Another module of this framework provides visualization capabilities for previously performed experiments. The connection to the experiment which should be post processed is done via the experiment folder provided to the class constructor.
import phs.post_processing # standalone import
post_pro = phs.post_processing.PostProcessing(experiment_dir='/absolute/path/to/an/experiment')Right now four different methods are available. All generated output will be stored inside the linked experiment folder, where a directory named 'post_processing' is created. Each method expects a parameter name to give unique file names.
The plot_3d method creates a three dimensional plot with individual choice of the dimensions. Therefor first, second and third need to get a parameter name used in the linked experiment. Note that result is valid for addressing the returned value from the target function. The third dimension is color encoded. Interpolated contour lines can be switched on with the flag contour. An animated .gif can be created with the number of samples animated_step_size added in a single step and animated_fps frames per second.
post_pro.plot_3d(name='name_of_generated_file',
first='parameter_name_of_first_dimension',
second='parameter_name_of_second_dimension',
third='parameter_name_of_third(color)_dimension',
contour=False,
animated=False,
animated_step_size=1,
animated_fps=1)The result (returned by the target function) is plotted over the indices of the parameter sets.
post_pro.result_evolution(name='name_of_generated_file')For each machine involved, a timeline is drawn with the parameter sets evaluated by it.
post_pro.create_worker_timeline(name='name_of_generated_file')For each parameter defined in the experiment, a seperate one dimensional scatter plot is drawn. The points belonging to the same parameter set are connected via lines. These lines show the color coded result.
post_pro.create_parameter_combination(name='name_of_generated_file')def main():
import phs.parameter_definition # standalone import
import phs.experiment_definition # standalone import
import phs.compute_definition # standalone import
import phs.post_processing
pardef = phs.parameter_definition.ParameterDefinition()
pardef.set_data_types_and_order([('x', float), ('y', float)])
pardef.add_individual_parameter_set(
number_of_sets=20,
set={'x': {'type': 'random', 'bounds': [-5, 5], 'distribution': 'uniform', 'round_digits': 3},
'y': {'type': 'random', 'bounds': [-5, 5], 'distribution': 'uniform', 'round_digits': 3}},
prevent_duplicate=True)
pardef.add_individual_parameter_set(
number_of_sets=20,
set={'x': {'type': 'bayesian', 'bounds': [-5, 5]},
'y': {'type': 'bayesian', 'bounds': [-5, 5]}})
expdef = phs.experiment_definition.ExperimentDefinition(
experiment_dir='/absolute/path/to/not/yet/existing/folder/your/experiments/should/be/saved',
target_module_root_dir='/absolute/path/to/root/dir/in/which/your/test_function/resides',
target_module_name='file_name_with_test_function_definition_(without_extension)',
target_function_name='test_griewank',
parameter_definitions=pardef.get_parameter_definitions())
compdef = phs.compute_definition.ComputeDefinition(
experiment_dir='/absolute/path/to/folder/with/existing/experiment',
parallelization='local_processes',
local_processes_num_workers=3,
redirect_stdout=False,
provide_worker_path=False,
bayesian_wait_for_all=True)
compdef.start_execution()
post_pro = phs.post_processing.PostProcessing(
experiment_dir='/absolute/path/to/an/experiment')
post_pro.plot_3d(name='plot_xyr_post',
first='x',
second='y',
third='result',
contour=True,
animated=True,
animated_step_size=1,
animated_fps=1)
post_pro.result_evolution(name='evo_post')
post_pro.create_worker_timeline(name='worker_timeline_post')
post_pro.create_parameter_combination(name='parameter_combination_post')
if __name__ == "__main__":
main()
