Data perturbation for escaping local maxima in learning
Bibliographic record
Abstract
Almost all machine learning algorithms—be they for regres-sion, classification or density estimation—seek hypotheses that optimize a score on training data. In most interesting cases, however, full global optimization is not feasible and local search techniques are used to discover reasonable solu-tions. Unfortunately, the quality of the local maxima reached depends on initialization and is often weaker than the global maximum. In this paper, we present a simple approach for combining global search with local optimization to discover improved hypotheses in general machine learning problems. The main idea is to escape local maxima by perturbing the training data to create plausible new ascent directions, rather than perturbing hypotheses directly. Specifically, we consider example-reweighting strategies that are reminiscent of boost-ing and other ensemble learning methods, but applied in a different way with a different goal: to produce a single hy-pothesis that achieves a good score on training and test data. To evaluate the performance of our algorithms we consider a number of problems in learning Bayesian networks from data, including discrete training problems (structure search), con-tinuous training problems (parametric EM, non-linear logistic regression), and mixed training problems (Structural EM)— on both synthetic and real-world data. In each case, we obtain state of the art performance on both training and test 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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.010 | 0.027 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.003 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.002 | 0.004 |
| Research integrity | 0.003 | 0.003 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
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 source (direct Gemma or distilled Codex), 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".