MetaRF: attention-based random forest for reaction yield prediction with a few trails

Table 4 The results of ablation study(in Buchwald-Hartwig HTE dataset)

	RMSE \(\downarrow\)	\(R^{2}\) \(\uparrow\)
Baseline	15.6535	0.6538
MetaRF + Dimension-reduction	12.6401	0.7738
MetaRF + Random\(^{ a }\)	14.5454	0.7003
MAML + Dimension-reduction\(^{ b }\)	21.0006	0.3730
MAML + Random	21.0204	0.3753
Transfer learning + Dimension-reduction\(^{ c }\)	14.4575	0.7038
Transfer learning + Random	15.2232	0.6705

The bold values represent the performance of the proposed method
^aThe first ablation test, replacing the dimension-reduction based sampling with random sampling
^bThe second ablation test, removing the random forest structure
^cThe third ablation test, replacing MAML with transfer learning framework

ISSN: 1758-2946