EEG 2025 Foundation Challenge#
End-to-end pipelines for the two tracks of the EEG 2025 Foundation Challenge: cross-task transfer learning, and subject-invariant representations for clinical-factor prediction. Difficulty 3; assumes the transfer and foundation-model track.
These tutorials are the runnable companions to the Foundation Challenge
manuscript. Each script ships with pre-trained weights so the build
reproduces the headline number without retraining from scratch, and
each frames the problem the way the Challenge protocol requires: the
training set, the held-out evaluation, and the metric. Pairs with
docs/tutorial_restructure_plan.md Category H (lines 458-470) and
follows the transfer principles in Schirrmeister et al. (2017).
What you will learn:
Challenge 1 (cross-task transfer): how to pretrain on a passive EEG task and transfer the learned representation to an active task, evaluated across held-out subjects.
Challenge 2 (subject-invariant representation): how to predict clinical factors (here: p-factor) with representations that generalise across subjects, evaluated against a per-subject baseline.
How to load the pre-trained weights that ship with this gallery and reproduce the published numbers without retraining.
How to extend either pipeline to a third task or a different clinical factor.
Run the tutorials:
tutorial_challenge_1.py– cross-task transfer learning.tutorial_challenge_2.py– subject-invariant representation for clinical-factor prediction.
How do I get my first baseline running for EEG2025 Challenge 1 (CCD)?
How do I submit a baseline to EEG2025 Challenge 2 (predict the p-factor)?