TITLE: Learning Preferences with Side-Information: Recovering 3-Dimensional Tensors

ABSTRACT:

A number of recent problems of great interest in e-commerce (such as context and location aware recommendations, personalized ’tag' learning, etc.) can be cast as large-scale problems of tensor recovery in three dimensions. Thus motivated, we consider the problem of recovering ‘simple’ tensors from their noisy observations in three dimensions. We provide an efficient algorithm to recover structurally ‘simple' tensors given noisy (or missing) observations of the tensor’s entries; our definition of simplicity subsumes low-rank tensors for a variety of definitions of tensor rank. Our algorithm is practical for large datasets and provides a significant performance improvement over state of the art incumbent approaches to Tensor completion. Further, we establish theoretical recovery guarantees that under certain assumptions are order optimal. Joint work with Andrew Li (MIT ORC).

Bio: 

Vivek is interested in the development of new methodologies for large scale dynamic optimization and applications in revenue management, finance, marketing and healthcare. He received his Ph.D. in Electrical Engineering from Stanford University in 2007 and has been at MIT since. Vivek is a recipient of an IEEE Region 6 Undergraduate Student Paper Prize (2002), an INFORMS MSOM Student Paper Prize (2006), an MIT Solomon Buchsbaum Award (2008), an INFORMS JFIG paper prize twice (2009, 2011), the NSF CAREER award (2011), a finalist for the Pierskalla award (2011), MIT Sloan’s Outstanding Teacher award (2013), the INFORMS Simulation Society Best Publication Award (2014), and the INFORMS Pricing and Revenue Management Best Publication Award (2015). Outside of academia, he has contributed to the design of the algorithmic trading strategies of GMO's (a USD 100B + money manager) first high frequency venture and is the co-founder and CTO of a venture-backed predictive analytics startup.