v0.1.0 — Initial release

AutoFMRI v0.1.0

First public release as a pip-installable Python package.

Install

pip install autofmri

Highlights

• AutoEstimator: Two-stage automatic voxel selection (atlas-based or cube-based patchification + SHAP feature importance)
• Unified CLI: autofmri train --dataset haxby --method autofmri --atlas yeo
• 6 built-in data loaders: Haxby, HCP Emotion/Gambling/Language, VKMC, GTZAN
• Multiple models: Random Forest, SVM, XGBoost (optional), 1D CNN (optional)
• No GPU required: cuML dependency removed; XGBoost and PyTorch are optional extras

Links

• PyPI: https://pypi.org/project/autofmri/0.1.0/
• Paper: ICASSP 2023