Fig. 4 | Scientific Reports

Fig. 4

From: Multichannel convolutional transformer for detecting mental disorders using electroancephalogrpahy records

Fig. 4

Histogram of the performances of the model without some specific modules. The x-axis is the model configuration, the y-axis is the accuracy. The label ’no_conv’ stands for no convolution layer, the label ’no_seq_pool’ stands for no sequence pooling module, and the label ’no_fusion’ stands for no fusion module. This experiment was conducted on the MODMA dataset. The most important module seems to be the convolutional layers.

Back to article page