Table 5 Effects of different attention spans \(\text{N}\) on experiments.

From: MMTransformer: a multivariate time-series resource forecasting model for multi-component applications

Attention heads

2

4

6

8

Metric

MSE

MAE

MSE

MAE

MSE

MAE

MSE

MAE

24

0.3995

0.4126

0.3918

0.4038

0.3994

0.4130

0.3985

0.4079

48

0.4079

0.4192

0.4013

0.4079

0.4021

0.4085

0.4031

0.4094

96

0.4412

0.4274

0.4383

0.4170

0.4398

0.4282

0.4363

0.4236

192

0.4605

0.4369

0.4522

0.4249

0.4623

0.4372

0.4650

0.4380

384

0.4702

0.4418

0.4662

0.4366

0.4777

0.4431

0.4796

0.4490

  1. Significant values are in [bold].
Back to article page