Drausin Wulsin

wulsin[at]seas.upenn.edu
[About]
As of September 2013, I will be a working as an engineer at the data analysis compay Palantir. I received my Ph.D. i in 2013 from the Department of Bioengineering at the University of Pennsylvania, working in the Translational Neuroengineering Lab with my advisor, Brian Litt. I graduated from Yale in 2007 with a B.S. in Biomedical Engineering.
[Interests]
My research focused on intelligently processing large amounts of clinical data using machine learning and statistical modeling techniques. I used nonparametric Bayesian models like hierarchical Dirichlet processes (HDP), HDP-hidden markov models, and beta processes to build large-scale models of seizures across many patients. This work aims to intelligently compare and analyse seizures and other epileptic events within a patient and between patients. In previous work, I have used models like Restricted Boltzmann Machines and Deep Belief Nets to learn clinical EEG waveforms and then do pattern recognition and anomaly detection on them.
[Publications]
D. F. Wulsin, Bayesian Nonparametric Modeling of Epileptic Events. Ph.D thesis. University of Pennsylvania, May 2013. pdf
D. F. Wulsin, E. B. Fox, B. Litt. Parsing Epileptic Events Using a Markov Switching Process for Correlated Time Series. International Conference on Machine Learning (ICML) 2013 [pdf, supp pdf]
D. F. Wulsin, S. T. Jensen, B. Litt. A Hierarchical Dirichlet Process Model with Multiple Levels of Clustering for Human EEG Seizure Modeling. International Conference on Machine Learning (ICML) 2012 [pdf, supp pdf]
D. F. Wulsin, J. R. Gupta, R. Mani, J. A. Blanco, B. Litt. Modeling electroencephalography waveforms with semi-supervised deep belief nets: faster classification and anomaly measurement. Journal of Neural Engineering 8 (3), 2011 [pdf]
J. Viventi, ... D. Wulsin, et al. Flexible, Foldable, Actively Multiplexed, High-Density Surface Electrode Arrays for Mapping Brain Activity in vivo with Single Trial Resolution. Nature Neuroscience, 14 (12): 1599-1605, 2011
D. Wulsin and B. Litt. An Unsupervised Method for Identifying Regions that Initiate Seizures on Intracranial EEG. Proceedings of the IEEE Engineering in Medicine and Biology Conference (EMBC), 2011 [pdf]
A. Pearce, D. Wulsin, B. Litt, J. Blanco. Does the Morphology of High-Frequency(100-500 Hz) Brain Oscillations Change During Epileptic Seizures? Proceedings of the IEEE Asilomar Conference on Signals, Systems, and Computers, 2011 (in press)
D. Wulsin, J. A. Blanco, R. Mani, B. Litt. Semi-Supervised Anomaly Detection of EEG Waveforms Using Deep Belief Nets. Proceedings of the International Conference on Machine Learning and Applications (ICMLA) 2010 [pdf]
T. Kyriakides, D. Wulsin, E. A. Skokos, P. Fleckman, A. Pirron, J. M. Shipley, R. M. Senior, P. Borstein. Mice that lack matrix metelloproteinase-9 display delayed wound healing associated with delayed reepitheliazation and disordered collagen fibrillogenesis. Matrix Biology 28 (2):65-73, 2009 [pdf]
[Code]
For simplicity, I have posted below code I think others might find most useful, but if you would like to see code from any of my publications, let me know. Of course, if you find bugs/errors or have general suggestions, please let me know.
BP-HMM Library (Matlab) BP-HMM_v1.0.zip
- OOP implementation of (BP-)HMM time series models, including the event-based BP-HMM model in Wulsin et al., ICML 2013
MLC-HDP Library (Matlab) MLC-HDP_v1.0.zip
- OOP implementation of the multilevel clustering hierarchical Dirichlet process (MLC-HDP)
Deep Learning Object Oriented Library (Matlab) DBNToolbox_v1.0.zip
- OOP implementatations of Restricted Boltzmann Machines (RBMs) and Deep Belief Nets
[Data]
Most of my datasets are too large to post here, but if you're interested in them, send me an email.