On Collaborative Content FIltering and Federated Learning
Details
For our final talk of a very productive 2016, we'll be meeting at Zillow to hear two talks from brilliant scientists from Zillow and Google, respectively. Please RSVP via Meetup here and we'll see you there.
Talk 1: Content based modeling and Collaborative Filtering for Home Recommendations
Recommendations products exist throughout Zillow Group. Homes for sale recommendations in email, search results personalized per user, similar homes for sale models on the property pages, and personalized collection of homes are some of the use cases. In this talk, Jasjeet Thind, Nicholas Stevens & Shruti Kamath will go in-depth into the core recommendation algorithms and technology.
About the Speakers
http://photos3.meetupstatic.com/photos/event/2/d/d/a/600_455471738.jpeg
Jasjeet Thind is the Senior Director of Data Science and Engineering at Zillow. His group focuses on machine-learned prediction models and big data systems that power use cases such as Zestimates, personalization, housing indices, search, content recommendations, and user segmentation. Prior to Zillow, Jasjeet served as director of engineering at Yahoo, where he architected a machine-learned real-time big data platform leveraging social signals for user interest signals and content prediction. The system powers personalized content on Yahoo (http://www.yahoo.com/), Yahoo Sports (http://www.sports.yahoo.com/), and Yahoo News (http://www.news.yahoo.com/). Jasjeet holds a BS and master’s degree in computer science from Cornell University.
Shruti Kamath is a Software Development Engineer, Machine Learning in the Data Science and Engineering team at Zillow. She is working on the recommendations platform for personalizing home recommendations at Zillow. Previously, she was a member of Amazon's Personalization team, primarily on building backend platform components for recommendations through Deep learning. She received her Master’s degree in Computer science from Columbia University, New York.
Nicholas Stevens is the Senior Program Manager for personalization and deep learning projects at Zillow. Before that he was the first Product Manager at Prismatic (content recommendations) and a Mobile Product Manager at Twitter. He has a Masters in Computer Science from Stanford and undergraduate degrees in Economics from Wharton and Systems Engineering from the University of Pennsylvania.
Talk 2: Federated Learning on Modern Mobile Devices
Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data center and training there using conventional approaches. We present an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. This talk will present the advantages and challenges of this approach, as well as a practical algorithm for the federated learning of deep networks.
About the Speaker
Brendan McMahan is a research scientist at Google Seattle. His interests include decentralized and distributed machine learning, online learning, differential privacy, and large-scale optimization.
On Collaborative Content FIltering and Federated Learning