Number of runs too small when using ParameterSpace

[leotozzi88]

Dear Experts,

I would like to run simulations exploring a parameter space for an ALN model.

I have read your examples online and I have managed to write my model evaluation function and run the search as follows:

parameters=ParameterSpace({"Ke_gl":[0.1, 500.0], "sigma_ou":[0.2, 0.3]})
search=BoxSearch(aln, parameters, evaluate)
search.run()

When I run this command, I get:

MainProcess root INFO     Number of processes: 8
MainProcess pypet.storageservice.HDF5StorageService INFO     I will use the hdf5 file `./data/hdf/exploration.hdf`.
MainProcess pypet.environment.Environment INFO     Environment initialized.
MainProcess root INFO     Number of parameter configurations: 2
MainProcess root INFO     BoxSearch: Environment initialized.

The simulations run, but I only get 2 runs of the simulation, whereas I would expect to have 4 (4 possible combinations of parameters).

I might be missing something obvious, but how can I use ParameterSpace to explore all possible combinations of parameters?

Thank you,

Leonardo Tozzi

[jajcayn]

Dear @leotozzi88,

For simple grids with only 2 values per parameter, neurolib is trying to automatically infer what kind of space you want to explore and incorrectly assumes those are just bounds. This should work:

...
parameters=ParameterSpace({"Ke_gl":[0.1, 500.0], "sigma_ou":[0.2, 0.3]}, kind="grid")
...

Thanks for using neurolib:)

N.

[leotozzi88]

Oh wow the library is too smart! :D

Thank you!

Leonardo Tozzi