Table 4 Multi-head attention configuration study.

From: Air quality prediction using multi-source remote sensing data integration with hybrid deep learning framework

Attention heads

PM2.5 RMSE

NO2 RMSE

Attention diversity

1

9.4

7.8

0.12

2

8.9

7.3

0.24

4

8.5

7.0

0.41

8

8.2

6.7

0.58

16

8.3

6.8

0.52

32

8.4

7.0

0.48