Connectionistic problem solving: computational aspects of biological learning

From MaRDI portal





This research monograph addresses issues concerning the efficient learning and representation of categories, and the assembly of action sequences to achieve specific goals. Various approaches, often with complementary characteristics, are presented. It contains ten main chapters followed by a summary and conclusions, further readings and notes. The initial chapters (1-6) develop the basic node and network structures, and general node and network learning techniques. These chapters address the problems of efficient category learning and representation. The later chapters are more concerned with the issues of the whole-system problem solving. Chapter 1 is an introduction where an overview of the presented model is given. In chapter 2, the basic node/neuron, a linear threshold unit (LTU), and a standard node training algorithm is described. In chapter 3, some basic properties of LTU training speed are considered. Techniques for accelerating perceptron training and dealing with training noise are discussed. In chapter 4, focusing is introduced as an alternative to perceptron training. In chapter 5, the basic structure of an operator is discussed. In chapter 6, two training schemes for a two-level network are described, one based on perceptron training and the other on focusing. In chapter 7, the issues of assembling action sequences is addressed. In chapter 8, a stimulus- stimulus (S-S) model of behavior is developed. In chapter 9, biological evidence for and against the S-S model is discussed. In chapter 10, a final Stimulus-Goal-Response (S-G-R) model is developed. The monograph includes an extensive bibliography of related works in the neural- modeling and animal-learning literature. It does not include, however, any reference to relevant works in the learning automata literature \textit{K. S. Narendra} and \textit{M. L. Thathachar} [Learning Automata, An Introduction, Englewood Cliffs: Prentice Hall (1989)]. I find the monograph rather difficult to read. The numerous notes, references and quotes dispersed all over the text are rather distracting. The monograph seems to be a collection of works rather than a homogeneous text.











This page was built for publication: Connectionistic problem solving: computational aspects of biological learning

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3996720)