Backend selection in the AMICA and AMICAICA wrappers

Open
#313 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
35/100
Issue type
Feature
Clarity
Needs clarification
Activity status
Active
Tech stack
python
Domain
backend, machine-learning

Research direction

Read pamica/amica.py and pamica/mne_compat/core.py:348-354 to trace how AMICA and AMICAICA select and expose their backend. Then review docs/api/mne-compat.md, backends.md, and the differences page. Done means the backend-selection decision is resolved and the resulting wrapper behavior and torch-only reality are documented, including the MNE export implications.

Written by the indexing model from the issue text.

Description

From the epic #278 integration-parity audit. Both convenience wrappers construct AMICATorchNG unconditionally: AMICA (pamica/amica.py) and, through it, AMICAICA (pamica/mne_compat/core.py:348-354). With the MLX backend now at full fitting + non-fitting parity (epic #278), every MLX user is by construction a raw-AMICAMLXNG user, losing the wrappers' conveniences (from_params_file, the #50 degenerate-fit usability contract, is_fitted_/converged_ surface), and — the sharpest case — an Apple-Silicon MNE user cannot get MLX-backed fits through AMICAICA at all without hand-reimplementing the to_mne_ica() export (mean/sphere/unmixing -> pca_mean_/pca_components_/unmixing_matrix_, including the rank-reduced SVD branch).

The torch-only nature of both wrappers is currently stated nowhere (checked docs/api/mne-compat.md, backends.md, changelog). Two separable decisions:

  • Whether AMICA (and thereby AMICAICA) grows a backend/engine selection parameter, or the raw-backend route stays the documented MLX path.
  • Either way, document the current torch-only wrapper reality explicitly (docs/api/mne-compat.md and the differences page), since the audit found it easy to over-read "no remaining MLX gaps" as covering the wrappers.

Relevant to the MLX audience specifically: individual experimenters on Apple hardware whose downstream is EEGLAB/MNE.

Dominant language
Python
Stars
8
Forks
3
Avg merge
5h 37m
Merged PRs (30d)
36

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from sccn/pAMICA

All issues in sccn/pAMICA

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.