Docs

https://neurord.github.io/ajustador/

Code

https://github.com/neurord/ajustador/

Setting up Optimizations on the Neuroscience Gate (NSG)

• create directory, such as my_NSGopt_dir, with the following files

  • main optimization scripts, example is gpNpas_opt.py, in https://github.com/neurord/optimization_scripts/blob/master/gp_opt/, which

    • specifies model type (moose_nerp package)

    • specifies neuron type

    • specifies data name (i.e., which experimental trace)

    • specifies optimization parameters, such as generations and population size

    • creates and moves to output directory according to some naming convention, e.g.

      rootdir=os.getcwd()+'/output'
      dirname='cmaes_'+dataname+'_'+str(seed)+'_'+str(popsiz)
      if not in dirname in os.listdir(rootdir):
         os.mkdir(rootdir+dirname)
      os.chdir(rootdir+dirname)
      

  • fit_commands.py

  • param_fitness_chan.py, which specifies

    • which parameters to change, and their ranges
    • the fitness function, and weights on fitness features

• in that same my_NSGopt_dir:

  • create empty /output subdirectory

  • copy (or link)

    • moose_nerp/moose_nerp
    • dill module
    • adjustador/ajustador
    • make sure __init__ in ajustador does not import xml.py, loadconc.py, nrd_fitness.py, drawing.py, nrd_output.py
    • make sure ajustador/helpers/save_params.py does not import xml

  • copy (or link) the data directory, and python file that specifies the data as class Params, e.g. from waves:

    • https://github.com/neurord/waves/blob/master/gpedata_experimental.py
    • https://github.com/neurord/waves/tree/master/gpedata-experimental

• create a zip file from the directory _above_ the directory you just created

  zip -r NSGopt.zip my_NSGopt_dir/