Encoding the Underlying Dynamics of Complex Time Series With a Focus on Healthcare Applications
Bibliographic record
Abstract
Time series data captures temporal patterns and provides invaluable insights into various dynamic phenomena. In healthcare, time series plays a crucial role in understanding the temporal dynamics of individuals' health, enabling early detection of medical conditions, personalized treatment plans, and proactive healthcare management. This thesis addresses some of the challenges associated with modelling real-world time series data and provides novel solutions for encoding generalizable knowledge from the data that is applicable to multitude of tasks. This study presents a self-supervised framework for learning representations tailored to non-stationary time series. This framework captures transferable insights from the data that can be effectively applied across a range of tasks. The results demonstrate the potential of this approach for downstream prediction, clustering as well as in tracking health state trajectories for early detection. To improve the interpretability of the underlying representations, a solution is presented that decouples different factors of variability into independent encodings with generative models and through counterfactual regularization. The proposed approach enables multitude of tasks including counterfactual sample generation with control over different underlying factors which is a powerful tool for applications such as modelling drug responses in individuals. This thesis also explores explainability in time series models and its impact on adoption of these models in clinical practice. It introduces an information-theoretic approach for explaining the output of time series models by taking into account the temporal correlations in the data. The overarching objective of this thesis is to design solutions that bring machine learning models closer to adoption in clinical practice. In the concluding chapter, this is demonstrated through a case study that illustrates the pathway from design to deployment of a risk estimator model, highlighting the challenges in this journey and introducing protocols for comprehensive evaluation of its performance, usability, and safety in practice. In conclusion, this thesis contributes novel machine learning solutions to tackle the challenges associated with modelling real-world time series and deployment. The proposed solutions have the potential to unlock the rich information embedded in time series, enabling their adoption in various applications, particularly in 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 | 0.002 | 0.010 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| 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".