English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Machine learning strategies like multi-task learning, meta-learning, andtransfer learning enable efficient adaptation of machine learning models tospecific applications in healthcare, such as prediction of various diseases, byleveraging generalizable knowledge across large datasets and multiple domains.In particular, Bayesian meta-learning methods pool data across relatedprediction tasks to learn prior distributions for model parameters, which arethen used to derive models for specific tasks. However, inter- and intra-taskvariability due to disease heterogeneity and other patient-level differencespose challenges of negative transfer during shared learning and poorgeneralizability to new patients. We introduce a novel Bayesian meta-learningapproach that aims to address this in two key settings: (1) predictions for newpatients (same population as the training set) and (2) adapting to new patientpopulations. Our main contribution is in modeling similarity between causalmechanisms of the tasks, for (1) mitigating negative transfer during trainingand (2) fine-tuning that pools information from tasks that are expected to aidgeneralizability. We propose an algorithm for implementing this approach forBayesian deep learning, and apply it to a case study for stroke predictiontasks using electronic health record data. Experiments for the UK Biobankdataset as the training population demonstrated significant generalizabilityimprovements compared to standard meta-learning, non-causal task similaritymeasures, and local baselines (separate models for each task). This wasassessed for a variety of tasks that considered both new patients from thetraining population (UK Biobank) and a new population (FinnGen).

Bayesian Meta-Learning for Improving Generalizability of Health Prediction Models With Similar Causal Mechanisms