Bibliographic record
Abstract
This thesis explores the challenges presented by label noise in machine learning (ML) for health. Label noise is unavoidable in real-world data; it can arise from variability in measurement, human annotation errors, and aleatoric uncertainty in the ground truth. Ignoring label noise when training classifiers with standard risk-minimization algorithms can result in faulty models that predict noise instead of signal; this is particularly problematic in healthcare where predictions can representmission-critical decisions. This thesis addresses the problem of label noise from several perspectives: exploring how and why label noise arises in health data, proposing robust algorithms to handle label noise, and studying the impact of label noise on individualized predictions while proposing methods for safer inference. Through a real-world study of stress detection in sensor data, the thesis first explores the significant inter- and intra-individual variability in label representations, undermining the reliability of classifiers trained on these labels. Motivated by this, the thesis introduces the concept of Temporal Label Noise, where noise levels vary over time, and proposes robust algorithms to handle this type of noise. It then examines the consequences of label noise on individual-level predictions, a gap in existing label noise research, while providing tools to mitigate the potential harm in high-stakes healthcare predictions. Finally, the work delves into a clinical case study in the Critical Care Unit, demonstrating how noisy sensor data can be leveraged to generate higher-quality annotations and improve patient safety initiatives. By identifying clinical challenges and proposing novel algorithmic solutions, this thesis intends to help advance the development of safe, robust, and trustworthy machine learning for healthcare.
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 | 0.023 | 0.092 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.002 | 0.006 |
| Scholarly communication | 0.007 | 0.008 |
| Open science | 0.003 | 0.007 |
| Research integrity | 0.004 | 0.007 |
| 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".