Fig. 4
From: MEMCAIN: a memory-enhanced hybrid CNN-attention model for network anomaly detection

The feature fusion block, which consists of three components: Concat, Multi-Dilated local Attention, and flatten.
From: MEMCAIN: a memory-enhanced hybrid CNN-attention model for network anomaly detection

The feature fusion block, which consists of three components: Concat, Multi-Dilated local Attention, and flatten.