Dynamic Learning Rate in Deep CNN Model for Metastasis Detection and Classification of Histopathology Images

Table 8

Comparison of execution time and loss.

Learning strategy	Execution time (sec)	Validation loss
CLR (triangular learning strategies) ()
Triangular	19190.43	0.1996
SGDWR (cosine annealing strategies)
	19064.76	0.2122
	18999.30	0.2765
	18998.74	0.2088
	18993.48	0.2303
SWA learning strategies
Cyclic	19117.48	0.2609
Constant	18996.33	0.2369
Conventional learning strategies
Fixed LR	19011.60	0.3298
Time-based decay	19001.54	0.3712
Step decay	19078.44	0.3891
Exponential decay	19057.34	0.3791