v1.6.0

MOABB 1.6.0 — 68 commits and 81 changelog entries since v1.5, from 21 credited contributors.

📖 Full changelog — this is a summary.

Highlights

Datasets. A large expansion of the catalogue: imagined speech arrives as its own category of imagery datasets, with a dedicated documentation section, summary table and API grouping — including AguileraRodriguez2025, the four Nguyen2017 variants, Nieto2022 and Pressel2016. New adapters also cover Wang2026 (sensory-guided joint learning), Lenaig2026 (SSAEP), Schrag2026Pediatric (pediatric SSVEP, 47 children), and BCI Competition 2020 Tracks 4 and 5. BCIComp2020IS was removed: its released files publish trials in a scrambled order, so it cannot be evaluated reliably.

NEMAR / downloads. Dataset downloads can now be served from NEMAR's sourcedata/ store via moabb.set_download_provider, with nemar_id declared for 15 datasets whose deposits are public. The download-flag contract is honoured across every dataset — path, force_update, update_path and verbose are forwarded, CacheConfig.overwrite_* works with use=False, and a regression test enforces it for future datasets.

Evaluations. Cross-subject transfer learning lands in CrossSubjectEvaluation through an optional target-calibration slice and the new CrossSubjectMode presets. Folds can be driven by any stock scikit-learn cross-validator via cv_class plus a groups argument. New analysis helpers: compute_pvals_corrected_ttest (Nadeau & Bengio) and compute_lowest_subject_scores.

Requirements. Minimum pyriemann is now 0.12; the pyxdf dependency is retired in favour of a built-in XDF reader.

Housekeeping in this release

CITATION.cff declared cff-version: 1.4, which is not a real CFF schema version — the file could not be parsed by cffconvert or by GitHub's citation widget. It now declares 1.2.0. All five citation surfaces (CITATION.cff, the README APA and BibTeX blocks, and the same two in cite.rst) were normalised to the same 43 authors, adding the contributors credited in this cycle.

Install

pip install moabb==1.6.0