Table 1.

Summary of the number of nodes in each layer, validation, and testing accuracy. The architecture was selected based on the model that highest accuracy when classifying cells in the validation data set. The testing accuracy arises from training a model with the specified architecture with training  + validation datasets and testing on previously unused test set. The mean (standard deviation) of 10 replications of each model is presented for accuracy and area under curve (AUC).

	Validation	First	Second	Testing	Testing
	Accuracy	Layer	Layer	Accuracy	AUC
No	94.8	100	50	94.9	0.999
Dropout	(0.457)			(0.248)	(1.28e−4)
Dropout	93.5	100	75	94.1	0.996
Only	(0.259)			(0.312)	(3.74e−4)
Dropout +	92.6	75	100	92.4	0.998
${\ell }_1$	(0.461)			(0.517)	(2.13e−4)
Dropout +	93.9	100	75	93.4	0.999
${\ell }_2$	(0.450)			(0.244)	(1.28e−4)