arxivcs.LGq-bio.NC2026-07-10
CoCoT-EEG: Contrastive-Pretrained Multiscale Convolutional Transformer for EEG Decoding
Gabriel Mahuas, Victoria Shevchenko, Ugo Tanielian, Yassir Bendou, Richard Gao
Self-supervised pretrained foundation models (FM) have shown early promise for non-invasive electroencephalogram (EEG) decoding applications. Many recent large-scale models converged on the approach of tokenizing raw EEG followed by masked reconstruction pretraining. However, thi…