Wednesday, August 21
10:00 am - 10:40 am
Machines as Artists: Making Music with Neural Networks
About the event
For a long time, we’ve been thinking about computer science as a tool for automation and computation. With the recent rise in popularity of artificial intelligence, researchers have started experimenting with using it for more artistic and creative applications.
In the first part of the talk, you’ll hear about an overview of the different tools that were developed to use machine learning to generate music, such as Google Magenta, MuseGAN, and MidiNet. With architectures from Variational Autoencoders (VAE) to Generative Adversarial Networks (GAN) and Long Short-Term Memory (LSTM); different tools are applied on different music data types, from raw audio files (e.g. mp3 files) to midi sequences.
The second part of the talk will walk you through how you can get started to work on this problem, create your own LSTM in a few steps, and understand many of the consideration of how music generation differers from other time-series-based applications of machine learning.
Husni is a Ph.D. candidate at Carnegie Mellon University. He spends a lot of his time writing Python programs to unravel the mysteries of the universe around us in general, and the mysteries of dark energy in particular.
He is passionate about artificial intelligence and generally enjoys solving challenging interdisciplinary problems -- from finance to music -- in novel computational ways.