AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

EpiTwin: Spatiotemporal Graph Transformers for Epileptic sEEG Signal Reconstruction

Jingbo Yang, Yunfeng Zhao, Chao Qiu, Yulin Sun, Xiuyun Liu, Xiaofei Wang

ICML 2026 · Graph learning › Graph self-supervised learning · approach: Brain and biological graph learning

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use EpiTwin instead of strongest baseline LaBraMsEEG reconstruction · sEEG reconstruction, single patientRMSE: 0.591 -> 0.492Improved
use EpiTwin instead of LaBraMsEEG reconstruction · sEEG reconstruction, single patientLSD / High-LSD: LSD 0.541 -> 0.463; High-LSD 0.525 -> 0.449Improved
combine geometric, topological and positional encodings (HSPE) rather than a single source (ablation)EpiTwin · sEEG reconstruction, single patientRMSE: about 0.51 -> 0.492Improved
use geometry-functional biased attention rather than vanilla attention (ablation)EpiTwin · sEEG reconstruction, single patientRMSE: 0.5220 -> 0.4920Improved
use geometry-functional biased attention rather than graph k-NN attention (ablation)EpiTwin · sEEG reconstruction, single patientLSD: 0.4877 -> 0.4630Improved
add adversarial refinement stage (ablation)EpiTwin · sEEG reconstruction, single patientLSD / High-LSDImproved
use EpiTwin instead of LaBraM with two patientssEEG reconstruction · sEEG reconstruction, two patientsLSD: 0.6984 -> 0.6986No clear effect
use EpiTwin instead of LaBraM with five patientssEEG reconstruction · sEEG reconstruction, five patientsRMSE: 0.6080 -> 0.5773Improved
use EpiTwin instead of LaBraM on unseen patientsEEG reconstruction · sEEG reconstruction, leave-one-subject-outRMSE: 1.253 -> 0.956Improved
evaluate with temporally contiguous split instead of random splitEpiTwin · sEEG reconstructionRMSE: 0.492 -> 0.504Hurt
add more subjects to the training dataall methods · sEEG reconstructionRMSE: EpiTwin 0.4920 (1 patient) -> 0.5773 (5 patients)Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy