Table 6 Effect of branching on model accuracy and parameters by changing the ___location and distribution of branching for a single occurrence of branching.

From: Improving deep learning model performance under parametric constraints for materials informatics applications

Branching

Branching

\(\#\)Model

\(\#\)Model

Dataset

Location

Distribution

Size (MiB)

Parameters

OQMD

AFLOWLIB

MP

JARVIS

Top

[2, 2]

18.3

4,548,385

0.041

0.048

0.106

0.070

 

[3, 1]

18.3

4,548,385

0.043

0.049

0.113

0.069

 

[2, 1, 1]

14.8

3,676,961

0.041

0.048

0.107

0.068

 

[1, 1, 1, 1]

11.3

2,805,537

0.041

0.047

0.108

0.069

Middle 1

[2, 1]

24.9

6,206,753

0.041

0.048

0.113

0.068

 

[1, 1, 1]

28.1

6,993,697

0.041

0.048

0.111

0.070

Middle 2

[2, 1]

22.6

5,616,673

0.040

0.048

0.111

0.069

 

[1, 1, 1]

23.4

5,813,537

0.042

0.048

0.113

0.071

Middle 3

[2, 1]

22.0

5,469,089

0.041

0.049

0.108

0.069

 

[1, 1, 1]

22.2

5,518,369

0.041

0.049

0.114

0.070

Bottom

[1, 1]

21.8

5,432,161

0.041

0.047

0.112

0.067

Back to article page