Question about the experiment results.
zw150668 opened this issue · 1 comments
zw150668 commented
Hi, sorry to bother you. I got two questions when running the code.
- The first is the same question with this
I update the RFFL_run.py in lines 361-363
rs = torch.div(rs, rs.sum())
r_threshold.append( threshold * (1.0 / len(R_set)) )
q_ratios = torch.div(rs, torch.max(rs))
Now, it's like this
rs = torch.div(rs, torch.max(rs))
rs = torch.tanh(rs * 2) # 2 is constant called the altruism degree
rs = torch.div(rs, torch.max(rs))
rs = torch.div(rs, rs.sum())
r_threshold.append( threshold * (1.0 / len(R_set)) )
q_ratios = torch.div(rs, torch.max(rs))
But it doesn't seem to work. The valid accuracy is still around 53(94). Did I misunderstand your modification to the code?
- When I run the code with cifar10 dataset, the CIFAR10 _UNI_10 accuracy is 44(44), the CIFAR10 _POW_10 accuracy is 43(47) and the CIFAR10 _CLA_10 accuracy is 29(41). But in the paper, the result is 61(62), 59(61) and 35(54).
That's my questions. Thanks.
Tecitomous commented
Hello, I met the same issues, have you already solve it?