Accuracy problem #1
Description
When I was attempting to reproduce the experimental results of layer == 9 in RQ2 of the paper, I encountered a problem that the experimental results could not reach the accuracy mentioned in the paper. After running the experiment multiple times and trying to disable the early stopping strategy, the best result I obtained was func = 0.2549, which was far from the 0.2272 mentioned in the paper. Other parameters and environments were basically the same. Could you please explain why the paper's results could not be reproduced? Are there any other parameters that need to be adjusted? Also, I would like to confirm again: Is the indicator in RQ2 func instead of prob as in RQ1?