31st Deep Learning Meetup in Vienna

Dear Deep Learners,

Our last meetup of the year is approaching: On 2nd of December we will have two exciting topics again: Singing Voice Detection in audio and a report about a Google AI residency and AutoML.

Talk 1:
Taming Horses in Singing Voice Detection
by Jan Schlüter, Researcher at Austrian Research Institute for Artificial Intelligence (OFAI) and contextflow GmbH

When listening to music, many humans can easily tell when somebody is singing, but nobody can give detailed instructions to implement this ability on a computer. I will show how to train a deep neural network to accurately detect and even extract vocals, only given music pieces tagged as vocal/non-vocal without any timing information. I will then demonstrate two surprising ways the network can be fooled, and effective measures to avoid this. Taking a broader perspective, these experiments teach a lesson on machine learning "horses" and the evaluation of automated decisions.

Jan Schlüter has been pursuing research on deep learning for audio processing since 2010. For his Ph.D. degree at the Johannes Kepler University Linz, he focused on music audio; currently he is working on bioacoustics at the Austrian Research Institute for Artificial Intelligence (OFAI) and medical imaging at contextflow GmbH.

Talk 2:
How to do ML if you have lots of Google’s GPUs
by Vladimir Macko, Freelancer - Machine Learning Consulting, ex Google AI resident

Journey of a student from UK Bratislava into doing research at Google New York as an AI resident. What is AI residency, what is AutoML, what is architecture search, how can one make use of tons of GPUs, why does AdaNet have such an amazing GIF, what does it mean to play at state-of-the-art levels of accuracy in image classification, and what is so far the best indicator for getting into the AI residency?

Conference Review:
Sebastian Böck and Katharina Prinz, researchers at OFAI, will give us a report from the 20th conference of the International Society for Music Information Retrieval (ISMIR 2019 in Delft, NL) and the importance of Deep Learning in this domain.

As usual, the meetup will be complemented by hot topics & latest news about Deep Learning, as well as networking.

Looking forward to seeing you at this last meetup of the year 2019,
Jan, Alex, René, Tom