Open russelljjarvis opened 6 years ago
Every neuron makes an appearance in at least one of the networks. The only pruning that happens is for graphical purposes to remove neurons that don't receive inputs or outputs in a selection visualization.
The raw output is below. The groupings show the neurons not displayed in each graph in the order: ii, ee, ei, ie
Given that Dan identified the connection matrices contained errors, how is analysis different now that errors have been fixed?
Are less network pruning required.
@danpetty can you still find a way to list which neurons where pruned if you get time?
Ideally no neurons that receive (even unitary) connections should get pruned, only cells that don't output projections onto other neurons should be pruned.
We are interested in neurons, whose state can at least be changed by the network, we should probably be interested in those cells, who can also change the state of others (those that have a unitary output). So I guess we probably shouldn't prune anything from the dynamic simulation.