Frozen AutoSOTA MoleculeACE ensemble. One unified training pipeline produces a target-specific model using Tanimoto-kernel ECFP SVR, stereochemical and pharmacophore Morgan experts, physicochemical features, random forest, and train-only OOF fusion.
No links found.
| Test set | Target label | Metric | Score |
|---|---|---|---|
| test | y | mean_squared_error | 0.38743026880651993 |
These are custom, user-defined attributes that are not required by the Polaris data model.
| Attribute | Value |
|---|---|
| Random seed | 42 |
| Frozen commit | 9aae09ae54aab3e60d857f7cd05fd5bf4f2c8c0f |
| Original-split primary metric | macro_cliff_rmse=0.745310113401 |