Deep learning is a form of machine learning that enables computers to learn from experience and understand the world in terms of a hierarchy of concepts. Because the computer gathers knowledge from experience, there is no need for a human computer operator to formally specify all the knowledge that the computer needs. The hierarchy of concepts allows the computer to learn complicated concepts by building them out of simpler ones; a graph of these hierarchies would be many layers deep. This book introduces a broad range of topics in deep learning. The text offers mathematical and conceptual background, covering relevant concepts in linear algebra, probability theory and information theory, numerical computation, and machine learning. It describes deep learning techniques used by practitioners in industry, including deep feedforward networks, regularization, optimization algorithms, convolutional networks, sequence modeling, and practical methodology; and it surveys such applications as natural language processing, speech recognition, computer vision, online recommendation systems, bioinformatics, and videogames. Finally, the book offers research perspectives, covering such theoretical topics as linear factor models, autoencoders, representation learning, structured probabilistic models, Monte Carlo methods, the partition function, approximate inference, and deep generative models. Deep Learning can be used by undergraduate or graduate students planning careers in either industry or research, and by software engineers who want to begin using deep learning in their products or platforms. A website offers supplementary material for both readers and instructors.

Deep Learning

Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning.

/pdf/reinforcement-learning-an-introduction-rzxgej9p17.pdf

Reinforcement Learning: An Introduction

In the first part of the paper we consider the problem of dynamically apportioning resources among a set of options in a worst-case on-line framework. The model we study can be interpreted as a broad, abstract extension of the well-studied on-line prediction model to a general decision-theoretic setting. We show that the multiplicative weight-update Littlestone?Warmuth rule can be adapted to this model, yielding bounds that are slightly weaker in some cases, but applicable to a considerably more general class of learning problems. We show how the resulting learning algorithm can be applied to a variety of problems, including gambling, multiple-outcome prediction, repeated games, and prediction of points in Rn. In the second part of the paper we apply the multiplicative weight-update technique to derive a new boosting algorithm. This boosting algorithm does not require any prior knowledge about the performance of the weak learning algorithm. We also study generalizations of the new boosting algorithm to the problem of learning functions whose range, rather than being binary, is an arbitrary finite set or a bounded segment of the real line.

/pdf/a-decision-theoretic-generalization-of-on-line-learning-and-3ykfx1u7er.pdf

A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting

https://www2.econ.iastate.edu/tesfatsi/DeepLearningInNeuralNetworksOverview.JSchmidhuber2015.pdf

Deep learning in neural networks

The game of Go has long been viewed as the most challenging of classic games for artificial intelligence owing to its enormous search space and the difficulty of evaluating board positions and moves. Here we introduce a new approach to computer Go that uses ‘value networks’ to evaluate board positions and ‘policy networks’ to select moves. These deep neural networks are trained by a novel combination of supervised learning from human expert games, and reinforcement learning from games of self-play. Without any lookahead search, the neural networks play Go at the level of stateof-the-art Monte Carlo tree search programs that simulate thousands of random games of self-play. We also introduce a new search algorithm that combines Monte Carlo simulation with value and policy networks. Using this search algorithm, our program AlphaGo achieved a 99.8% winning rate against other Go programs, and defeated the human European Go champion by 5 games to 0. This is the first time that a computer program has defeated a human professional player in the full-sized game of Go, a feat previously thought to be at least a decade away.

/pdf/mastering-the-game-of-go-with-deep-neural-networks-and-tree-3viuxo6uxm.pdf

Mastering the game of Go with deep neural networks and tree search

We consider the problem of learning a concept from examples in the distribution-free model by Valiant. (An essentially equivalent model, if one ignores issues of computational difficulty, was studied by Vapnik and Chervonenkis.) We introduce the notion of dynamic sampling, wherein the number of examples examined may increase with the complexity of the target concept. This method is used to establish the learnability of various concept classes with an infinite Vapnik-Chervonenkis dimension. We also discuss an important variation on the problem of learning from examples, called approximating from examples. Here we do not assume that the target concept T is a member of the concept class C from which approximations are chosen. This problem takes on particular interest when the VC dimension of C is infinite. Finally, we discuss the problem of computing the VC dimension of a finite concept set defined on a finite domain and consider the structure of classes of a fixed small dimension.

Results on learnability and the Vapnik-Chervonenkis dimension

We introduce a new general scheme for shared memory nonpreemptive scheduling policies. Our scheme utilizes a system of inequalities and thresholds and accepts a packet if it does not violate any of the inequalities. We demonstrate that many of the existing policies can be described using our scheme, thus validating its generality. We propose a new scheduling policy, based on our general scheme, which we call the harmonic policy. Our simulations show that the harmonic policy both achieves high throughput and easily adapts to changing load conditions. We also perform a theoretical analysis of the harmonic policy and demonstrate that its throughput competitive ratio is almost optimal.

Harmonic buffer management policy for shared memory switches

We study the problem of controlling linear time-invariant systems with known noisy dynamics and adversarially chosen quadratic losses. We present the first efficient online learning algorithms in this setting that guarantee $O(\sqrt{T})$ regret under mild assumptions, where $T$ is the time horizon. Our algorithms rely on a novel SDP relaxation for the steady-state distribution of the system. Crucially, and in contrast to previously proposed relaxations, the feasible solutions of our SDP all correspond to "strongly stable" policies that mix exponentially fast to a steady state.

Online Linear Quadratic Control

Phantom: a simple and effective flow control scheme

We consider the amount of randomness used in private distributed computations. Specifically, we show how n players can compute the exclusive-or (xor) of n boolean inputs t-privately, using only O(t2 log (n/t)) random bits (the best known upper bound is O(tn)). We accompany this result by a lower bound on the number of random bits required to carry out this task; we show that any protocol solving this problem requires at least t random bits (again, this significantly improves over the known lower bounds).
For the upper bound, we show how, given m subsets of {1,...,n}, to construct in (deterministic) polynomial time a probability distribution of n random variables (i.e., a probability distribution over {0,1}n) such that (1) the parity of random variables in each of these m subsets is 0 or 1 with equal probability, and (2) the support of the distribution is of size at most 2m. This construction generalizes previously considered types of sample spaces (such as k-wise independent spaces and Schulman's spaces [Sample spaces uniform on neighborhoods, in Proc. of the 24th Annual ACM Symposium on Theory of Computing, ACM, New York, 1992, pp. 17--25]). We believe that this construction is of independent interest and may have various applications.

Yishay Mansour

Papers

Results on learnability and the Vapnik-Chervonenkis dimension

Harmonic buffer management policy for shared memory switches

Online Linear Quadratic Control

Phantom: a simple and effective flow control scheme

Randomness in Private Computations