rei-forks/precise-lite-amd64aarch64
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Increase default weight of weighted_log_loss to 0.7

Browse Source 
No point in using it by default if it isn't weighted. This makes it reduce false positives by default
This commit is contained in:
Matthew D. Scholefield 2018-04-18 23:11:54 -05:00
parent bc7c73a4a9
commit 0cbb98fa78
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
precise/functions.py
View File

@ -21,7 +21,7 @@ def weighted_log_loss(yt, yp) -> Any:
yp: Prediction
"""
from keras import backend as K
weight = 0.5 # [0..1] where 1 is inf bias
weight = 0.7 # [0..1] where 1 is inf bias
pos_loss = -(0 + yt) * K.log(0 + yp + K.epsilon())
neg_loss = -(1 - yt) * K.log(1 - yp + K.epsilon())