Binary Rating Estimation With Graph Side Information

Authors:
Kwangjun Ahn Korean Augmentation To the United States Army (KATUSA)
Kangwook Lee KAIST
Hyunseung Cha Kakao Brain
Changho Suh KAIST

Abstract:

Rich experimental evidences show that one can better estimate users' unknown ratings with the aid of graph side information such as social graphs. However, the gain is not theoretically quantified. In this work, we study the binary rating estimation problem to understand the fundamental value of graph side information. Considering a simple correlation model between a rating matrix and a graph, we characterize the sharp threshold on the number of observed entries required to recover the rating matrix (called the optimal sample complexity) as a function of the quality of graph side information (to be detailed). To the best of our knowledge, we are the first to reveal how much the graph side information reduces sample complexity. Further, we propose a computationally efficient algorithm that achieves the limit. Our experimental results demonstrate that the algorithm performs well even with real-world graphs.

You may want to know: