Please note in order to gain entry, you MUST sign up on both this site & Dataiku's MeetUp page
6:30pm: Pizza + Beer networking
7:00pm: TBD by Data Scientist at Dataiku
7:30pm: Nonlinear Latent Factorization for Personalized Recommender System by TaeYoung Choi, Data Scientist at Conde Nast
Nonlinear Latent Factorization for Personalized Recommender System by TaeYoung Choi, Data Scientist at Conde Nast The New Yorker publishes about 90 new articles every week & about half of the pageviews are made within the first 24hrs. As The New Yorker is renowned for its political & cultural commentary, its articles in general have shorter lifespans compared to other items such as IPhone cases on Amazon or Bad Guy by Billie Eillish on Spotify. In other words, we do not have time to gather enough data - we are constantly facing the cold-start problem. This talk will cover how I approached the problem by creating user embeddings that will allow us to make recommendations as soon as a new article gets published. The model burrows the famous word2vec architecture & contains additional layers that incorporate nonlinearity to capture multiple interests for our readers.
Speaker bio: TaeYoung a data scientist at Cond Nast, home to some of the world's most iconic brands, including Vogue, The New Yorker, GQ, Vanity Fair & Wired. At Cond Nast, he has led various data science projects such as traffic forecasting, subscription churn, user segmentation & classification, & content personalization. He's passionate about personalization & recommendation. He's previously participated in developing a data science pipeline at MyCelebs, (a Korean-based data science AI start-up), & worked as a graduate researcher at Columbia Business School, building matching algorithms from unstructured data for Executive Education Program.