Experimental Results of Participant 1 (P1)

The table below presents a summary of results for the final task of Participant 1 (P1). It consists of three tables, each illustrating the accuracy of all classes, new classes, old classes, and computation time across different holdout sizes in each scenario for each dataset.

The evaluated methods include Random (Rand), EWC-Replay (EWC-R), iCaRL, LUCIR, TaskVAE (with and without filtering), and TaskVAE-GMM.

For a larger view, right-click on the image and select "Open in a new window."

P1 Results

All Accuracy

New Accuracy

Old Accuracy

Computation Time