API documentation: | spikes | analysis | benchmark | parameters | stgen | utilities | facets |

NeuroTools.benchmark

Data

__name__ = "NeuroTools.benchmark"

Functions

get_experiment_dict(params)

Takes params = dict with vectors of all parameters values Calculates cross product of all and returns a dict with all experiments.

Classes

Benchmark

__init__(self, filename, model, experiments)

Generates the benchmark files at the specified name

A benchmark is a list of experiments to test the effect of parameter changes on a model. It is defined by: - filename is the folder where the benchmark is run - model is the Model object - params is its operating point (a dict) - experiments is the list of experimental parameters (in a dict) that change and producing the list of experiments

execute(self, experiment_list)

Execute all experiments from experiment_list.

This should be a batch of experiments to launch on one core.

experiment_filename(self, experiment="")

Returns the filename for an experiment.

This will generate one file per experiment and a default

TODO : use a more sexy filename with:

name=""
for key in data_root['run'][experiment].keys():
  name = name+"_"+key+"_"+str(data_root["run"][experiment][key])

get(self, key="", experiment="")

Gets dictionary at key from experiment file.

opens the main file if experiment = ""

get_figures(self)

Get figure list

list_experiments(self)

Returns the list of experiments in the benchmark

list_param(self, param)

Lists values for param in experiments = the reverse of get_experiment_dict

list_unfinished_experiments(self)

returns a list of the experiments in the benchmark that did not end

list_where(self, param, value)

Lists experiments where param is equal to value.

put(self, key, data, experiment)

Puts dict at key in experiment file.

run_simulations(self)

Runs the benchmark simulations

The strategy is applied: - only simulate what was not already simulated (did your process break?) - run the different independent simulations on different ipython1 engines and then gather all data

show_experiments(self)

Show a nice list of experiments

show_figures(self, list=[])

Show a list of figures saved as pylab objects experiments = <property object at 0x48bf80>


Model

Base class for standardized models.

A model is a system that has some parameters (params) and outputs data in a dict (out) thanks to a python call or TODO a standard unix command (cmd).

TODO: define more exactly the standard

__init__(self, params, cmd)

run(self)