Thanks for the presentation. Looking forward to the presentation tomorrow!
I'm totally new to this area, just curious about the testing method. In the paper, to test the validity of the new method, you "measure performance on 100 sequentially presented digit classification tasks on the permuted MNIST digit classification task". Just curious if there's any sort of "standard" testing method or "standard" dataset when you want to prove the validity of the method (to make the results comparable to other research)?
Thanks for the presentation. Looking forward to the presentation tomorrow!
I'm totally new to this area, just curious about the testing method. In the paper, to test the validity of the new method, you "measure performance on 100 sequentially presented digit classification tasks on the permuted MNIST digit classification task". Just curious if there's any sort of "standard" testing method or "standard" dataset when you want to prove the validity of the method (to make the results comparable to other research)?
Thanks in advance!