Semester-Long Programs

Current Programs (2019-2020)
1) Deep Learning

  • Deep Learning will be presented in the fall semester of 2019. The program will focus on statistical strategies for improving machine learning. There is vast interest in automated methods for complex data analysis. However, there is a lack of consideration of: (1) interpretability; (2) uncertainty quantification; (3) applications with limited training data; and (4) selection bias. Statistical methods can achieve (1)-(4) through a change in focus.

2) Causal Inference

  • Causal Inference will be presented in the spring semester of 2020. Medical and health applications will be a significant theme, but other applications will be considered. Much of the new work in causal inference entails modern machine learning tools, and this perspective will be important to the program.

Upcoming Programs (2020-2021)
1) Program on Numerical Analysis in Data Science

  • Novel and efficient numerical techniques are undeniably needed to process and interpret massive data sets generated by modern technological and scientific developments; e.g., surveillance, space observation, medical data. Three overlapping themes in emerging numerical methods for this program are: (i) analysis of deep learning (DL) techniques; (ii) finding underlying dynamics of time dependent data sets; (iii) Randomized Numerical Linear Algebra (RandNLA) algorithms.

2) Program on Quantum Computing and Algorithms

  • Quantum computing is on the cusp of changing our computational paradigm, with progress exponentially expanding our computational reach. These challenges and opportunities require close collaborations between computer scientists, mathematicians, statisticians, physicists, and engineers, with impacts in scientific research, business, entertainment, and transportation.

3) Program on Combinatorial Probability

  • Placing a probability distribution over rankings and decomposition of rankings is also a probability model with combinatorial parameters, and can be used to extend classically deterministic optimization-based methods to stochastic models. Partition parameters arise in modeling gerrymandering in voting districts, where one is interested in distributions of demographic, political affiliation, and social affiliation variables conditional on the partition. Topological data analysis will be part of this program.

4) Program on Data Science in the Social and Behavioral Sciences

  • This program will address topics in computational social science, including social networks, machine learning, simulation methods, and other innovative data analysis procedures suitable for the complexity of such data.