Comparing Performance of Neural Networks Applied to a Simplified Recognition Problem

May 2000

Proceedings of the Thirteenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2000)

Volume

Proceedings of the Thirteenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2000)

All Papers

In this note we present and discuss results of experiments comparing the performance of six neural network architectures (back propagation, recurrent network with dampened feedback, network with multiple hidden layers each with a different activation function, jump connection networks, probabilistic neural networks and general regression neural networks) applied to a simplified multi-font recognition problem.