Memory controlled sequential self attention for sound recognition
Each of the Figure below visualizes self attention plots obtained using frame-level pairwise cosine similarity of the Memory Controlled Self Attention feature embeddings of different models.
Input-MelSpec: pairwise cosine similarity of input mel-spectrogram feature
Baseline_RNN: pairwise cosine similarity of the GRU feature embeddings of the baseline model
SelfAttn: pairwise cosine similarity of the self attention feature embeddings of the model without memory control
SelfAttn_L: pairwise cosine similarity of the memory controlled self attention feature embeddings of the model with attention width L
MultiHead_SelfAttn: pairwise cosine similarity of the memory controlled self attention feature embeddings of the MultiHead model
Figure 1
Figure 2
Figure 3
Figure 4
Memory controlled multi head self attention is developed using Keras Self-Attention.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent