Table 6 Active learning results
Target | Model | Sampling | Standard | Generalization | ||
|---|---|---|---|---|---|---|
Whole | OOD Data Bin | Whole | ID Bins | |||
Solubility | MDM | Data Density (EB) | 0.14% | 0.25% | 0.08% | 0.03% |
\(\varvec{p=0.015}\) | \(\varvec{p=0.013}\) | \(p=0.146\) | \(p=0.326\) | |||
Diversity (EB) | 0.00% | 0.13% | −0.03% | −0.07% | ||
\(p=0.483\) | \(p=0.098\) | \(p=0.674\) | \(p=0.793\) | |||
MCDO | 0.13% | 0.08% | 0.10% | 0.11% | ||
\(\varvec{p=0.032}\) | \(p=0.204\) | \(\varvec{p=0.036}\) | \(\varvec{p=0.024}\) | |||
OOD Only | – | 1.93% | 0.18% | −0.28% | ||
– | \(\varvec{p<0.001}\) | \(\varvec{p=0.007}\) | \(p=0.992\) | |||
Solubility | GBM | Data density (EB) | −0.01% | 0.24% | 0.04% | −0.03% |
\(p=0.714\) | \(\varvec{p<0.001}\) | \(p=0.084\) | \(p=0.868\) | |||
GBM | 0.06% | 0.11% | 0.05% | 0.03% | ||
\(\varvec{p=0.002}\) | \(\varvec{p=0.012}\) | \(\varvec{p=0.045}\) | \(p=0.203\) | |||
OOD only | – | 1.82% | 0.14% | −0.31% | ||
– | \(\varvec{p<0.001}\) | \(\varvec{p=0.009}\) | \(p=0.996\) | |||
Redox | MDM | Data density (EB) | −0.09% | 0.42% | 0.06% | −0.05% |
\(p=0.986\) | \(\varvec{p<0.001}\) | \(p=0.055\) | \(p=0.900\) | |||
MCDO | 0.01% | 0.18% | 0.05% | 0.01% | ||
\(p=0.377\) | \(\varvec{p=0.001}\) | \(p=0.102\) | \(p=0.487\) | |||
OOD only | – | 2.09% | 0.41% | −0.10% | ||
– | \(\varvec{p<0.001}\) | \(\varvec{p<0.001}\) | \(p=0.977\) | |||
Redox | GBM | Data density (EB) | 0.02% | 0.21% | 0.06% | 0.01% |
\(p=0.168\) | \(\varvec{p<0.001}\) | \(\varvec{p=0.003}\) | \(p=0.265\) | |||
GBM | 0.02% | 0.12% | 0.05% | 0.02% | ||
\(p=0.123\) | \(\varvec{p=0.007}\) | \(\varvec{p=0.013}\) | \(p=0.076\) | |||
OOD only | - | 1.60% | 0.11% | −0.29% | ||
- | \(\varvec{p<0.001}\) | \(\varvec{p=0.025}\) | \(p=0.994\) | |||