A LEARNING ENVIRONMENT FOR NEURAL NETWORKS AND TRANSIENT ACOUSTICS

THE PROPOSED RESEARCH IS CONCERNED WITH DEVELOPING AN INTELLIGENT TUTOR WHICH GOES BEYOND STATIC CONTEXT-SENSITIVE HELP SCREENS. NETROLOGIC'S PROPOSED SYSTEM WILL JUDGE THE EPISTEMIC STATE OF A USER AND DEVISE ALTERNATIVE GOAL-DIRECTED STRATEGIES. THE EXPERTISE OF THE INTELLIGENT TUTOR WILL COVER NEURAL MODELS AND TRANSIENT ACOUSTICS. SEVERAL NEW LEARNING MODELS WILL BE INCORPORATED IN THE TUTOR, SUCH AS KOHONEN LEARNING, ADAPTIVE VECTOR QUANTIZATION, TABER'S NEURON RING AND HEBBIAN LEARNING. THE PROPOSED AUTOMATED DECISION SUPPORT AID WILL ENABLE AN ACCELERATION IN THE DEVELOPMENT OF FINELY TUNED, HIGHLY EFFECTIVE SONAR SIGNAL CLASSIFIER. PHASE I OF THE RESEARCH WILL BE BOTH A THEORETICAL AND DEVELOPMENTAL STUDY. NEURAL NETWORK EXPERIMENTS WILL BE RUN ON NETROLOGIC'S IN-HOUSE NEUROCOMPUTER.