Adaptive, Learning, and Pattern Recognition Systems; theory and applications

This volume offers a glimpse of the status of research in adaptive and learning systems in 1985. In recent years these areas have spawned a multiplicity of ideas so rapidly that the average research worker or practicing engineer is overwhelmed by the flood of information. The Yale Workshop on Applications of Adaptive Systems Theory was organized in 1979 to provide a brief respite from this deluge, wherein critical issues may be examined in a calm and collegial environment. The fourth of the series having been held in May 1985, it has now become well established as a biennial forum for the lively exchange of ideas in the ever changing domain of adaptive systems. The scope of this book is broad and ranges from theoretical investigations to practical applications. It includes twenty eight papers by leaders in the field, selected from the Pro ceedings of the Fourth Yale Workshop and divided into five sections. I have provided a brief introduction to each section so that it can be read as a self-contained unit. The first section, devoted to adaptive control theory, suggests the intensity of activity in the field and reveals signs of convergence towards some common themes by workers with rather different moti vation. Preliminary results concerning the reduced order model problem are dramatically changing the way we view the field and bringing it closer to other areas such as robust linear control where major advances have been recently reported.

In this book, we study theoretical and practical aspects of computing methods for mathematical modelling of nonlinear systems. A number of computing techniques are considered, such as methods of operator approximation with any given accuracy; operator interpolation techniques including a non-Lagrange interpolation; methods of system representation subject to constraints associated with concepts of causality, memory and stationarity; methods of system representation with an accuracy that is the best within a given class of models; methods of covariance matrix estimation;methods for low-rank matrix approximations; hybrid methods based on a combination of iterative procedures and best operator approximation; andmethods for information compression and filtering under condition that a filter model should satisfy restrictions associated with causality and different types of memory.As a result, the book represents a blend of new methods in general computational analysis,and specific, but also generic, techniques for study of systems theory ant its particularbranches, such as optimal filtering and information compression.- Best operator approximation,- Non-Lagrange interpolation,- Generic Karhunen-Loeve transform- Generalised low-rank matrix approximation- Optimal data compression- Optimal nonlinear filtering

The work described in this publication was initiated at the General Electric Company's Electronics Laboratory, Syracuse, N.Y., U.S.A. The author would like to take this opportunity to express his gratitude to the Electronics Laboratory for its support and encouragement in this work. Thanks are in particular due to Mr. J.J. Suran for his continued interest and help. It is impossible to acknowledge all the help the author has re ceived from members of the Laboratory staff. However, the author is par ticularly indebted to Mr. T.C. Robbins for managing the building of the word recognizer (described in Section 7.4) and for many helpful discussions. Thanks are also due to Mr. W.E. Sollecito for valued support and direction, and to S.M. Korzekwa, S.B. Akers, Jr., and B.L. Crew for many discussions on implementation and design of pattern recognizers. Part of the work has been sponsored by two departments of the General Electric Company, the Large Jet Engine Department and the Apollo Support Department. The author is grateful for the permission from the two departments to publish results of theoretical interest in this dissertation. The work was later continued in Denmark, supported by two grants: no.

Estuarine Perspectives presents most of the invited papers presented at the Fifth Biennial International Research Conference on Estuarine Research. The book includes information on one tropical and two Arctic estuaries; contemporary techniques as applied to estuarine research; and some hypotheses of estuarine ecology. The text also describes value and management of wetlands as well as the chemical cycles and fluxes. The primary production and photosynthesis; the physical and biological factors of estuarine sediment; and the ecosystem dynamics are also encompassed.

Introduction to Statistical Pattern Recognition introduces the reader to statistical pattern recognition, with emphasis on statistical decision and estimation. Pattern recognition problems are discussed in terms of the eigenvalues and eigenvectors. Comprised of 11 chapters, this book opens with an overview of the formulation of pattern recognition problems. The next chapter is devoted to linear algebra, with particular reference to the properties of random variables and vectors. Hypothesis testing and parameter estimation are then discussed, along with error probability estimation and linear classifiers. The following chapters focus on successive approaches where the classifier is adaptively adjusted each time one sample is observed; feature selection and linear mapping for one distribution and multidistributions; and problems of nonlinear mapping. The final chapter describes a clustering algorithm and considers criteria for both parametric and nonparametric clustering. This monograph will serve as a text for the introductory courses of pattern recognition as well as a reference book for practitioners in the fields of mathematics and statistics.