Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day.

Gradient-based learning applied to document recognition

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

There is, I think, something ethereal about i —the square root of minus one. I remember first hearing about it at school. It seemed an odd beast at that time—an intruder hovering on the edge of reality.

Usually familiarity dulls this sense of the bizarre, but in the case of i it was the reverse: over the years the sense of its surreal nature intensified. It seemed that it was impossible to write mathematics that described the real world in …

I and i

The architecture and learning procedure underlying ANFIS (adaptive-network-based fuzzy inference system) is presented, which is a fuzzy inference system implemented in the framework of adaptive networks. By using a hybrid learning procedure, the proposed ANFIS can construct an input-output mapping based on both human knowledge (in the form of fuzzy if-then rules) and stipulated input-output data pairs. In the simulation, the ANFIS architecture is employed to model nonlinear functions, identify nonlinear components on-line in a control system, and predict a chaotic time series, all yielding remarkable results. Comparisons with artificial neural networks and earlier work on fuzzy modeling are listed and discussed. Other extensions of the proposed ANFIS and promising applications to automatic control and signal processing are also suggested. >

ANFIS: adaptive-network-based fuzzy inference system

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

Deep learning in neural networks

For pt.I see Proc. 37th IEEE Conf. on Decision and Control, pp. 2977-82 (1998). In this part, the identification and control problems of a nonlinear dynamical system are considered in a simplified framework. First it is shown that when the linearized system is observable, the state vector of the nonlinear system can be reconstructed from the input and the output together with their past values. This establishes the existence of solutions to regulation problems using only the input and the output. If the linearized system has a relative degree and asymptotically stable zero dynamics, under certain conditions, the nonlinear system also has a well-defined relative degree and asymptotically stable zero dynamics. In such a case, an input-output representation (the NARMA model) can be obtained in which the output is affected by the input through linear terms no later than through nonlinear terms, and by using the implicit function theorem, the existence of a control law can be established such that the output follows any given reference trajectory in a neighborhood of the origin.

http://ieeexplore.ieee.org/iel5/7965/22042/01024834.pdf

Identification and control of a nonlinear dynamical system based on its linearization. II

It is now well realized [1] that two unstable dynamical systems, attempting to stabilize each other using the error between their outputs for adjusting their parameters, may not always succeed. The fact that adaptation may result in instability makes the mathematical problem a very interesting one. Since similar problems are arising in many other branches of science (e.g psychology, biology, medicine etc), obtaining precisely the conditions under which stability is achieved is also assuming greater importance. This paper may be considered as a first attempt to discuss the many aspects of this intriguing and difficult problem. Work carried out during the past year has shown that the interaction of two nth order systems, results in a 4nth order differential equation, whose stability has to be investigated. Even in the simple case when n = 2, determining necessary and sufficient conditions for stability is a formidable undertaking. However, through extensive simulation studies and the theoretical analysis of special cases suggested by them, numerous insights have been obtained. The objective of this paper is to convey these insights, and discuss wherever possible, their implications to higher order systems.

Unstable systems stabilizing each other through adaptation - part II

The current status of control theory is examined briefly. A few examples are given to emphasize that the intelligent control of the present will be the conventional control of the future. The roles of prior information and neural networks in intelligent control are discussed. >

Intelligent control

This is the first of a series of papers in which a detailed investigation of the adaptive identification and control of linear dynamical systems with periodic coefficients is undertaken. In this paper the identification of stable linear periodic systems is investigated when all the state variables of the system are accessible. Error models similar to those known for LTI systems are derived and used to demonstrate the stability of the identification algorithms. The same concepts are applied to the adaptive control of a special class of plants with all state variables accessible. Towards the end of the paper results are also derived related to Floquet theory which will be useful for formulating more general adaptive control problems.

Adaptive Identification and Control of Linear Periodic Systems

http://ieeexplore.ieee.org/iel5/9/24097/01098693.pdf

Kumpati S. Narendra

Papers

Identification and control of a nonlinear dynamical system based on its linearization. II

Unstable systems stabilizing each other through adaptation - part II

Intelligent control

Adaptive Identification and Control of Linear Periodic Systems

Stability of nonlinear time-varying systems