NEWS

Machine Learning

RIT Americas issues poster at DCC 2020

2020.04.29

Dr. Yang Shi of RIT Americas, in collaboration with co-author Animashree Anandkumar of the California Institute of Technology, had her poster accepted at this year’s Data Compression Conference (DCC). The DCC is an internationally-renowned forum for theoretical and experimental projects on data compression and related applications. It brings together members of academic institutions and industry scientists.

Given the explosive growth of data, it has become critical to develop efficient techniques to store and transmit data, to minimize the size of data being stored or communicated as well as relax computational costs. “Higher-order Count Sketch: Dimension Reduction that Retains Efficient Tensor Operations” addresses employing a data compression technique known as sketching, a randomized dimensionality reduction method that aims to preserve relevant information in large-scale datasets. The poster, a reference to a previously submitted work under the same name, proposes a novel extension known as Higher-order Count Sketch (HCS) and derives efficient approximation of various tensor operations such as tensor products and contractions. HCS is the first sketch to fully exploit the multi-dimensional nature of higher-order tensors.

Due to the COVID-19 pandemic, the 2020 conference convened the broad research community and industry experts in a virtual format. Dr. Shi was still able to share her contribution and ideas to diverse program committee members and network with attendees working at other data-driven companies in the bay area and globally, including Ebay, Google, Alibaba, and Bytedance.

RIT would like to thank the DCC, members, and sponsors for transitioning smoothly and creating an impactful virtual event.

Copied! instagram
Research Areas : #Machine Learning