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

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

Deep learning in neural networks

New developments in the science of learning science of learning overview mind and brain how experts differ from novices how children learn learning and transfer the learning environment curriculum, instruction and commnity effective teaching - examples in history, mathematics and science teacher learning technology to support learning conclusions from new developments in the science of learning.

How people learn: Brain, mind, experience, and school.

Background on genetic algorithms, LISP, and genetic programming hierarchical problem-solving introduction to automatically-defined functions - the two-boxes problem problems that straddle the breakeven point for computational effort Boolean parity functions determining the architecture of the program the lawnmower problem the bumblebee problem the increasing benefits of ADFs as problems are scaled up finding an impulse response function artificial ant on the San Mateo trail obstacle-avoiding robot the minesweeper problem automatic discovery of detectors for letter recognition flushes and four-of-a-kinds in a pinochle deck introduction to biochemistry and molecular biology prediction of transmembrane domains in proteins prediction of omega loops in proteins lookahead version of the transmembrane problem evolutionary selection of the architecture of the program evolution of primitives and sufficiency evolutionary selection of terminals evolution of closure simultaneous evolution of architecture, primitive functions, terminals, sufficiency, and closure the role of representation and the lens effect Appendices: list of special symbols list of special functions list of type fonts default parameters computer implementation annotated bibliography of genetic programming electronic mailing list and public repository

Genetic Programming: On the Programming of Computers by Means of Natural Selection

1 GAs: What Are They?.- 2 GAs: How Do They Work?.- 3 GAs: Why Do They Work?.- 4 GAs: Selected Topics.- 5 Binary or Float?.- 6 Fine Local Tuning.- 7 Handling Constraints.- 8 Evolution Strategies and Other Methods.- 9 The Transportation Problem.- 10 The Traveling Salesman Problem.- 11 Evolution Programs for Various Discrete Problems.- 12 Machine Learning.- 13 Evolutionary Programming and Genetic Programming.- 14 A Hierarchy of Evolution Programs.- 15 Evolution Programs and Heuristics.- 16 Conclusions.- Appendix A.- Appendix B.- Appendix C.- Appendix D.- References.

Genetic Algorithms + Data Structures = Evolution Programs

This book is a followon to the book in which John Koza introduced genetic programming (GP) to the world "enetic Programming: On the Programming of Computers by Means of Nataral Selection " [5] 1. As such, the primary intended audience is someone already familiar with GP; however, Koza does provide introductory material to both genetic algorithms (GA) and GP. The driving force behind this book is a method to automatically decompose a program into solvable components. The book presents this method, called automatically defined functions (ADF), and then presents case studies of the application of this method to a variety of problems. While this book's size is intimidating, there is a wealth of information to be found by the reader willing to conduct a prolonged campaign. The reader is advised to study the first seven chapters of the book to gain an understanding of the concepts behind ADFs. Then the reader should be able to select which case studies he or she finds to be of interest. If the reader is feeling overwhelmed by the information presented in the book, there are several concise chapters dealing with ADFs by Koza and others in Advances in Genetic Programming edited by Kenneth Kinnear [3] 2.

Genetic Programming II: Automatic Discovery of Reusable Programs

Computer simulations of artificial ecologies typically model the interactions of a population of independently acting, spatially situated, resource-restricted, differently structured, self-reproducing agents. The interactions in such systems are typically highly nonlinear, subject to random perturbation, executed in parallel, and coevolutionary in the sense that the agents interact with each other as well as their environment. Examples of artificial ecologies that possess the above characteristics include Skipper's \"computer zoo\" of migrating flocks of hierarchically invokable fragments of assembly code [10], Shanahan's populations of evolutionary automata [9], Lindgren's coevolving populations of strategies for playing the iterated prisoner's dilemma game [5], the Turing gas of self-organizing groups of autocatalytic algorithmic fragments of Rasmussen, KnudLsen, and Feldberg [6], Ray's populations of parasitic and symbiotic self-reproducing assembly code programs [8], the populations of redcode coreworld creatures of Rasmussen, Knudsen, Feldberg, and Hindsholm [7], Werner and Dyer's populations of males and females that coevolve communication [11], and Ackley and Littman's artificial world for testing the Baldwin effect concerning learning and evolution [1]. Each of these complex adaptive systems (and many others like them) succeed in illustrating one or more key features of living systems. However, each of these systems live up to their name in the worst possible way—they are all exceedingly complex. The emergent behavior and other interesting phenomena are obscured by so much modelspecific detail that it is rarely clear whether the observed phenomena represent any important general principles or are merely artifacts of the details. Moreover, all of these existing systems are computationally expensive and deliver little in the way of important emergent phenomena in relation to the amount of computational effort expended. This excessive overhead is not merely a matter of inefficiency or inconvenience; it may actually preclude emergence of important phenomena that can only materialize in the presence of certain minimum amounts of time or matter. John Holland's new book, Hidden Order: How Adaptation Builds Complexity [3],

Hidden Order: How Adaptation Builds Complexity.

Many seemingly different problems in machine learning, artificial intelligence, and symbolic processing can be viewed as requiring the discovery of a computer program that produces some desired output for particular inputs. When viewed in this way, the process of solving these problems becomes equivalent to searching a space of possible computer programs for a highly fit individual computer program. The recently developed genetic programming paradigm described herein provides a way to search the space of possible computer programs for a highly fit individual computer program to solve (or approximately solve) a surprising variety of different problems from different fields. In genetic programming, populations of computer programs are genetically bred using the Darwinian principle of survival of the fittest and using a genetic crossover (sexual recombination) operator appropriate for genetically mating computer programs. Genetic programming is illustrated via an example of machine learning of the Boolean 11-multiplexer function and symbolic regression of the econometric exchange equation from noisy empirical data.

http://www.genetic-programming.com/jkpdf/scjournallong.pdf

John R. Koza

Papers

Genetic Programming: On the Programming of Computers by Means of Natural Selection

Genetic Programming II: Automatic Discovery of Reusable Programs

Hidden Order: How Adaptation Builds Complexity.

Genetic programming as a means for programming computers by natural selection

Genetic Programming II