MCPcopy Index your code
hub / github.com/HiLab-git/SSL4MIS / get_comp_loss

Function get_comp_loss

code/train_fixmatch_standard_augs.py:148–177  ·  view source on GitHub ↗

get complementary loss and adaptive sample weight. Compares least likely prediction (from strong augment) with argmin of weak augment. Args: weak (batch): weakly augmented batch strong (batch): strongly augmented batch Returns: comp_loss,

(weak, strong)

Source from the content-addressed store, hash-verified

source not stored for this graph (policy: none)

Callers 1

trainFunction · 0.85

Calls 1

logMethod · 0.80

Tested by

no test coverage detected