Convolutional Neural Networks for Recognition of Lymphoblast Cell Images

Table 4

Summary of parameter settings for the ConVNet, SVM-GA, MLP, and random forest.

Methods	Parameters	Setting
ConVNet	Filter size (convolutional layer)	3 × 3
	Filter size (max pooling layer)	2 × 2
	Batch size	121
	Epoch	50
	Learning rate	0.001
SVM-GA	Population size	100
	Number of generations	100
	Probability of crossover	0.80
	Probability of mutation	0.06
	Rate of elitism	0.05
		20
		42
MLP	Number of hidden layers	1
	Number of neurons in hidden layer	69
	Activation function	Logistic
	Batch size	121
	Epoch	10000
	Learning rate	0.001
	Momentum	0.7
Random forest	Number of classifiers	100
	Maximum depth	2