Tree smoothing: Post-hoc regularization of tree ensembles for interpretable machine learning
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Random Forests (RFs) are powerful ensemble learning algorithms that are widely used in various machine learning tasks. However, they tend to overfit noisy or irrelevant features, which can result in decreased generalization performance. Post-hoc regularization techniques aim to solve this problem by modifying the structure of the learned ensemble after training. We propose a novel post-hoc regularization via tree smoothing for classification tasks to leverage the reliable class distributions closer to the root node whilst reducing the impact of more specific and potentially noisy splits deeper in the tree. Our novel approach allows for a form of pruning that does not alter the general structure of the trees, adjusting the influence of nodes based on their proximity to the root node. We evaluated the performance of our method on various machine learning benchmark data sets and on cancer data from The Cancer Genome Atlas (TCGA). Our approach demonstrates competitive performance compared to the state-of-the-art and, in the majority of cases, and outperforms it in most cases in terms of prediction accuracy, generalization, and interpretability. • A novel post-regulation technique for Tree Ensembles called BBTS is introduced. • Interpretability is improved through posterior distributions in the leaf nodes. • This method allows the incorporation of domain knowledge through prior beliefs. • It was tested using ML benchmarks and real-world cancer data from the TCGA database. • BBTS has proven itself with the state-of-the-art and exceeds them on real-world cancer data.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.008 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it