within active learning query strategies, with up to 14% less tokens and concepts to manually annotate than with state-of-the-art query strategies.

Harnessing information from injury narrative in the 'big data' era: Understanding and applying machine learning for injury surveillance

Objective Vast amounts of injury narratives are collected daily and are available electronically in real time and have great potential for use in injury surveillance and evaluation. Machine learning algorithms have been developed to assist in identifying cases and classifying mechanisms leading to injury in a much timelier manner than is possible when relying on manual coding of narratives. The aim of this paper is to describe the background, growth, value, challenges and future directions of machine learning as applied to injury surveillance. Methods This paper reviews key aspects of machine learning using injury narratives, providing a case study to demonstrate an application to an established human-machine learning approach. Results The range of applications and utility of narrative text has increased greatly with advancements in computing techniques over time. Practical and feasible methods exist for semi-automatic classification of injury narratives which are accurate, efficient and meaningful. The human-machine learning approach described in the case study achieved high sensitivity and positive predictive value and reduced the need for human coding to less than one-third of cases in one large occupational injury database. Conclusion The last 20 years have seen a dramatic change in the potential for technological advancements in injury surveillance. Machine learning of ‘big injury narrative data’ opens up many possibilities for expanded sources of data which can provide more comprehensive, ongoing and timely surveillance to inform future injury prevention policy and practice.

Natural actor-critic algorithms

We present four new reinforcement learning algorithms based on actor-critic, natural-gradient and functi approximation ideas,and we provide their convergence proofs. Actor-critic reinforcement learning methods are online approximations to policy iteration in which the value-function parameters are estimated using temporal difference learning and the policy parameters are updated by stochastic gradient descent. Methods based on policy gradients in this way are of special interest because of their compatibility with function-approximation methods, which are needed to handle large or infinite state spaces. The use of temporal difference learning in this way is of special interest because in many applications it dramatically reduces the variance of the gradient estimates. The use of the natural gradient is of interest because it can produce better conditioned parameterizations and has been shown to further reduce variance in some cases. Our results extend prior two-timescale convergence results for actor-critic methods by Konda and Tsitsiklis by using temporal difference learning in the actor and by incorporating natural gradients. Our results extend prior empirical studies of natural actor-critic methods by Peters, Vijayakumar and Schaal by providing the first convergence proofs and the first fully incremental algorithms.

Incremental natural-gradient actor-critic algorithms

We present four new reinforcement learning algorithms based on actor-critic and natural-gradient ideas, and provide their convergence proofs. Actor-critic rein- forcement learning methods are online approximations to policy iteration in which the value-function parameters are estimated using temporal difference learning and the policy parameters are updated by stochastic gradient descent. Methods based on policy gradients in this way are of special interest because of their com- patibility with function approximation methods, which are needed to handle large or infinite state spaces. The use of temporal difference learning in this way is of interest because in many applications it dramatically reduces the variance of the gradient estimates. The use of the natural gradient is of interest because it can produce better conditioned parameterizations and has been shown to further re- duce variance in some cases. Our results extend prior two-timescale convergence results for actor-critic methods by Konda and Tsitsiklis by using temporal differ- ence learning in the actor and by incorporating natural gradients, and they extend prior empirical studies of natural actor-critic methods by Peters, Vijayakumar and Schaal by providing the first convergence proofs and the first fully incremental algorithms.

Reinforcement learning with average cost for adaptive control of traffic lights at intersections

We propose for the first time two reinforcement learning algorithms with function approximation for average cost adaptive control of traffic lights. One of these algorithms is a version of Q-learning with function approximation while the other is a policy gradient actor-critic algorithm that incorporates multi-timescale stochastic approximation. We show performance comparisons on various network settings of these algorithms with a range of fixed timing algorithms, as well as a Q-learning algorithm with full state representation that we also implement. We observe that whereas (as expected) on a two-junction corridor, the full state representation algorithm shows the best results, this algorithm is not implementable on larger road networks. The algorithm PG-AC-TLC that we propose is seen to show the best overall performance.

Contributions to unsupervised and supervised learning with applications in digital image processing

311 p. : il.

Efficient learning of decomposable models with a bounded clique size

The learning of probability distributions from data is a ubiquitous problem in the fields of Statistics and Artificial Intelligence. During the last decades several learning algorithms have been proposed to learn probability distributions based on decomposable models due to their advantageous theoretical properties. Some of these algorithms can be used to search for a maximum likelihood decomposable model with a given maximum clique size, k, which controls the complexity of the model. Unfortunately, the problem of learning a maximum likelihood decomposable model given a maximum clique size is NP-hard for k > 2. In this work, we propose a family of algorithms which approximates this problem with a computational complexity of O(k · n^2 log n) in the worst case, where n is the number of implied random variables. The structures of the decomposable models that solve the maximum likelihood problem are called maximal k-order decomposable graphs. Our proposals, called fractal trees, construct a sequence of maximal i-order decomposable graphs, for i = 2, ..., k, in k − 1 steps. At each step, the algorithms follow a divide-and-conquer strategy based on the particular features of this type of structures. Additionally, we propose a prune-and-graft procedure which transforms a maximal k-order decomposable graph into another one, increasing its likelihood. We have implemented two particular fractal tree algorithms called parallel fractal tree and sequential fractal tree. These algorithms can be considered a natural extension of Chow and Liu’s algorithm, from k = 2 to arbitrary values of k. Both algorithms have been compared against other efficient approaches in artificial and real domains, and they have shown a competitive behavior to deal with the maximum likelihood problem. Due to their low computational complexity they are especially recommended to deal with high dimensional domains.

Spinal cord injury therapy through active learning

Therapy employing epidural electrostimulation holds great potential for improving therapy for patients with spinal cord injury (SCI) (Harkema et al., 2011). Further promising results from combined therapies using electrostimulation have also been recently obtained (e.g., van den Brand et al., 2012). The devices being developed to deliver the stimulation are highly flexible, capable of delivering any individual stimulus among a combinatorially large set of stimuli (Gad et al., 2013). While this extreme flexibility is very useful for ensuring that the device can deliver an appropriate stimulus, the challenge of choosing good stimuli is quite substantial, even for expert human experimenters. To develop a fully implantable, autonomous device which can provide useful therapy, it is necessary to design an algorithmic method for choosing the stimulus parameters. Such a method can be used in a clinical setting, by caregivers who are not experts in the neurostimulator's use, and to allow the system to adapt autonomously between visits to the clinic. To create such an algorithm, this dissertation pursues the general class of active learning algorithms that includes Gaussian Process Upper Confidence Bound (GP-UCB, Srinivas et al., 2010), developing the Gaussian Process Batch Upper Confidence Bound (GP-BUCB, Desautels et al., 2012) and Gaussian Process Adaptive Upper Confidence Bound (GP-AUCB) algorithms. This dissertation develops new theoretical bounds for the performance of these and similar algorithms, empirically assesses these algorithms against a number of competitors in simulation, and applies a variant of the GP-BUCB algorithm in closed-loop to control SCI therapy via epidural electrostimulation in four live rats. The algorithm was tasked with maximizing the amplitude of evoked potentials in the rats' left tibialis anterior muscle. These experiments show that the algorithm is capable of directing these experiments sensibly, finding effective stimuli in all four animals. Further, in direct competition with an expert human experimenter, the algorithm produced superior performance in terms of average reward and comparable or superior performance in terms of maximum reward. These results indicate that variants of GP-BUCB may be suitable for autonomously directing SCI therapy.

Deep learning review and its applications

Deep neural networks have recently gained popularity for improv- ing state-of-the-art machine learning algorithms in diverse areas such as speech recognition, computer vision and bioinformatics. Convolutional networks especially have shown prowess in visual recognition tasks such as object recognition and detection in which this work is focused on. Mod- ern award-winning architectures have systematically surpassed previous attempts at tackling computer vision problems and keep winning most current competitions. After a brief study of deep learning architectures and readily available frameworks and libraries, the LeNet handwriting digit recognition network study case is developed, and lastly a deep learn- ing network for playing simple videogames is reviewed.

Bayesian Active Learning for Classification and Preference Learning

Information theoretic active learning has been widely studied for probabilistic models. For simple regression an optimal myopic policy is easily tractable. However, for other tasks and with more complex models, such as classification with nonparametric models, the optimal solution is harder to compute. Current approaches make approximations to achieve tractability. We propose an approach that expresses information gain in terms of predictive entropies, and apply this method to the Gaussian Process Classifier (GPC). Our approach makes minimal approximations to the full information theoretic objective. Our experimental performance compares favourably to many popular active learning algorithms, and has equal or lower computational complexity. We compare well to decision theoretic approaches also, which are privy to more information and require much more computational time. Secondly, by developing further a reformulation of binary preference learning to a classification problem, we extend our algorithm to Gaussian Process preference learning.