External validation of a transformer triage model across four health systems
Iyer S, Novotny J, Haddad A and 2 others
Why it’s here In Machine Learning, which you follow · Field fit: 88 of 100
Today’s digest
Friday 2 October · 8 papers, 6 in the digest and 2 also worth a look
Showing 3 papers of 8
In today’s digest 1
Also worth a look 2
Self-supervised pretraining on unlabelled echocardiograms improves view classification
Tanaka R, Silva T, Achebe K
Why it’s here In Machine Learning, which you follow · Field fit: 79 of 100
From the abstract Labelled echocardiography data are scarce. Pretraining on 1.2 million unlabelled clips before fine-tuning on 5,000 labelled ones improved view classification over training from scratch.
Federated learning across hospitals without sharing records: a benchmark on readmission prediction
Fischer L, Osei B, Kowalski J and 2 others
Why it’s here In Machine Learning, which you follow · Field fit: 76 of 100
From the abstract We benchmark six federated learning methods for 30-day readmission prediction across eleven hospitals and find that simple federated averaging performs within two points of centralised training.