Ours-WRN (in paper) is written as 59.06 (0.41)
andreirusu opened this issue · comments
Yet the paper says 59.60 (0.41). Which one is correct?
I have a related question. I have run the codes, and get the following results:
0.54075058430718 0.087718464622564
0.69464093959732 0.064621921214153
There is some difference of the mean, but I am more curious about the std, which is as high as 8% , while in the paper it is just around 1% .
Any idea why?
Edit: I know the reason. It is using confidence interval of 95%, which is $1.96 * \frac{\sigma}{\sqrt{n}}$
, as shown in link