Jets and Sharks
|
|
- We reimplemented McClelland's (1981) Jets and Sharks simulation.
- He used nodes that had variable strength.
- Our simulation used orthogonal
CAs of 200 fLIF neurons.
- Each feature had its own subnet with distance biased
connectivity within for excitatory neurons, and random
connectivity for inhibitory neurons and between nets.
- The individual features (the CAs) were learned first.
- Items were copresented (e.g. Art, Jet, 40's, Junior High,
Single, Pusher) afterward.
