Results
- I worked for years with people (the ANT group at the U of
Michigan).
- I decided it was time to actually start implementing something.
- Small number of neurons ( n < 1000)
- Small number of axons (for each neuron a < 30)
- No learning
- Varied connectivity, connectivity strength, activation threshold,
recovery rate, fatigue, and decay rate.
- No snoopy curve with 400 neurons.
- Snoopy Curve with 900 in several places.
- Learning
- I would have been working on learning, but I wrote this talk.
- I'm hoping learning will solve this problem.
- After about 3 hours working on learing, I did manage ongoing
activation via learning, but not a snoopy curve.
- Clearly, this is work in progress.
