pbarbarant/fmri_alignment_benchmark

BenchOpt benchmark for fMRI alignment

fMRI Alignment Benchmark

Benchopt is a package to simplify and make more transparent and reproducible the comparisons of optimization algorithms. This benchmark is dedicated to the resolution of the functional brain alignment problem. Specifically, we evaluate the decoding accuracy of a classifier on a left-out subject after training it on data from other subjects, following functional alignment using the methods detailed below. We benchmark the following methods, all implemented in the [fmralign](https://github.com/Parietal-INRIA/fmralign/) package:

• Piecewise Procrustes
• Ridge Regression
• FastSRM
• Optimal Transport
• FUGW

on five different datasets:

• IBC RSVP
• IBC Sounds
• Courtois Neuromod
• BOLD5000

Install

This benchmark can be run using the following commands:

$ pip install -U benchopt
$ git clone https://github.com/pbarbarant/fmri_alignment_benchmark
$ benchopt run fmri_alignment_benchmark

Apart from the problem, options can be passed to benchopt run, to restrict the benchmarks to some solvers or datasets, e.g.:

$ benchopt run fmri_alignment_benchmark -s solver1 -d dataset2 --max-runs 10 --n-repetitions 10

Use benchopt run -h for more details about these options, or visit https://benchopt.github.io/api.html.

fmri_alignment_benchmark's Issues

Deprecated nil earn/fugw calls + dysfunctional Ridge/Ot solvers

Multiple problems in the solver:

• There seems to have outdated / deprecated calls to fugw methods/attribute (as well as nilearn) in the fugw solver.
• Ridge/OT are not "installed" according to benchopt

Add Feeling's Hyperalignment to the benchmark