Skip to content

Code for NeurIPS 2019 Neuro-AI workshop, paper : "Estimating encoding models of cortical auditory processing using naturalistic stimuli and transfer learning"

Notifications You must be signed in to change notification settings

iqdecay/neurips19_neuroai_encoding

Repository files navigation

Companion code for the accepted paper at NeurIPS'19 Workshop "Real Neurons & Hidden Units: Future directions at the intersection of neuroscience and artificial intelligence".

Link to the paper on OpenReview : "Estimating encoding models of cortical auditory processing using naturalistic stimuli and transfer learning"

Written by :

Our encoding models are trained on SoundNet features. We provided the SoundNet features, and also explain the method to regenerate them in step 1.

Code :

Requirements

  • Python 3.7
  • nilearn
  • sklearn
  • pandas
  • matplotlib
  • numpy
  • scipy
  • tqdm
  • pytorch
  • soundfile
  • librosa

To extract features from SoundNet :

To estimate encoding models :

Run the notebook "2_encoding_with_parcellation.ipynb"

Please read the instructions inside carefully, and configure the various path accordingly.

Data :

We include R2 maps for encoding models estimated on conv7 layer, using 1000 neurons in the hidden layers. ALl other maps can be regenerated using the provided code.

About

Code for NeurIPS 2019 Neuro-AI workshop, paper : "Estimating encoding models of cortical auditory processing using naturalistic stimuli and transfer learning"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published