Added a new manually executable Git Actions workflow for generating benchmark tables for SPI testing called run_datset_generation.yaml. Upon running (via Git Actions portal), the CML7.npy dataset will be loaded and SPIs will be computed on the same dataset for a fixed number of iterations (using a new Calculator each time). The mean and standard deviation for each SPI's output matrix (MPI) is then saved as a dictionary and uploaded as a .pkl file that can be downloaded as an artefact in Git Actions. Each time a new SPI is added, or a major change (e.g., dependency update) is made, the workflow can be manually triggered to create a new baseline set of SPI outputs.
Added a new pairwise distance measure - Gromov-Wasserstain (GWTau) distance. See here for more details about the specific implementation.
Added a new in-depth tutorial for integrating SPIs into the pyspi framework using GWTau as an example. See the docs for a full explanation of the validation/testing process of this SPI using the dataset provided in the original research paper.
Added new benchmark dataset tables which include reference outputs for the new GWTau SPI.
Updated the default (config.yaml) and fast (fast.yaml) subset configuration files to include GWTau.
In this PR, I've made the following changes:
run_datset_generation.yaml. Upon running (via Git Actions portal), theCML7.npydataset will be loaded and SPIs will be computed on the same dataset for a fixed number of iterations (using a new Calculator each time). The mean and standard deviation for each SPI's output matrix (MPI) is then saved as a dictionary and uploaded as a .pkl file that can be downloaded as an artefact in Git Actions. Each time a new SPI is added, or a major change (e.g., dependency update) is made, the workflow can be manually triggered to create a new baseline set of SPI outputs.GWTau) distance. See here for more details about the specific implementation.config.yaml) and fast (fast.yaml) subset configuration files to includeGWTau.