Reproduce “STST-JEPA: Shallow-Target Spatio-Temporal Joint Embedding Prediction Architecture For EEG Self-Supervised Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Roy Segal, Yoni Svechinsky, Tomer Fekete
STST-JEPA learns reusable EEG representations by predicting masked structure in a learned space while lightly reconstructing waveforms. Pretrained on 47,703 sessions, it supports an age readout. The paper-reported validation result is 3.06 years mean absolute error across 3,367 sessions, but the paper gives conflicting accounts of the validation cohort. In NeuralBench, age ranks fourth with standard 2-second inputs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.