Doing real experiments

Structure

For highly complex experiments it is useful to split the scripts into several files. We normally try to follow the scheme below.

Folders:

Results:contains pickles of the results under nicely-named folders biases:bias (text) files of the chips setupfiles:xml files of the setups (*) chipfiles:csv and xml files representing the chips (*)

(*) these folders are usually in a common folder somewhere else because they are

not experiment dependent.

Files and correspondent actions:

expSetup:load the setup expMap:prepare connectivity expPop:create populations expStim:load data set and convert into AER expTools:useful functions, code snippets, ... expOsc:prepare acquisition of analog signals if needed expMon:define monitors (?) expRun:run the experiment and save values expData:analysis of the data expPlot:plot data

Experiment tools

We developed a specific class to save the data of our experiments.

Save

et = pyNCS.et
et.mksavedir() # creates a new folder under Results/
et.globaldata.variable_to_be_saved = v1
et.globaldata.an_other_one = v4
et.save() # saves everything from globaldata and all expXXX scripts.

Some comments. The Results folder has to be created in advance. The function mksavedir creates, by default, a folder of the format NNN__DD-MM-YYYY. You can assign every pickable variable you want to globaldata attributes. The save function saves:

• all the python files starting by exp in a single tar.gz file inside the new folder
• a single pickle of et.globaldata
• an empty file with the time of the experiment as name

Warning

Don’t forget to save chip biases!

Load

loaded_data = pyNCS.et
et.globaldata.directory = 'Results/128__19-04-2011'
et.globaldata = et.globaldata.load()

The load functions grabs the pickle from the specified directory. You can find the variables you saved in et.globaldata.