Network Weights Shock dialog box

 

Use this dialog box to shock the current training network weights. Shocking network weights is necessary at times when a neural network appears to be caught in local minima.

 

Options

 

Shock Factor

Specifies the amount by which to alter the network weights. Type in this box a floating-point number greater than zero and smaller than or equal to one. The value of each connection weight is changed by generating uniformly distributed random numbers between (where x is the shock factor), that are multiplied and added to the weights, as follows:

 

 

Change Factor Only

Select this option if you want to change the current shock factor without affecting the network weights.

 

 

OK

Click this button to close the dialog box and save any changes you have made.

 

 

Cancel

Click this button to close the dialog box without saving any changes you have made.