Doerschuk’s research group develops, validates, and uses mathematical models of primarily biological and medical systems, mostly with the goal of understanding experimental data. The range of problems has been broad, e.g., ethanol pharmacokinetics in human subjects based on breath-analyzer data to understanding resting-state brain behavior via functional MRI to the 4-D spatial-temporal dynamics of virus particles based on single-particle cryo electron microscopy images. Most of the work is in collaboration with a domain expert whose research group performs the experiments. Most of the work is of a statistical character. The group is trying to integrate these statistical ideas, which can incorporate large amounts of prior knowledge such as physics, with the “learning from examples” approaches of artificial intelligence, especially “deep learning”. The most successful work often involves systems where something is understood about the physical – chemical basis of the system and/or the measurement process. Some of the problems are “big data” problems, e.g., 100,000 images in single-particle cryo electron microscopy. The group’s two goals are to develop new interesting ideas in modeling and inference and to solve data understanding problems of interest to our collaborators!
Much of the group’s work involves custom solutions to challenges posed by collaborators. Such solutions are often a mix of symbolic components and computational components. Applying the ideas at relevant scales often involves substantial computations and the group has extensive experience with practical computing up to the level of writing software systems for distributed-memory cluster computing with C/C++/OpenMP/CUDA/MPI and Python. But the group also writes a lot of Matlab for multi-core execution on a desktop computer!
In summary, the group’s interests include:
time series and networks
If you are interested in this group’s research, please contact me at pd83 AT cornell.edu!