Output Results (CUL)

This page summarizes the output results of continual unlearning experiment and evaluation pipelines. Their existence, file or folder names, and formats can be customized.

Output Results of Main Experiment and Evaluation

The output results of continual unlearning main experiment and continual unlearning main evaluation is exact the same as the output results of continual learning main experiment or continual learning main evaluation. Please refer to Output Results (CL). The difference is that some tasks are requested to be unlearned during the process.

Output Results of Full Experiment and Evaluation

The following output results are produced after running continual unlearning full experiment or continual unlearning full evaluation, in addition to output results of main experiment and evaluation.

Folder or File	Description	Customization
<output_dir>/refretrain/	Contains all output results of the reference retrain experiment	Cannot be customized
<output_dir>/reforiginal/	Contains all output results of the reference original experiment	Cannot be customized
<dd_save_dir>/	Contains all data and figures of DD metrics	To include or exclude these outputs, enable or disable CULDistributionDistance metric callback (see Configure Metrics) <dd_save_dir>: Field dd_save_dir in CULDistributionDistance. It is recommended to be set as ${output_dir}/results/ to make sure outputs are under <output_dir>/ Cannot be customized in full experiment
<dd_save_dir>/dd.csv	The original data of DD metrics, stored as CSV format	dd.csv: Field dd_csv_name in CULDistributionDistance. Cannot be customized in full experiment. Default is dd.csv
<dd_save_dir>/dd_plot.png	A plot of DD	dd_plot.png: Field dd_plot_name in CULDistributionDistance To exclude this output, set this field to null or remove this field Cannot be customized in full experiment
<ad_save_dir>/	Contains all data and figures of AD metrics	To include or exclude these outputs, enable or disable CULAccuracyDifference metric callback (see Configure Metrics) <ad_save_dir>: Field ad_save_dir in CULAccuracyDifference. It is recommended to be set as ${output_dir}/xxx/ to make sure outputs are under <output_dir>/ Cannot be customized in full experiment
<ad_save_dir>/dd.csv	The original data of AD metrics, stored as CSV format	ad.csv: Field ad_csv_name in CULAccuracyDifference. Cannot be customized in full experiment. Default is ad.csv
<ad_save_dir>/ad_plot.png	A plot of AD	ad_plot.png: Field ad_plot_name in CULAccuracyDifference To exclude this output, set this field to null or remove this field Cannot be customized in full experiment