DimredRun

This class can be used in order to comfortably represent multiple D3plot samples in a 3D graph through dimensionality reduction. In the 3D graph every simulation is a data point and the closeness of the data points represents the similarity of the result field. It offers functions to:

• Subsample simulations to reduce the computational effort and account for different meshing.
• Reduce order which is the black magic reducing complex field results to a few numbers which can then be plotted in 3D.
• Clustering with sklearn to group simulations semi-automatically (mostly just a visual feature).
• Output 3D similarity plot as a webpage

For ease of use, check out the Tool section, which explains the command line tool for this dimensionality reduction feature.

Class to control and run the dimensionality reduction process

__init__(simulation_runs, start_stage, end_stage, project_dir, html_name='3d-beta-plot', html_set_timestamp=False, reference_run=None, console=None, img_path=None, exclude_runs=None, logfile_filepath=None, n_processes=1, part_id_filter=None, timestep=-1, show_output=True, cluster_args=None, outlier_args=None)

Class handling a dimensionality reduction

Parameters:

Name	Type	Description	Default
simulation_runs	Sequence[str]	simulation runs to analyze	required
start_stage	str	where to start	required
end_stage	str	where to stop	required
project_dir	Union[None, str]	required project directory for creation of buffer files. Allows restart in between.	required
html_name	str	Name of the output .html file	'3d-beta-plot'
html_set_timestamp	bool	If true, the output .html will include a timestamp (hh_mm_ss) at the end of the filename	False
reference_run	str	filepath to the reference run. If not set, first entry in simulation_runs will be used as reference run.	None
console	Union[Console, None]	Console for information printing and logging. Rich offers pretty text printing, syntax highlighting etc.	None
img_path	Union[None, str]	optional image directory to show images in visualization.	None
exclude_runs	Union[None, Sequence[str]]	optional list of runs to exclude from processing	None
logfile_filepath	Union[str, None]	path of the log file (always appends)	None
n_processes	int	number of processes to use during execution	1
part_id_filter	Union[None, Sequence[int]]	which part ids to process	None
timestep	int	specifies timestep to analyze in clustering and show in output visualization	-1
show_output	bool	Set to false not to show the output html in the browser	True
cluster_args	Union[None, Sequence[str]]	Arguments for cluster algorithm	None
outlier_args	Union[None, Sequence[str]]	Arguments for outlier detection algorithm	None

Notes

Using a project directory allows to restart stages of the entire
process.

process_reference_run()

Process the reference run

subsample_to_reference_run()

Subsamples all runs

dimension_reduction_svd()

Calculate V_ROB and Betas

clustering_results()

clustering results

visualize_results()

creates an output .html file