NIPS 2013 Workshop: Perturbations, Optimization, and Statistics

Schedule

7:30-7:50 Danny Tarlow
Microsoft Research
Introduction
7:50-8:20 Andrea Montanari
Stanford (invited)
High-Dimensional Robust Estimation
[abstract]
8:20-8:50 Stefano Ermon
Cornell (invited)
Probabilistic Inference by Universal Hashing and Optimization
[abstract]
8:50-9:05 Discussion Time (to be used as appropriate in above window)
9:05-9:30 Coffee Break
9:30-10:00 Moshe Ben-Akiva
MIT (invited)
Discrete Choice Models with Large Choice Sets
10:00-10:30 Karthik Sridharan
UPenn (invited)
Power of Randomization and Perturbations in Online Learning
[abstract]
10:30-10:45 Gergely Neu
INRIA (contributed)
An Efficient Algorithm for Learning with Semi-Bandit Feedback
10:45-11:05 Spotlight Session Training Restricted Boltzmann Machine by Perturbation
Efficient Feature Learning Using Perturb-and-MAP
Does Better Inference mean Better Learning?
On Robustness and Regularization of Structural Support Vector Machines
11:05-break Poster Session (all contributed papers)
15:30-16:00 Percy Liang
Stanford (invited)
Dropout Regularization
[abstract]
16:00-16:30 David McAllester
TTI-C (invited)
PAC-Bayes and Model Perturbations
[abstract]
16:30-16:45 Ben London
UMD (contributed)
PAC-Bayes Generalization Bounds for Randomized Structured Prediction
16:45-17:05 Discussion Time (to be used as appropriate in above window)
17:05-17:30 Coffee Break
17:30-18:00 Subhransu Maji
TTI-C (invited)
Efficient Multi-scale Boundary Annotation using Random Maximum A-Posteriori Perturbations
[abstract]
18:00-18:30 Panel discussion + Wrapup

Accepted Contributed Papers

The following contributed papers will be presented at the poster session and also as spotlights or short talks.

  1. Andrew E. Gelfand, Rina Dechter, and Alexander Ihler
    Does Better Inference mean Better Learning? (spotlight)
  2. Gergely Neu and Gabor Bartok
    An Efficient Algorithm for Learning with Semi-Bandit Feedback (short talk)
  3. Ben London, Bert Huang, Ben Taskar, and Lise Getoor
    PAC-Bayes Generalization Bounds for Randomized Structured Prediction (short talk)
  4. Siamak Ravanbakhsh, Russell Greiner, and Brendan J. Frey
    Training Restricted Boltzmann Machine by Perturbation (spotlight)
  5. Ke Li, Kevin Swersky, and Richard Zemel
    Efficient Feature Learning Using Perturb-and-MAP (spotlight)
  6. MohamadAli Torkamani and Daniel Lowd
    On Robustness and Regularization of Structural Support Vector Machines (spotlight)