In https://github.com/PPPLDeepLearning/plasma-python/pull/22 we have applied clip(0, infinity) in the beginning of Normalizer.apply method for all normalizers, now, we also clip the signal further to the statistical range clip(-N,N) for all *Var* normalizers (Var, MeanVar, AveragingVar)
The value of N which defines the statistical range is defined in the config file, and is a tunable parameter.
The results for tuning N for Var normalizer (only) follow:
This PR extends https://github.com/PPPLDeepLearning/plasma-python/pull/22
In https://github.com/PPPLDeepLearning/plasma-python/pull/22 we have applied
clip(0, infinity)
in the beginning ofNormalizer.apply
method for all normalizers, now, we also clip the signal further to the statistical rangeclip(-N,N)
for all*Var*
normalizers (Var
,MeanVar
,AveragingVar
)The value of
N
which defines the statistical range is defined in the config file, and is a tunable parameter.The results for tuning N for
Var
normalizer (only) follow: