Combinatorial Scientific Computing
We focus on discrete optimization problems on large-scale graphs which are often used to accelerate the performance of scientific computing algorithms. Typical combinatorial scientific computing tasks include such problems as (hyper)graph partitioning, reordering, and coloring to improve load-balancing, task mapping, and data locality on modern high-performance computing systems.
We are interested in fundamental modeling and computational principles underlying various multiscale methods. A broad range of scientific problems involve multiple scales. Traditional monoscale approaches have proven to be inadequate, even with the largest supercomputers, because of the prohibitively large number of variables involved. Thus, there is a growing need to develop multiscale approaches in which a hierarchy of coarse scale approximations is used to solve large-scale problems efficiently.
Machine Learning and Data Mining
Many standard machine learning and data mining algorithms are prohibitive for large-scale number of variables. For example, this can happen because of the slow convergence or NP-hardness of underlying optimization problems (such as in support vector machines and cut-based clustering). We are interested in algorithms that cope with such problems.
Hypothesis Generation and Text Mining
Hypothesis generation is becoming a crucial time-saving family of techniques which allow researchers to quickly discover implicit connections between important concepts. We are interested in such techniques and complex text mining problems, in general.
We are interested in computational, modeling, theory and data problems related to complex networks in social/natural/information sciences, and engineering. The types of analysis often include frequent pattern discovery, outliers detection, quantitative methods for importance ranking of network elements, time-dependent data analysis, network evolution modeling, visualization, and community detection.
We are always looking for talented graduate and undergraduate students. Please feel free to apply and suggest exciting projects.