Here you can set learning parameters.

**Order** determines how learning points are sorted during learning. There are severals ways to set order:
**Fixed** - Always the same order of learning points.
**Random** - Always completely different order, new permutation.
**Random Err/2** - This is not permutation! It works in the following way: select random 2 points, choose one which was wrongly classified last time, if none or both, choose first one.
**Random Err/10** - The same as above, but select random 10 points.
**Swapping Permutation** - Random permutation at start. After each epoch, two random points are exchanged.

**Premphase errors** - Before each epoch, all wrongly classified points are tested/learned.
It makes learning phase longer.

**Verify** - It works only with predefined shapes. After each epoch, verification is done. It is not learning phase. Always the same random chosen 400 points are used for verification (not visible on screen). Verification slows down the whole process, and has no effect on learning.
It only produces brown plot (see Error graphs).
If you don't want this plot, you should switch it off to make the learning faster.

English | Polski

Copyright © 2009 SharkTime Software