​
Bio: Oren Somekh received his BSc, MSc, and PhD in Electrical Engineering from the Technion in 1989, 1991, and 2005, respectively. During 1991-1996 he served in the IDF Signals Corps. During 1998-2002 he was the VP R&D and later CTO of Surf Communication Solutions Ltd. During 2005-2008 he has been a visiting research fellow at the EE Departments of NJIT and Princeton University. Dr. Somekh Joined Yahoo Labs in 2009 and serves there as a senior Research Scientis​t since. He was a recipient of the European Community Marie-Curie Outgoing International Fellowship, a co-recipient of the first IEEE Information Theory Society ISIT best student paper award, a co-recipient of Yahoo 2015 Master Inventor award, and holds patents in the fields of Communications and Internet technologies. His current research interests are scientific aspects of Internet technologies, such as Recommendation and Computational Advertising. 

Item Cold-Start Handling in Collaborative Filtering Recommenders

​Abstract​:​ The item cold-start problem is inherent in collaborative filtering (CF) recommenders where items and users are represented by vectors in a latent space. It emerges since CF recommenders rely solely on historical user interactions to characterize their item inv​entory. As a result, an effective serving of new and trendy items to users may be delayed reducing both users' and content suppliers' satisfaction. Much work has been devoted to mitigate this problem mostly by employing hybrid approaches that combine context- and content-based recommendation techniques or by devoting a portion of the users' traffic for exploration to gather interactions from random users.

In this talk we focus on pure CF recommender systems (i.e., operating without content or context information) and consider several approaches for mitigating the item cold-start problem in both online and offline settings. We start with an offline setting where all users are available for exploration and propose greedy approximation algorithms to select a set of users for rating the new item in order to minimize the prediction error of our model. Next, we consider an online setting where users arrive randomly one after the other and propose algorithms for the system to immediately decide whether the arriving user will participate in the exploration of the new items. Finally, we present a hybrid approach for learning a mapping between the item attribute space and the CF latent feature space. The mapping is then used to characterize the new items providing initial estimates for their latent vectors. 

​​​