This research aims to create an interactive module for investigating and demonstrating various competitve learning methods implemented on neural networks. In the course of completing numerous software units on competitive neural networks, the need for visualization, animation, and interaction has emerged."One picture is worth a thousand words" somebody said. In this case it not only clarifies the scientific and theoretical concept, but also illustrates the internal changes that take place in the neural system,e.g. change of weights, neural activations, activation functions, and output levels.
Competitive methods for learning in neural networks represent a well of knowledge, because of the multitude of problems that could be modeled and the likeness to the biological nervous system. Our goal is to convey these characteristics by illustrating the neural networks states as well as allowing the user maximum freedom in choice of architectures, inputs,form of output, and interpretations.
This research is based on current and past projects on self-organizing maps with number of applications ,
adaptive resonance theory neural network implementations , and competitive clustering experiments.
