Batch simulations using 'evol' method

[Maliha A]

Hi, 

Just wondering how the batch.py file from Tutorial 8 can be modified if I want to use the evolutionary parameter optimization method instead of a grid search?

Thank you
Maliha

[J C]

https://github.com/suny-downstate-medical-center/netpyne/blob/291af6be4f08307213f25be909a574e5317f35ed/examples/evolutionary/src/batchRun.py#L49-#L96

something similar to this will need to be implemented for Tutorial 8.

including at least these major steps:

specifying that the batch method is 'evol' here 

defining a fitness function (which will be able to reference anything from the generated sim pkl/json file here

defining the evolutionary config parameters here

Best,

James

[Maliha A]

Thanks James, this is helpful!

Regards

Maliha

[Maliha A]

When the batch method is set to 'evol', is my understanding correct in that the the following lists are parameter ranges? 

params = specs.ODict()

params['prob'] = [0.01, 0.5]
params['weight'] = [0.001, 0.1]
params['delay'] = [1, 20]

[J C]

Sorry, yes you are correct--

this will sample: 

0.01 < prob < 0.5

0.001 < weight < 0.1

1 < delay < 20