recent updates
- N Sismanis, DL Sussman, JT Vogelstien, et al. Extracting Proximity for Brain Graph Voxel Classification. 5th Panhellenic Conference of Biomedical Technology, Athens, Greece.
- CE Priebe, DL Sussman, M Tang, JT Vogelstein. Statistical inference on errorfully observed graphs. under review at JASA. arxiv.
- Q Wu, D Carlson, W Lian, M Zhou, CR Stoetzner, D Kipke, D Weber, JT Vogelstein, D Dunson, L Carin. Multichannel Electrophysiological Spike Sorting via Joint Dictionary Learning & Mixture Modeling. under review at IEEE TBME. arxiv.
- JT Vogelstein, et al. Large (Brain) Graph Matching via Fast Approximate Quadratic Programming. under review at PAMI, arxiv, code, repo.
- JT Vogelstein, et al. Shuffled Graph Classification: Theory and Applications in Statistical Connectomics. round 2 @ J of Classification, arxiv, repo.
philosophy
This world is already an incredibly beautiful and lovely place to live (for many of us). Our thesis is that by improving our models of how we think and act, we will be able to make this world even better for more of us. More specifically, we aim to contribute to explaining mental properties (e.g., thoughts, emotions, memories, psychiatric conditions) in terms of brain properties (e.g., connectivity structure, patterns of spike trains, BOLD signal) in ways that lead to qualitative insight and quantitative predictions. We contribute to the design of neuroscience, psychology and psychiatry experiments. The data objects under investigation consist of longitudinal multi-modal 3D brain images, genetics, behavioral assessments (spanning a wide variety of spatiotemporal scales, ranging from cubic nanometers and microseconds to cubic millimeters and years). Thus these data are ultrahigh-dimensional non-Euclidean objects. We believe our best hope in making progress is to fuse these multiple massive disparate and dynamic data sets to perform various statistical inference investigations. Applied investigations including obtaining clinically useful predictive analytics by finding low dimensional structure. Basic investigations include estimating the generative model of the data at various scales. All investigations operate via the development and extension of tools from statistical machine learning, applied mathematics, and computer science, incorporating domain knowledge as much as possible in every step along the way. A primary motivating factor is that all humans have brains and therefore such ideas could directly benefit all of humankind. In alignment with our intentions, all of our research products are freely available to all, including: publications; conferences; invited talks (ie, not conference talks); my code and the open connectome project's github page; data.