Supervised learning with potentials for neural network-based object recognition

Supervised learning techniques are widely used in object recognition based on neural networks. Presenting class-labelled samples to the neural network and employing certain learning criteria accomplish the supervised learning process. In this research we present a learning algorithm which uses the potential function between cluster centers and samples as the learning criterion. A learning process using Euclidean distance as the criterion is also performed. Results from both methods are compared.<
>