Heuristics-free self-supervised representation learning for time series with SIGReg (LeJEPA). Disentangles time-axis collapse, positional structure, and representation richness across PatchTST, TCN, and bag-of-patches encoders. Reproducible, seeded, significance-tested.
machine-learning research deep-learning time-series pytorch transformer hydra representation-learning anomaly-detection tcn time-series-forecasting self-supervised-learning financial-time-series jepa dimensional-collapse patchtst representation-collapse lejepa sigreg
-
Updated
Jun 18, 2026 - Python