about me
Update 2/10/2014: I have now started at Google as a Research Scientist!
Previously, I was a graduate student in the Computer and Information Science (CIS) Department at the University of Pennsylvania. I was advised by Ben Taskar.
I do research in machine learning and specialize in computer vision and natural language processing. My research focuses on enabling more complex models via efficient, practical learning and inference procedures. I co-organized the NIPS 2010 workshop on Coarse-to-Fine Learning and Inference. I still enjoy neuroscience (as is my background); ultimately, I hope to promote the integration of machine learning and neuroscience, and to change the way we analyze and understand our minds.
I am currently a member of the GRASP Lab and the Penn Research in Machine Learning (PRIML) research groups. Previously, I was a research specialist at the Princeton Computational Memory Lab. I graduated from Princeton University with a degree in Computer Science and a certificate in Neuroscience in 2007, advised by Dave Blei (CS) and Ken Norman (Psychology).
In my spare time, I enjoy partaking in machine learning competitions. Most recently, I won 2nd place in the CHALEARN Gesture Challenge as part of Team Pennect. Before that, I was a member of Team Dinosaur Planet, Grand Prize Team, and The Ensemble, the greatest Netflix Prize wrecking crew ever created. Our team was written up in the New York Times.
Finally, I'm also a founding software developer of MedForward Inc. and maintain several software packages of my own. I also contributed significantly to the MVPA Toolbox for Matlab.
highlighted projects
- Structured Prediction Cascades. A general
method for speeding up inference and enabling more complex
models. This has led to state-of-the-art results
in OCR,
human pose estimation and
tracking,
and natural
language parsing (extension by others). Code for linear-chain
models is
available here,
human pose estimation in still
images here,
and in
video here. Read
about it in
detail here.
- Object segmentation with SCALPEL. A state-of-the-art image object segmentation algorithm that generates hundreds of object segmentation proposals per image with high recall. It uses refinement of localized shape priors with a cascade of greedy superpixel selectors. Read about it here and download code here.
- One-shot gesture recognition. We used an HMM-based approach to win second place in the CHALEARN One-shot Learning Challenge. Learn more from my slides (with videos) in keynote or quicktime format.
- Teaching Machine Learning and AI. I have worked with Professor Taskar for the past four years to advance the Penn Machine Learning cirriculum. This includes building an automated Matlab based homework submission and grading system, a live-updated Matlab based ML competition leaderboard, and even online educational videos, like the one below: Additionally, I designed and co-taught a novel undergraduate course using the 2011 Google AI Challenge as a motivating basis. As part of this I developed a user-friendly Python framework for students to use to complete assignments.
- Enhanced Matlab toolboxes. In my own work and collaborations with Ben Sapp I've maintained two useful Matlab toolboxes: one for managing distributed experiments and another filled with generally useful functions. See my code page for more information.
publications
-
PDF |
Poster
Code TBA Learning adaptive value of information for structured prediction.
David Weiss and Ben Taskar.
Neural Information Processing Systems (NIPS), December 2013. -
PDF |
Poster
Code TBA Dynamic structured model selection.
David Weiss, Benjamin Sapp, and Ben Taskar.
International Conference on Computer Vision (ICCV), December 2013. -
PDF
Code SCALPEL: Segmentation CAscasdes with Localized Priors and Efficient Learning.
David Weiss and Ben Taskar.
Computer Vision and Pattern Recognition (CVPR), June 2013. -
PDF
Structured Prediction Cascades.
David Weiss, Benjamin Sapp, and Ben Taskar.
Pre-print; under review at JMLR. -
PDF
Slides
Code Parsing Human Motion with Stretchable Models.
Benjamin Sapp, David Weiss, and Ben Taskar.
Computer Vision and Pattern Recognition (CVPR), June 2011.
(Oral presentation, 3.5% acceptance) -
PDF |
Supp. Info
Poster
Code Sidestepping Intractable Inference with Structured Ensemble Cascades.
David Weiss, Benjamin Sapp, and Ben Taskar.
Neural Information Processing Systems (NIPS), December 2010. -
PDF |
Supp. Info
Code Mixed Membership Matrix Factorization.
Lester Mackey, David Weiss, and Michael I. Jordan.
International Conference on Machine Learning (ICML), June 2010. -
PDF
Poster
Code Structured Prediction Cascades.
David Weiss and Ben Taskar.
International Conference on Artificial Intelligence and Statistics (AISTATS), May 2010. -
URL
Listening for recollection: A multi-voxel pattern analysis of recognition memory retrieval strategies.
Joel R. Quamme, David J. Weiss, and Kenneth A. Norman.
Frontiers in Human Neuroscience, 4(0), 2010. - PDF
Probabilistic Additive Component Analysis: A Latent
Variable Model for Dimensionality Reduction of Human fMRI
Datasets.
David Weiss. Senior Thesis, Princeton University, May 2007. -
PDF
Haptic Rendering of Tissue Cutting with Scissors in a Virtual
Environment.
David Weiss and Alison Okamura.
Medicine Meets Virtual Reality 12, J.D. Westwood, et al. (Eds.), IOS Press, 2004, pp. 407-409.