An artificial synaptic plasticity mechanism for classical conditioning with neural networks

Abstract

We present an artificial synaptic plasticity (ASP) mechanism that allows artificial systems to make associations between environmental stimuli and learn new skills at runtime. ASP builds on the classical neural network for simulating associative learning, which is induced through a conditioning-like procedure. Experiments in a simulated mobile robot demonstrate that ASP has successfully generated conditioned responses. The robot has learned during environmental exploration to use sensors added after training, improving its object-avoidance capabilities.