D-semble: Efficient Diversity-Guided Search for Resilient ML Ensembles
Bibliographic record
Abstract
Supervised Machine Learning (ML) is used in many safety-critical applications, such as self-driving cars and medical imaging. Unfortunately, many training datasets have been discovered to contain faults. The accuracy of individual models when trained with faulty datasets can significantly degrade. In comparison, ensembles, consisting of multiple models combined through simple majority voting, are able to retain accuracy despite training data faults, due to their classification diversity, and are thus more resilient. However, there are many different ways to generate ML ensembles, and their accuracy can significantly differ. This creates a large search space for ensembles, making it challenging to find ensembles that maximize accuracy despite training data faults. We identify three different ways to generate diverse ML models, and present D-semble, a technique that uses Genetic Algorithms and diversity to efficiently search for resilient ensembles. We evaluate D-semble by measuring the balanced accuracies and F1-scores of ensembles it finds. Compared with bagging, greedy search, random selection, and the best individual model, ensembles found by D-semble are on average 9%, 16%, 28%, 32% more resilient respectively.
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.
How this classification was reachedexpand
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 | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.001 |
| 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".