Categorisation
- This network is set up to categorise something as one of
two classes.
- The classes are the top or the bottom.
- What should happen is that a new pattern is presented (randomly
selected).
- Either the top half or the bottom half of the net will
come on, and neurons will continue firing until the end
of the epoch.
- CANT of course learns this.
- As it's set, it won't learn it.
- You need to change some parameters.
- I got it to work by changing fatigue rate from 0.8 to 0.1.
- You do this by
- Selecting the Set Parameters/Fatigue Rate menu item.
- The old value should appear in the text box.
- Change the value to the one you want.
- Hit the Enter Parameter button
- You can select the menu item again to make sure it's changed.
- Now run the system for a while.
- You might want to change Execution/Set Run Delay to make
it go faster.
- Mine worked pretty well at 2300 cycles.
