Sparse Linear Methods for Top-N Recommender Systems

GRAND Seminar Nov. 15, 3pm, Thursday, 2012, ENGR 4801

Xia Ning
Research staff
Autonomic Management Department
NEC Labs America

Host: Huzefa Rangwala

Abstract:

Recommender systems represent a set of computational methods that produce recommendations of interesting entities (e.g., products, friends, etc) from a large collection of such entities by retrieving/filtering/learning information from their own properties (e.g., product attributes, personal profiles, etc) and/or the interactions between these entities and other parties (e.g., user-product ratings, friend-friend trust relations, etc). Recommender systems are particularly important for E-commerce applications, where the overwhelming amount of items makes it extremely difficult for users to manually identify those items that best fit their personal preferences. Recommender systems have attracted significant research interests from academia, and meanwhile, they have been functioning as a critical revenue enhancer for major E-commerce websites such as Amazon.com and eBay.com.

In this talk, we will address a core task for recommender systems, that is, top-N recommendation, in which a size-N list of items that most conform to the user's interests and preference is to be generated for recommendation. We have developed 1). a novel sparse linear method for top-N recommendation, which utilizes regularized linear regression with sparsity constraints to model user-item purchase patterns, and it significantly outperforms the current state-of-the-art methods; 2). a set of novel sparse linear methods with side information for top-N recommendation, which use side information to regularize sparse linear models or use side information directly to model user-item purchase behaviors, and they stand for the best performing methods with side information incorporated. These sparse linear methods are particularly suitable for Big-Data environment, where the huge amount of information leaves the conventional computation-intensive recommendation algorithms inapplicable, whereas the sparse linear methods are easy to be distributed on large-scale computing systems like Hadoop.

Bio:

Dr. Xia Ning is a research staff member at the Autonomic Management Department, NEC Labs America. Dr. Ning received her PhD in Computer Science from the University of Minnesota, Twin Cities, 2012. Her doctoral research focuses on recommender systems for large-scale e-commerce applications and data mining/machine learning methods for drug discovery and medical informatics. She has also been working on applying and developing statistical analytics/data mining algorithms for real “big-data” problems arising in text mining, social network mining and autonomic system management.