Pavlo Apisov graduated with a degree computer engineering in 2013 and finished his master’s in computer systems and networks in 2014. Since then, he has made Android applications. He also enjoys traveling around the world. One of his major interests is music. He has been writing music for 10 years already, and, taking into account his passion for programming, the next possible move is to combine these things. Lately, he has been trying to make a transition from Android development to machine learning research, and attending the Wolfram Summer School is a first serious step to doing this.
Project: Generating Music with Expressive Timing and Dynamics
The idea behind this project is to generate a sequence of notes that represent a musical performance. We build a neural network that knows nothing about music and try to teach it with thousands of songs given in MIDI format.
Main Results in Detail
After the first 10 rounds of training, the model gives pretty musical performances. The model starts overfitting during the second session of training and tends to generate shorter performances.
Exploring possibilities of learning long-term structures in multi-instrumental music using the MusicVAE model made by Magenta team.