Transfer, Foundation Models, and EEG2025#

Estimated reading time:3 minutes

Four advanced lessons on transfer learning and foundation-model fine-tuning, framed around the EEG2025 Foundation Challenge. Difficulty 3; assumes the core workflow, features, and evaluation tracks.

Transfer is where the EEG decoding field is moving fastest, and it is also where most of the unprincipled choices accumulate: tasks selected to make the transfer score look good, evaluation that does not respect subject boundaries, fine-tuning learning rates pulled from thin air. These lessons follow Schirrmeister et al. (2017) for the architecture and training principles, and use the EEG2025 Challenge tasks as the concrete, reproducible benchmark.

What you will learn:

How EEGChallengeDataset differs from EEGDashDataset and when to reach for which.
How to set up a cross-task transfer experiment (Challenge 1): resting-state pretraining transferred to contrast-change detection.
How to run subject-invariant regression for clinical-factor prediction (Challenge 2): predict p-factor across held-out subjects.
How to fine-tune a Braindecode pretrained model on a downstream task with sane hyperparameter choices.
How to read a transfer result critically: what scores actually mean when the source and target tasks share subjects.

Run the lessons in order:

plot_70_challenge_dataset_basics.py – EEGChallengeDataset basics.
plot_71_cross_task_transfer.py – EEG2025 Challenge 1.
plot_72_subject_invariant_regression.py – EEG2025 Challenge 2.
plot_73_finetune_pretrained_model.py – fine-tune a Braindecode pretrained model.

How do I get started with the EEG2025 Foundation Challenge dataset?

Pretrain on resting-state, fine-tune on contrast-change detection (Simulated Data)

Subject-invariant p-factor regression (EEG2025 Challenge 2)

How do I adapt a pretrained EEG model to a new task?

How do I plug EEGDash into the Meta NeuroAI ecosystem?