34th Deep Learning Meetup in Vienna (virtual edition)

Dear Deep Learners,

Welcome back to the Vienna Deep Learning Meetup talk series after a long break! We will start with a stellar contribution by Sander Dieleman from Google DeepMind London on music generation. Please join us on Thursday, August 20 at 19:00 via Zoom! (Link will be provided to registered participants.)

Talk:
Generating music in the waveform domain
by Sander Dieleman, Google DeepMind

Almost all research in computational music generation so far has focused on symbolic representations: scores, MIDI sequences and other representations that make abstraction of certain aspects of the music, such as the idiosyncrasies of a particular performance. However, this removes a lot of nuance from the generated content that can be quite important musically, and more importantly, can greatly impact our enjoyment of the music. Generating music directly in the waveform domain is a more challenging task, but it allows us to capture musicality at all scales, from musical form and phrasing all the way down to timbral variations of the instruments. Recent advances in deep learning research have made this approach tractable. In this talk, I will give an overview of the different approaches that have been proposed in the literature and compare them, with a focus on data efficiency, scalability and malleability of the generated audio signals.

The talk will be followed by a Q&A with Sander Dieleman.

As usual, we will also include time for networking - have your favorite drinks and snacks ready :-)

Looking forward to seeing you virtually at the meetup,
Tom, Jan, Alex, René