Cumulative normalization, wrong variance computation
zmolikova opened this issue · comments
Hi,
I suspect that the computation of variance in CumLN
is missing a division by the length of the sequence. That is:
asteroid/asteroid/masknn/norms.py
Line 101 in 140733b
should be
cum_var = cum_pow_sum / cnt - cum_mean.pow(2)
as in
You're fully right, thank you !
Could you please submit a PR ?