Performance Psychonet can tend to be extremely CPU intensive due to the sheer volume of data that is being computed (especially during training). Here are some performance tips:
Shut down everything that you can, except for Psychonet.
Turn off memory monitor using Display menu
Set number of iterations per cycles (Training tab of preference panel ) to higher value. This will slow down user feedback but will increase performance of the net as a whole.
The smaller the number of input parms, the more generalized the network will be. The more generalized the network is, the better the pattern recognition will be.
Current error on training display value will start out large, but should quickly (within 100 cycles) start approaching 0.
There is a point of diminishing returns on training where additional training provides very little gain. If this occurs, you have exceeded capacity of the network and need to add additional Hidden Layers.
Start Small - When first starting to train your data, use small data sets and train for a small number of iterations. Once you have ran smaller tests with some success, then increase the volume of data to
larger values as needed.
