Announcing the Winners of Bloomberg's First Scientific Research Program

Originally posted on Bloomberg Now.

Bloomberg, we are committed to funding high-quality computer science research in areas that are critically shaping the computer science world and impacting our own research and business including machine learning, natural language processing, machine translation, statistics, and theory.

It is why we are happy to announce the winners of the first-ever Bloomberg Scientific Research Award program. A committee of Bloomberg researchers – from our R&D and CTO Office -selected proposals by Viswanath Nagarajan from the University of Michigan, Jinwoo Shin from KAIST, Shay Cohen from the University of Edinburgh, and Giorgio Satta from the University of Padua. Each award fully supports a graduate student for one year, including conference travel and hardware expenses.

About the Winners:

Scalable Probabilistic Deep Learning, Professor Jinwoo Shin, KAIST, South Korea

Deep learning is a popular approach to solve speech, vision, and text analytics problems. However, training models for deep learning is an expensive process. Prof. Shin’s proposal seeks to significantly speedup training of deep learning models by addressing fundamental algorithmic challenges inherent to deep learning.

Algorithms for Offline, Online and Stochastic Clustering, Professor Viswanath Nagarajan, University of Michigan

Clustering is of central importance in machine learning. However, many clustering problems are NP-complete; therefore, approximations are in common use. Prof. Nagarajan’s proposal attempts to improve on the state-of-the-art approximate algorithms for clustering in offline, online, and stochastic settings.

Latent-Variable Learning for Transition-Based Parsing, Professor Shay Cohen, University of Edinburgh and Giorgio Satta, University of Padua

Parsing is a critical step in natural language analysis and understanding. Prof. Cohen and Prof. Satta will investigate ways to accelerate parsing for streams of text from different sources. To do so, they propose to improve classifiers that resolve ambiguity where it arises.

We are currently accepting applications for our second round of grants. Scientific faculty at universities around the world can apply for funding from Bloomberg for consideration of an unrestricted grant to support their work. The next deadline for proposals is May 20th 2015.