The use of deep neural networks has exploded in popularity recently. Thinking that music information retrieval should not be left out of this trend in machine learning, we explore two different applications of this technology in the field.

The first we looked at was genre identificaton, using the initial categories of 'popular music,' 'art music,' and 'traditional music.' This was found to be a difficult problem - classifying music into these categories can be challenging even for experts, and assembling a large dataset for use in training represents a significant problem.

The second approach we took to using these techniques was looking at instrument identification, specifically for the purpose of identifying the time and category (from "guitar", "vocal", or "drum") of solos in popular music.