Skip to content

Experiments for Learning Features of Music From Scratch, ICLR 2017

Notifications You must be signed in to change notification settings

jthickstun/thickstun2017learning

Repository files navigation

thickstun2017learning

Experiments for Learning Features of Music From Scratch, ICLR 2017

Included are 6 notebooks: one for each of the 5 non-convex models discussed in the paper, and a notebook outlining the experimental setup for the linear models. For each non-convex model we include a set of weights and optimization statistics that gave us our results. Each notebook is set up to optionally load this set of weights. This allows you to continue optimization from this initialization, or to re-run the analysis for the numbers reported in the paper.

To run the notebooks, you will need to set an environment variable MUSICNET to the location where you have stored the MusicNet dataset. You can download MusicNet at our website:

http://homes.cs.washington.edu/~thickstn/start.html

About

Experiments for Learning Features of Music From Scratch, ICLR 2017

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published