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

From the Publisher:
This is the first comprehensive treatment of feed-forward neural networks from the perspective of statistical pattern recognition. After introducing the basic concepts, the book examines techniques for modelling probability density functions and the properties and merits of the multi-layer perceptron and radial basis function network models. Also covered are various forms of error functions, principal algorithms for error function minimalization, learning and generalization in neural networks, and Bayesian techniques and their applications. Designed as a text, with over 100 exercises, this fully up-to-date work will benefit anyone involved in the fields of neural computation and pattern recognition.

/pdf/neural-networks-for-pattern-recognition-34puigiau1.pdf

Neural networks for pattern recognition

This paper proposes a paradigm for managing the dynamic aspects of organizational knowledge creating processes. Its central theme is that organizational knowledge is created through a continuous dialogue between tacit and explicit knowledge. The nature of this dialogue is examined and four patterns of interaction involving tacit and explicit knowledge are identified. It is argued that while new knowledge is developed by individuals, organizations play a critical role in articulating and amplifying that knowledge. A theoretical framework is developed which provides an analytical perspective on the constituent dimensions of knowledge creation. This framework is then applied in two operational models for facilitating the dynamic creation of appropriate organizational knowledge.

/pdf/a-dynamic-theory-of-organizational-knowledge-creation-22t9dxu55o.pdf

A Dynamic Theory of Organizational Knowledge Creation

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

Deep learning in neural networks

pH OF THE CEREBRAL CORTEX AND ARTERIAL BLOOD UNDER INSULIN

Some of you may be old enough to remember how, during World War I , the Austrian Empire laughed itself to pieces at the Good Soldier Schweik who had officially been declared feebleminded. Observation had convinced him that those of his officers who thought often made mistakes, and advised them t o avoid errors by not thinking-so you see, he was a Good Soldier. But the gambler and the banker have greater liberty to think within the limit set by their law: “If you can’t afford to lose, you can’t afford to win.” Scientists can afford to lose. We learn by making mistakes and correcting them. A quarter of a century ago, when a handful of us hazarded time and reputation on guesses about the behavior of closed loops, including computers, none of us dreamed of the proliferation of these ideas, which Norbert Wiener christened “Cybernetics.” He and I guessed that his book would sell about one thousand copies in the first year, and John Wiley & Sons took a chance on what turned out to be a best seller. Like us, the gods must have loved Norbert, for he died between lectures on two new themes. He taught his students how to build theories and invent new theorems; many of whom have said, “He taught me my business.” One day, he came splayfoot and triumphant from one of his inspiring lectures, but stopped abruptly in the hall, took his cigar from his mouth, and with a look of astonishment said, “My God! I’ve proved too much. There are no prime numbers,” and returned to the blackboard before his students had departed, to try to find his mistake. They did. Such was the teacher who considered the proofs of theorems in the textbooks “a way of hiding how mathematics is made.” Our proofs are afterthoughts and have little to d o with the ar t of conjecture and discovery. Teaching and learning are not what Donald MacKay calls competitive monologue, but dialogue in which each exposes the inadequacies of his map of the world for his partner to complete or correct. I shall have more to say of the formal properties of dialogue. But we have not always been so lucky in this field. I n 1943, Pitts and I were able to prove only three theorems concerning nets with circles. They were just enough to show that such nets could compute only those numbers that a Turing machine could compute with a finite tape. The years went by without new theorems.

After‐dinner address: the fun of failures*

I want to talk about neurons, which I hold to be coupled nonlinear oscillators. Unfortunately, I lack the mathematics to handle them. Van der Pol (personal communication) opined that whether an ocean of them, like the cerebral cortex, given any constant coefficient of coupling, would lock in phase, as in a convulsion, depended only on the density of the Gaussian primes as one goes away from the origin. He died too soon to prove or disprove it.

Of Digital Oscillators

Santiago Ramón y Cajal and his investigation of the nervous system.

I would like first to place bionics in a broader framework because there are several fields of interest which surround it on all sides. I therefore pick for my title “The Imitation of One Form of Life by Another—Biomimesis.” There is nothing new in biomimesis. It is so important in avoiding enemies and catching prey that it is determined in the genes of many insects: the walking stick, the velvet ant, and so on. It has been of enormous importance; it has given us the images of our gods and the costumes of our witch doctors. It has given us, since the wings of Daedalus, all sorts of transportation. We have mimicked and mimicked. There are a few things of which we can boast, like the wheel and the seeking of power from indirect sources—first from the sun by eating each other, so to speak, and then from steam engines and what not—all one way or another from the sun until we got fission and now fusion. But aside from sources of power, and from the wheel, most of what we have done has been an imitation. The imitation used to be primarily for our muscular chores:to pump out a mine, to carry material from one place to another, and so forth. The first major deviation from that, curiously enough, was in the steam engine itself. It was not the replacement of the boy who used to push the slide valve by a stick; the boy still did that. That gave us the regenerative cycle. It was by replacement of the boy at the throttle, the boy who controlled the rate at which the steam was admitted to the slide valve, that the first major deviation from doing mere mechanical work came about. I would like to start from that point with what seems to me the major problem we have to face in bionics, as opposed to cybernetics.

Warren S. McCulloch

Papers

pH OF THE CEREBRAL CORTEX AND ARTERIAL BLOOD UNDER INSULIN

After‐dinner address: the fun of failures*

Of Digital Oscillators

Santiago Ramón y Cajal and his investigation of the nervous system.

The Imitation of One Form of Life by Another—Biomimesis