Advances in Information Systems Science

Information systems science embraces a broad spectrum of topics. It is vir­ tually impossible to provide comprehensive and in-depth discussion, other than simple recitals of recent results, of every important topic in each volume of this annual review series. Since we have chosen the former approach, each volume will only cover certain aspects of recent advances in this bur­ geoning field. The emphasis in this volume, the third of a continuing series, is focussed upon pattern recognition, pictorial information manipulation, and new approaches to logical design of information networks. In Chapter 1, V. A. Kovalevsky presents a tutorial survey of practical and theoretical developments in pattern recognition. He categorizes the basic developments in three different directions. The first direction is charac­ terized by an empirical treatment with highly specialized recognition schemes. In the second direction, the major efforts are centered upon the cre­ ation of learning systems capable of improving recognition performance on the basis of past experience. The majority of the work in the third direction is devoted to the study of the basic structure of complex patterns, the con­ struction of mathematical models for pattern recognition, and the analysis of complex pictorial representations. The author elucidates the "heuristics" approach and the "science" approach to pattern recognition problems. This chapter together with Chapter 2 of this volume supplements the chapter on Engineering Principles of Pattern Recognition in Volume 1 to provide a more complete treatment of this subject.

1 Pattern Recognition: Heuristics or Science?.- 1. Introduction.- 2. Principal Directions in Pattern Recognition.- 2.1. Basic Concepts.- 2.2. Heuristic Recognition Methods.- 2.3. Perceptrons.- 2.4. Learning As Approximation to a Decision Function.- 2.5. The Method of Stochastic Approximation.- 2.6. Methods Based on Assumptions About the Properties of the Observed Signals.- 2.7. Applied Results.- 3. Parametric Models of Signals.- 3.1. Distributions with Interfering Parameters.- 3.2. The Problem of Recognition of Complex Signals.- 3.3. The Statistical Problems of Supervised and Nonsupervised Learning.- 3.4. Parametric Models with Reference Patterns.- 4. The Method of Permissible Transformations.- 4.1. Formalization of the Concept of Resemblance.- 4.2. Permissible Transformations.- 4.3. Correlation Method.- 4.4. Effectiveness of the Correlation Method.- 5. Methods of Analyzing Complex Pictures.- 5.1. Formal Syntactic Rules for Constructing Complex Pictures.- 5.2. Description of Complex Pictures in the Presence of Noise (the Method of Reference Sequences).- 5.3. Examples of the Use of the Reference-Sequences Method.- 6. Conclusions.- References.- 2 Feature Compression.- 1. The Role of “Features” in Pattern Recognition.- 1.1. Four Kinds of Pattern Recognition and Features.- 1.2. Component and Composition—Structure Analysis.- 1.3. Pattern Recognition As Induction.- 1.4. Decision Procedure and Features.- 1.5. Selection of Variables.- 1.6. Distance and Feature.- 2. A Concrete Example of Feature Compression—Handwritten ZIP Code Reader.- 2.1. Nature of the Problem.- 2.2. Compression of Invariants.- 2.3. Local Features.- 2.4. Horizontal Zone Feature.- 2.5. Global Features.- 2.6. Feature Compression As Structural Analysis.- 3. Discriminatory Feature Compression—SELFIC.- 3.1. Rotations in Representation Space.- 3.2. Minimum-Entropy Principle.- 3.3. Basic Theorem of SELFIC.- 3.4. Discriminatory Feature Space and SELFIC.- 3.5. Object-Predicate Reciprocity.- 4. Characteristic Feature Compression—CLAFIC.- 4.1. Class-Feature Space.- 4.2. Subspace Model Versus Zone Model.- 4.3. Decision Procedures by Projection and by Entropy.- 5. Implications of Subspace Model—Fuzzy Class.- 5.1. Modular Nondistributive Predicate Lattice.- 5.2. Implications of the New Logic.- 5.3. Fuzzy Class.- References.- 3 Image Processing Principles and Techniques.- 1. Introduction.- 1.1. Central Problems.- 1.2. Processing for Data Compression.- 1.3. Processing for Enhancement.- 1.4. Processing for Classification.- 2. Filter Theory Applied to Images.- 2.1. Spatial Frequency Filtering.- 2.2. Matched Filtering.- 3. Statistical Decision Theory.- 3.1. Decision Theory Formalisms.- 3.2. Special Cases.- 3.3. Commentary on Applications.- 4. Adaptive Network Approaches.- 5. Image Features.- 5.1. Approximating Functions.- 5.2. Random Features.- 5.3. Feature Adaptation.- 5.4. Shape Features.- 5.5. Textural Features.- 5.6. Serially Derived Features.- 5.7. Picture Linguistics.- 5.8. Distance Features.- 6. Implementations: Staging.- 6.1. Realizable Decision Functions.- 6.2. Number of Stages.- 7. Implementations: Parallelism.- 7.1. All-Serial Methods.- 7.2. Parallel Operator, Serial Image Processing.- 7.3. Serial Operator, Parallel Image Processing.- 7.4. All-Parallel Methods.- 8. Electrooptical Devices.- 8.1. Point and Aperture Scanners.- 8.2. Image Parallel Devices.- 9. Digital Computers.- 9.1. The Fast Fourier Transform.- 9.2. Parallel Computers.- 10. Optical Techniques.- 10.1. Coherent Optics.- 10.2. Incoherent Optics.- 11. Comparison of Implementations.- 12. Conclusions.- References.- 4 Computer Graphics.- 1. Introduction.- 2. Devices for Computer Graphics.- 2.1. Noninteractive Graphic Output Devices.- 2.2. Noninteractive Graphic Input Devices.- 2.3. Input for Interaction.- 2.4. Interactive Display Operations.- 3. Modes of Interactive Graphic Systems.- 3.1. Shared Memory with Stand-Alone Dedicated Processor.- 3.2. Buffered Memory Systems.- 3.3. Large Machine with Satellite.- 3.4. Multiaccess Graphics.- 4. Data Structures.- 4.1. The Nature of Data Structure.- 4.2. List Structures.- 4.3. Ring and Associative Structures.- 4.4. Data Structure Operations.- 4.5. Choice of Data Structures.- 5. Graphics Software.- 5.1. Introduction.- 5.2. Techniques for Generation of Display File.- 5.3. Special Techniques.- 6. Graphic Languages.- 6.1. Introductory Remarks.- 6.2. Graphic Command Languages.- 6.3. Picture Processing Languages.- 7. Conclusions.- Appendix 1. Choice of Equations for Generating a Circle.- Appendix 2. Method Given by Forrest for Parametrizing a Conic.- References.- 5 Logical Design of Optimal Digital Networks by Integer Programming.- 1. Introduction.- 2. Features of Logical Design by Integer Programming.- 3. Design of an Optimal Combinational Network with a Given Type of Gate by Integer Programming.- 3.1. General Mathematical Formulation of Design Procedures with Threshold Gates.- 3.2. Design of an Optimal Network with NOR Gate or Other Types of Gates.- 4. Design of an Optimal Combinational Network with Building Blocks (or Composite Gates) by Integer Programming.- 4.1. Feed-Forward Network Formulation and Design Procedure of an Optimal Combinational Network.- 4.2. Computational Examples.- 4.3. Design of Optimal Networks with Composite Gates.- 5. Other Applications of the Integer Programming Logical Design Method.- 5.1. Design of Combinational Optimal Networks under Miscellaneous Conditions.- 5.2. Design of an Error-Correcting Optimal Network.- 5.3. Diagnosis of a Network by Integer Programming.- 5.4. Design of Optimal Sequential Networks by Integer Programming.- 6. Concluding Remarks.- References.