Inst., 19600 NW vonNeumann Dr, Beaverton, OR 97006 Abstract We propose a very simple, and well principled way of computing the optimal step size in gradient descent algorithms. The online version is very efficient computationally, and is applicable to large backpropagation networks trained on large data sets. The main ingredient is a technique for estimating the principal eigenvalue(s) and eigenvector(s) of the objective function's second derivative matrix (Hessian),which does not require to even calculate the Hessian. Severalother applications of this technique are proposed for speeding up learning, or for eliminating useless parameters. 1 INTRODUCTION Choosing the appropriate learning rate, or step size, in a gradient descent procedure such as backpropagation, is simultaneously one of the most crucial and expertintensive partof neural-network learning. We propose a method for computing the best step size which is both well-principled, simple, very cheap computationally, and, most of all, applicable to online training with large networks and data sets.

Artificial neural networks are comprised of an interconnected collection of certain nonlinear devices; examples of commonly used devices include linear threshold elements, sigmoidal elements and radial-basis elements. We employ results from harmonic analysis and the theory of rational approximation toobtain almost tight lower bounds on the size (i.e.

Language inference and automata induction using recurrent neural networks has gained considerable interest in the recent years. Nevertheless, success of these models hasbeen mostly limited to regular languages. Additional information in form of a priori knowledge has proved important and at times necessary for learning complex languages(Abu-Mostafa 1990; AI-Mashouq and Reed, 1991; Omlin and Giles, 1992; Towell, 1990). They have demonstrated that partial information incorporated in a connectionist model guides the learning process through constraints for efficient learning and better generalization. 'Ve have previously shown that the NNPDA model can learn Deterministic Context 65 66 Das, Giles, and Sun Output

This paper describes a technique called Input Reconstruction Reliability Estimation (IRRE) for determining the response reliability of a restricted class of multi-layer perceptrons (MLPs). The technique uses a network's ability to accurately encode the input pattern in its internal representation as a measure of its reliability. The more accurately a network is able to reconstruct the input pattern from its internal representation, the more reliable the network is considered to be. IRRE is provides a good estimate of the reliability of MLPs trained for autonomous driving. Results are presented in which the reliability estimates provided by IRRE are used to select between networks trained for different driving situations. 1 Introduction In many real world domains it is important to know the reliability of a network's response since a single network cannot be expected to accurately handle all the possible inputs.

Untill recently, state-of-the-art, large-vocabulary, continuous speech recognition (CSR) has employed Hidden Markov Modeling (HMM) to model speech sounds. In an attempt to improve over HMM we developed a hybrid system that integrates HMM technology with neural We present the concept of a "Segmental Neural Net"networks.

One way to speed up reinforcement learning is to enable learning to happen simultaneously at multiple resolutions in space and time. This paper shows how to create a Q-Iearning managerial hierarchy how to set tasks to their submanagersin which high level managers learn how to satisfy them. Sub-managerswho, in turn, learn understand their managers' commands. Theyneed not initially simply learn to maximise their reinforcement in the context of the current command. We illustrate the system using a simple maze task .. As the system learns how to get around, satisfying commands at the multiple than standard, flat, Q-Iearninglevels, it explores more efficiently and builds a more comprehensive map. 1 INTRODUCTION Straightforward reinforcement learning has been quite successful at some relatively thecomplex tasks like playing backgammon (Tesauro, 1992).

For over a decade now, AI techniques have been applied to some of the hardest problems faced by business today, often with stellar results and a tenfold-plus return on investment. One of the major problems faced by businesses in the 1990s is how to produce environmentally friendly products and stay profitable. A pioneering consortium at Carnegie Mellon University (CMU) is using AI, combined with operations research, environmental science, public policy, and other disciplines, to build tools for green engineering. Green engineering is an approach to product development that balances environmental compatibility against economic profitability. It looks at the entire life cycle of the product, from design to disposal, and seeks to extend this life cycle through remanufacturing, reusing, and recycling products and components.

Celite corporation and Andersen Consulting have developed an advanced approach to traditional software development called the application software factory (ASF)." The approach is an integration of technology and total quality "management" techniques that includes the use of an expert system to guide module design and perform "module programming." The expert system component is called the knowledge-based design assistant and its inclusion in the ASF methodology" has significantly reduced module development time, training time, and module and communication errors.

Analogical although analogy can help, as note that although still in its infancy reasoning is thus achieved in such well as hamper, learning. The role of and somewhat simplistic in character, systems by mainly keeping the analogy in learning is discussed by connectionist research might prove abstract relational microfeatures. Ann Brown and by Rand Spiro et al., to have an edge in tackling these Rumelhart proposes another way and the role of analogy in knowledge problems. The research described in for achieving analogical reasoning, acquisition is discussed by Brian Ross this book presents a grand challenge that is, "soft clamp," in which input and by John Bransford et al.; Stella and a future prospect for AI clamps can be overridden, and the Vosniadou studies the developmental researchers (traditional or connectionistic) rule of thumb is that the more concrete change in the use of analogy. Because in their endeavor to find a a feature is, the easier it can be part 3 of the book is of marginal better and more cognitively plausible overridden. The system finds the interest to AI, I do not discuss it any representation scheme.

The methodological underpinnings of AI are slowly changing. Benchmarks, test beds, and controlled experimentation are becoming more common. Although we are optimistic that this change can solidify the science of AI, we also recognize a set of difficult issues concerning the appropriate use of this methodology. We discuss these issues as they relate to research on agent design. We survey existing test beds for agents and argue for appropriate caution in their use. We end with a debate on the proper role of experimental methodology in the design and validation of planning agents.