Table 5 Results of the Buchwald-Hartwig dataset under reactant-based out-of-sample conditions
From: Prediction of chemical reaction yields with large-scale multi-view pre-training
Split type | Measure | YieldBERT | YieldBERT-DA | UA-GNN | ReaMVP |
|---|---|---|---|---|---|
Halide Br | MAE | \(7.882\pm 0.311\) | \(8.431\pm 0.415\) | \(7.336\pm 0.824\) | \(\mathbf {7.118\pm 0.873}\) |
| Â | RMSE | \(11.180\pm 0.379\) | \(12.457\pm 0.508\) | \(10.185\pm 1.334\) | \(\mathbf {10.034\pm 1.126}\) |
| Â | R\(^2\) | \(0.803\pm 0.013\) | \(0.756\pm 0.020\) | \(0.834\pm 0.044\) | \(\mathbf {0.840\pm 0.037}\) |
Halide Cl | MAE | \(18.727\pm 2.130\) | \(17.769\pm 0.735\) | \(26.822\pm 1.243\) | \(21.664\pm 1.588\) |
| Â | RMSE | \(25.184\pm 2.781\) | \(21.253\pm 0.571\) | \(33.169\pm 0.783\) | \(25.881\pm 0.936\) |
| Â | R\(^2\) | \(-0.434\pm 0.316\) | \(-0.010\pm 0.054\) | \(-1.459\pm 0.117\) | \(-0.498\pm 0.107\) |
Halide I | MAE | \(10.359\pm 0.422\) | \(9.201\pm 0.419\) | \(15.950\pm 2.924\) | \(\mathbf {8.877\pm 0.254}\) |
| Â | RMSE | \(14.388\pm 0.398\) | \(\mathbf {12.419\pm 0.631}\) | \(20.793\pm 3.359\) | \(13.084\pm 0.314\) |
| Â | R\(^2\) | \(0.676\pm 0.018\) | \(\mathbf {0.758\pm 0.025}\) | \(0.307\pm 0.225\) | \(0.732\pm 0.013\) |
Pyridyl | MAE | \(17.443\pm 1.009\) | \(18.406\pm 0.480\) | \(\mathbf {16.946\pm 0.214}\) | \(17.172\pm 0.833\) |
| Â | RMSE | \(23.904\pm 1.126\) | \(26.300\pm 0.328\) | \(24.819\pm 0.632\) | \(\mathbf {21.401\pm 1.147}\) |
| Â | R\(^2\) | \(0.350\pm 0.060\) | \(0.215\pm 0.020\) | \(0.301\pm 0.035\) | \(\mathbf {0.479\pm 0.056}\) |
Nonpyridyl | MAE | \(\mathbf {14.143\pm 0.684}\) | \(15.043\pm 0.351\) | \(18.802\pm 1.216\) | \(17.259\pm 1.320\) |
| Â | RMSE | \(19.075\pm 0.751\) | \(\mathbf {18.580\pm 0.244}\) | \(23.610\pm 2.059\) | \(21.171\pm 1.141\) |
| Â | R\(^2\) | \(0.308\pm 0.055\) | \(\mathbf {0.344\pm 0.017}\) | \(-0.067\pm 0.188\) | \(0.146\pm 0.090\) |