I noticed that this released code has some discrepencies with the implementation details mentioned in the paper. For example, in the paper, the optimizer is SGD, learning rate is 0.001, total epochs is 500, while in the code, the optimizer is Adam, learning rate is 0.0001, total epochs is 1000.
Could you please explain the rationale behind these choices? Also, would be possible to use the code as it is without any modifications for fair comparisons?
Hello, congratulations on your great work!
I noticed that this released code has some discrepencies with the implementation details mentioned in the paper. For example, in the paper, the optimizer is SGD, learning rate is 0.001, total epochs is 500, while in the code, the optimizer is Adam, learning rate is 0.0001, total epochs is 1000.
Could you please explain the rationale behind these choices? Also, would be possible to use the code as it is without any modifications for fair comparisons?