|
Alphabet
|
% Acc. on CN dataset
|
% Acc. on RSA dataset
|
|---|
|
AA
|
74.0 ± 0.6
|
70.7 ± 0.4
|
|
DualRMI
|
73.3 ± 0.5
|
70.3 ± 0.4
|
|
WW5
|
73.1 ± 0.7
|
69.6 ± 0.4•
|
|
SR5
|
73.1 ± 0.7
|
69.6 ± 0.4•
|
|
MU4
|
72.6 ± 0.7•
| 69.4 ± 0.4•▼ |
|
MM5
|
73.1 ± 0.6
| 69.3 ± 0.3•▼ |
|
HD1
|
72.9 ± 0.6
| 69.3 ± 0.4•▼ |
|
HD2
|
73.0 ± 0.6
| 69.3 ± 0.4•▼ |
|
HD3
|
73.2 ± 0.6
|
69.9 ± 0.4•
|
- A • marks reduced datasets where BioHEL performs significantly worse than the AA type dataset according to the t-tests with a 99% confidence level. A ▼ marks the alphabets that performed significantly worse than the DualRMI strategy.
