Automatic Prediction of Meningioma Grade Image Based on Data Amplification and Improved Convolutional Neural Network

Table 3

Impact of different activation functions on network test results.

	Sigmoid	Tanh	ReLU	ELU
Characteristics	Gradient disappears	Convergence speed is faster than Sigmoid; gradient disappears	The input is positive, the gradient does not disappear; the input is negative, the gradient disappears.	It combines sigmoid and ReLU; and gradient disappears
Test accuracy rate	70.00%	56.67%	76.67%	83.33%