Early Recognition of Skin Malignancy in Images Based on Convolutional Networks by Using Dynamic System Model

Table 5

Table of experiments carried out with AlexNet, applying data augmentation and different subdivisions of the dataset. Best results are highlighted in bold, italic, and bold-italic, respectively.

Exp.	Basic network	Data augmentation			Proportion	Better accuracy
		Flip H	Flip V	Jitter	Train (%)/val (%)	Train	Val
1	AlexNet	No	No	No	80/20	7.73E-01	6.76E-01
2					70/30	7.78E-01	7.02E-01
3					55/45	7.74E-01	7.41E-01
4		Yes			80/20	7.76E-01	6.85E-01
5					70/30	7.74E-01	7.02E-01
6					55/45	7.74E-01	7.40E-01
7			Yes		80/20	7.72E-01	6.82E-01
8					70/30	7.74E-01	7.13E-01
9					55/45	7.72E-01	7.36E-01
10				Yes	80/20	7.45E-01	6.73E-01
11					70/30	7.44E-01	7.11E-01
12					55/45	7.44E-01	7.31E-01