My CA Model
- Other models: Hetherington (93), Rochester (56), Bower's Genesis
- Essential Features of my model
- Neurons are "leaky integrators"
- Connections are distance-biased.
- Neurons fatigue
- Neurons are excitatory or inhibitory but not both.
- All learning is local: Hebbian or Anti-Hebbian
- Current Properties
- Time is discrete. Each cycle is supposed to map to 10 ms.
- The topology is toroidal. Currently using 400-900 neurons with
each neuron connected to roughly 20 others.
10,000 neuron simulations run 100 cycles per second on a PC.
- Neurons may spontaneously activate.
