Backpropagation

honzour

À propos de l'application

A simple example of backpropagation algorithm. Learn small neural network basic functions like predefined examples: AND, XOR or 2D distance. Import and export of custom tasks from and to xml or well readable csv. Integration with www browser. Creating new or editing loaded tasks in an editor is also possible. You can show the network anatomy and all weights and also the result with marked training set or exact output for set input. If learning failes, restart whole learning algorithm or just single neuron weights. Alpha learning speed is automatic and configurable. Help on each screen. This program may be useful for students of a basic course of artificial neural networks. It illustrates well that learning of neural networks is a complex task and basic backpropagation without any improvements can only solve very simple tasks.

Nouveautés

Java export. Adds button. Targeted to 29 SDK.