Here is the source code developed during an internship "Decoding auditory cortex using high-dimensional regression" at PARIETAL (Inria lab).

It is released under BSD 3-clause license.

First of all, you need stimuli: they were communicated privately by Michael Hanke. So they are absent in this repo. Essentially it is audio files and text annotations (in JSON).

Then you should download original dataset. I would recommend to rsync from here. Put your paths in fg_constants.py and preprocess data using preprocess.py. Then look at cross_validation.py to choose what exactly you compute. If audio, run convert_audio.py, if word2vec, download a wikidump (German) and run train_word2vec.py...

After that, run cross_validation.py and view_batch_3d.py to look at results.

Plenty of them. NumPy, SciPy, scikit-learn, gensim, nilearn, nsgt. Optionally, pycortex or FSL to visualization, glove, FFTW, features.