2-way XOR
Back to the Example

This Neural Net example implements a consultation module for a 2-way XOR problem. It takes three input values (v1, v2, v3) and produces one output which is the results of...

XOR( XOR(v1, v2), v3)
for example: for the inputs (1, -1, 1) the expected response is -1.

This is simply an even parity problem where the net was trained on values of ( -1, 1 ) rather than ( 0, 1 ).

To cycle the net, simply enter the three input values (range of +/- 1) and click the Cycle button. You can select input data from the Data Selection frame to the left. You can select an input pattern from either the training set or the example test set.

You will notice higher error for values in the +/- .3 range as can be expected.

You may modify the Input weight values for all hidden and output nodes by simply changing the value in the weight display and clicking on the Modify Weights button. To restore the original weights, click on the Reset Weights button.

NEW - 1-23-96 - You may now SAVE and RESTORE network states (weights and Bias values).

I M P O R T A N T - N O T E:
Network saves performed prior to 1-25-96 @ 10:am will not work due to a change in the order of parameter saves.
No further changes are planned.

Network state is saved and restored using the Netscape Cookie feature. Saved states will be retained for 72 hours. This feature will be improved as time permits.

Also, you may now to examine a list of saved states using the View Saved ID List button.