From: Spatial attention and quantization-based contrastive learning framework for mmWave massive MIMO beam training
Number of attention modules
Total parameters
Success rate (%)
Achievable rate
1
31.0K
76.9
6.43
2
34.4K
95.1
7.37
4
41.0K
95.5
7.36
6
47.7K
8
54.3K
95.9
7.38