Dice

This simulates a two-input perceptron, a neural network with one neuron.
Input data is generated by a combat simulator where the game statistics for the opponent are randomly generated with each fight.
By examining the opponents defensive armor level and amount of health, the perceptron must attempt to predict whether the combat will end in victory or defeat.
The line on the plot represents the current threshold decision level.
The last 30 combats are plotted as points as a visual aid for you to gauge the performance of the perceptron.
The red dots indicate defeat, green victory.
The perceptron does not remember past combats other than the impression they left in its two synaptic weights.
Over time, the threshold decision level will settle as the perceptron learns.

Click your mouse to start and stop the thread from running.


http://www.alumni.caltech.edu/~croft/java/Dice/Dice.html
last updated 1996-08-23 by David Croft.