Comparing Performance of Neural Networks Applied to a Simplified Recognition Problem

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.

• Type: Conference
• Year: 2000
• Where: FLAIRS
• Authors: Marcin Paprzycki, Rick Niess, Jason Thomas, Lenny Scardino, William Douglas
• Keywords: Artificial Intelligence, FLAIRS 2000, Neural Network Architectures, Neural Networks, Probabilistic Neural Networks