Question regarding the average and last accuracy.

[vgthengane]

Dear authors,

Thank you for open-sourcing your work. I am slightly confused about the metrics you guys used for the evaluation. Here is my understanding from your readme.

1. accuracy_n: accuracy evaluation on only the n-th task after training for the n-th task.
2. forgetting: Average forgetting up until the current task.
3. avg_acc: Average evaluation accuracy after training for the n-th task.

for (3), after training for the n-th task, we got,
acc_per_task = [a1, a2,...an];
avg_acc = average(acc_per_task);

Is it right?

[KingSpencer]

Yeah, I think your understanding is correct:)