Back to home
Planned
When does an EEG decoder generalize?
A planned reproducible comparison of simple motor imagery decoding baselines, with evaluation on held out runs and people.
Why
The intention behind it.
Motor imagery decoding results are easy to report optimistically. I want to build a small, honest comparison where the evaluation split is the point, not an afterthought.
Deliverables
What would exist if this is actually completed.
- 01
A public repository with the analysis code
- 02
A reproducible environment specification
- 03
A written protocol fixed before running the comparison
- 04
A results table across held out runs and held out people
- 05
A subject level figure rather than a single averaged number
- 06
A written limitations section
Open questions
What I do not know yet.
- How much does accuracy drop moving from held out runs to held out people?
- Which baseline is the fairest reference point for a beginner to start from?
- How should uncertainty be reported when subject counts are small?
Results
Results.
No results yet
This project is planned. No data has been analysed, so there are no figures, metrics, code links, or participant counts to show. This page will only change when real outputs exist.