Neural nets and alphabets: introducing students to neural networks

Abstract

Three student projects involving neural networks are described. The projects include recognizing handwritten letters of the alphabet, determining the orientation of an imaged line, and recognizing particular rooms of a house based on samples of furniture found in the rooms. All projects were run on a back propagation neural network program implemented in Modula-2. A description of the program is presented and a sample module for simulating the behavior of an OR gate is included as an appendix. The program has been successfully used in several Artificial Intelligence classes for classroom demonstrations and carrying out various cognitive science experiments.

Hopfield networks, neural data structures and the nine flies problem

“…Th e diagram on the right shows which three flies moved, and th e new arrangement . Schweller and Plagman (1989) and Wallace and Wallace (1991) agree that it is extremely important that students are introduced to both interesting and active research area s early in the curriculum, and strongly emphasise th e importance of exposing them to neural networ k programming . It also familiarizes students with a nontraditional programming paradigm .…”