Reducing cross-sample prediction churn in scientific machine learning
About
Scientific machine learning reports predictive performance. It does not report whether the same prediction would survive a different draw of training data. Across $9$ chemistry benchmarks, two classifiers trained on independent bootstraps of the same training set agree on aggregate accuracy to within $1.3\text{--}4.2$ percentage points but disagree on the class label of $8.0\text{--}21.8\%$ of test molecules. We call this gap \emph{cross-sample prediction churn}. The standard parameter-side techniques (deep ensembles, MC dropout, stochastic weight averaging) do not reduce this gap; two data-side methods do. The first is $K$-bootstrap bagging, which cuts the rate $40\text{--}54\%$ on every dataset at no accuracy cost ($K{\times}$-ERM compute). The second is \emph{twin-bootstrap}, our proposal: two networks trained jointly on independent bootstraps with a sym-KL consistency loss between their predictions, which at matched $2{\times}$-ERM compute reduces churn a further median $45\%$ beyond bagging-$K{=}2$. Cross-sample prediction churn deserves a column alongside predictive performance in scientific-ML benchmark reports, because without it the parameter-side and data-side methods are indistinguishable on the metric they actually differ on.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Distributional disagreement (sym-KL) | CYP2D6-Sub (id-test) | Delta Sym-KL-0.63 | 5 | |
| Distributional disagreement (sym-KL) | Pgp (id-test) | Delta Symmetric KL Divergence (delta sym-KL)-0.5 | 5 | |
| Distributional disagreement (sym-KL) | BACE (dev) | Delta sym-KL-0.74 | 5 | |
| Distributional disagreement (sym-KL) | TADF (id-test) | Delta Symmetric KL Divergence (delta sym-KL)-0.39 | 5 | |
| Distributional disagreement (sym-KL) | MOF-thermal (id-test) | Delta sym-KL-0.38 | 5 | |
| Distributional disagreement (sym-KL) | BBBP (id-test) | Delta Sym-KL Divergence-0.46 | 5 | |
| Distributional disagreement (sym-KL) | AMES (id-test) | Delta Sym-KL-1.11 | 5 | |
| Molecular Property Classification | CYP2D6-Sub | ∆ ID-Churn-9.3 | 5 | |
| Molecular Property Classification | BBB-Martins | Delta ID Churn-7.4 | 5 | |
| Molecular Property Classification | AMES | Delta ID Churn-8.3 | 5 |