| < Previous page | Next page > |
Trimming
Trimming.
An EasyNN-plus network can be trimmed while it is learning.
Pruning weights, disconnecting input nodes and excluding example rows can be enabled while the network is learning.
Pruning has a low weight threshold. If the weight is below the threshold for the number of cycles it will be no longer be updated during learning.
Input nodes can be set to disconnect if they are the least important for the number of cycles.
Example rows that have the greatest error for the number of cycles can be excluded.
|
Check to enable connection pruning when the weight is below the level set for the number of learning cycles. 
Disconnect the input node if it is the lease important for the number of cycles.
Exclude the example row if it has the greatest for the number of cycles.
Click to accept all settings and close the dialog.
Click to reject all settings and close the dialog.