 This paper proposes a new method called symmetric deep convolutional adversarial network, SDCN. It combines convolutional neural networks, CNNs, and adversarial learning to extract and transfer invariant and discriminative features from electroencephalogram, e.g., signals in order to accurately classify different levels of mental stress. The proposed method outperforms existing methods in terms of both accuracy and generalization ability. This article was authored by Rui Qifu, Yifong Chen, Yongqi Huang, and others.