Estimating categorical counterfactuals via deep twin networks
Athanasios Vlontzos, Bernhard Kainz, Ciarán M. Gilligan-Lee
Machine learning touches every aspect of Spotify’s business. It is used to help listeners discover content via recommendations and search, generate playlists, extract audio content-rich signals for cataloging and other content-based applications, understanding voice commands, serve ads, develop business metrics and optimization algorithms, create music with AI-assisted tools, and more. Central to these endeavors is a commitment to cultivate expertise in the latest approaches as we advance the state of the art in machine learning methodology and applications. Of particular interest are approaches in reinforcement learning, approximate inference, graphical models, causal inference, deep learning, time series modeling, and meta-model learning.
Athanasios Vlontzos, Bernhard Kainz, Ciarán M. Gilligan-Lee
Himan Abdollahpouri, Zahra Nazari, Alex Gain,Clay Giibson, Maria Dimakopoulou, Jesse Anderton, Benjamin Carterette, Mounia Lalmas, Tony Jebara
Yan Chen, Perry Dong, Qinxun Bai, Maria Dimakopoulou, Wei Xu, Zhengyuan Zhou