search

ioam / topographica

A general-purpose neural simulator focusing on topographic maps.
topographica.org
BSD 3-Clause "New" or "Revised" License
53 stars 32 forks source link

from snapshot to output the parameters used in simulation #685

Closed dancehours closed 6 years ago

dancehours commented 6 years ago

Dear Bednar,

if I load a snapshot saved before, for example,GCAL model running 20000 times, is there a way to know the parameters I used in that simulation ? e.g. the size of area I used.

jbednar commented 6 years ago

After loading it back in, you can try saving the "script_repr" of the simulation in memory, which will write out a Python text file showing how the objects can be recreated. topo.sim.script_repr(), I think?

dancehours commented 6 years ago

Yes,it shows all the parameters of the simulation.Thanks !

dancehours commented 6 years ago

After I load a snapshot, could I continue running the model by using a different parameter, e.g. learning rate ? Before it is 0.1 and now I change it to 0.2 to test.

jbednar commented 6 years ago

Some parameters only take effect during initialization, such as any of the size parameters for a CFProjection; there is no meaning to changing those after the weights have been initialized and where practical we try to mark those values constant so that you get a useful error message if you try. But most other parameters, including all learning rates, thresholds, and similar parameters apply to every iteration, and can thus be changed whenever you like.