Existing FER methods typically report overall accuracy on highly imbalanced test sets but exhibit low performance in terms of the mean accuracy across all expression classes.

Recentstudieshaveshownthat the transferability of adversarial examples exists for CNNs, and the same holds true for ViTs.

However, these models suffer from quadratic computational cost in the input sequence lengthn to compute pairwise attention in each layer.

Wethenseparate theoptimization process into two steps, corresponding to weight update and structure parameter update. For the former step, we use the conventional chain rule, which can be sparse via exploiting the sparse structure.