Term Project


Due on 18 January 2011
You need to prepare a poster presentation, which will be scheduled for 2011/1/18, 11am-2pm, at EECS B1.

You also need to upload a report of your term project to the ftp server.



Option 1:
Apply one of the following methods
  1. Support Vector Machines
  2. Boosting algorithms
  3. Gaussian processes, or 
  4. EM algorithms
  5. Deep learning (http://deeplearning.net/)
to your current research topics.

Option 2:
Join one of the following competitions:

  1. RTA Freeway Travel Time Predictio
    http://kaggle.com/RTA

  2. Predict Grant Applications
    http://www.kaggle.com/unimelb

Option 3:
Implement one of the following papers:

1. Generative Local Metric Learning for Nearest Neighbor Classification
Y. Noh, B. Zhang, D. Lee

2. Simultaneous Object Detection and Ranking with Weak Supervision
M. Blaschko, A. Vedaldi, A. Zisserman

3. Worst-Case Linear Discriminant Analysis
Y. ZhangD. Yeung

4. Random Projections for $k$-means Clustering
C. Boutsidis, A. Zouzias, P. Drineas

5. Deep Coding Network
Y. Lin, T. Zhang, S. Zhu, K. Yu

6. Deterministic Single-Pass Algorithm for LDA
I. Sato, K. Kurihara, H. Nakagawa

7. Feature Set Embedding for Incomplete Data
D. Grangier, I. Melvin

8. Generating more realistic images using gated MRF's
M. Ranzato, V. Mnih, G. Hinton

9. Kernel Descriptors for Visual Recognition
L. Bo, X. Ren, D. Fox

10. Large Margin Multi-Task Metric Learning
S. Parameswaran, K. Weinberger

11. New Adaptive Algorithms for Online Classification
F. Orabona, K. Crammer

12. Unsupervised Kernel Dimension Reduction
M. Wang, F. Sha, M. Jordan

13. Learning Convolutional Feature Hierarchies for Visual Recognition
k. kavukcuoglu, P. Sermanet, Y. Boureau, K. Gregor, M. Mathieu, Y. LeCun

14. Learning via Gaussian Herding
K. Crammer, D. Lee

15. Universal Kernels on Non-Standard Input Spaces
A. Christmann, I. Steinwart

16. Exploiting weakly-labeled Web images to improve object classification: a domain adaptation approach
A. Bergamo, L. Torresani

17. Pose-Sensitive Embedding by Nonlinear NCA Regression
G. Taylor, R. Fergus, G. Williams, I. Spiro, C. Bregler

18. Label Embedding Trees for Large Multi-Class Tasks
S. Bengio, J. Weston, D. Grangier



A Primal-Dual Algorithm for Group Sparse Regularization with Overlapping Groups
S. Mosci, S. Villa, A. Verri, L. Rosasco

A Theory of Multiclass Boosting
I. Mukherjee, R. Schapire

Beyond Actions: Discriminative Models for Contextual Group Activities
T. Lan, Y. Wang, W. Yang, G. Mori

Boosting Classifier Cascades
M. Saberian, N. Vasconcelos

A Reduction from Apprenticeship Learning to Classification
U. Syed, R. Schapire

Convex Multiple-Instance Learning by Estimating Likelihood Ratio
F. Li, C. Sminchisescu

Discriminative Clustering by Regularized Information Maximization
R. Gomes, A. Krause, P. Perona

Distributed Dual Averaging In Networks
J. Duchi, A. Agarwal, M. Wainwright

Efficient Optimization for Discriminative Latent Class Models
A. Joulin, F. Bach, J. Ponce

Exact learning curves for Gaussian process regression on large random graphs
M. Urry, P. Sollich

Factorized Latent Spaces with Structured Sparsity
Y. Jia, M. Salzmann, T. Darrell

Gated Softmax Classification
R. Memisevic, C. Zach, G. Hinton, M. Pollefeys

Gaussian Process Preference Elicitation
E. Bonilla, S. Guo, S. Sanner

Gaussian sampling by local perturbations
G. Papandreou, A. Yuille

Getting lost in space: Large sample analysis of the resistance distance
U. von Luxburg, A. Radl, M. Hein

Graph-Valued Regression
H. Liu, X. Chen, J. Lafferty, L. Wasserman

Group Sparse Coding with a Laplacian Scale Mixture Prior
P. Garrigues, B. Olshausen

Guaranteed Rank Minimization via Singular Value Projection
P. Jain, R. Meka, I. Dhillon

Hashing Hyperplane Queries to Near Points with Applications to Large-Scale Active Learning
P. Jain, S. Vijayanarasimhan, K. Grauman

Implicitly Constrained Gaussian Process Regression for Monocular Non-Rigid Pose Estimation
M. Salzmann, R. Urtasun

Joint Cascade Optimization Using A Product Of Boosted Classifiers
L. Lefakis, F. Fleuret

Large-Scale Matrix Factorization with Missing Data under Additional Constraints
K. Mitra, S. Sheorey, R. Chellappa

Layered image motion with explicit occlusions, temporal consistency, and depth ordering
D. Sun, E. Sudderth, M. Black

Learning To Count Objects in Images
V. Lempitsky, A. Zisserman

Object Bank: A High-Level Image Representation for Scene Classification Semantic Feature Sparsification
L. Li, H. Su, E. Xing, L. Fei-Fei

Occlusion Detection and Motion Estimation with Convex Optimization
A. Ayvaci, M. Raptis, S. Soatto

On the Theory of Learnining with Privileged Information
D. Pechyony, V. Vapnik

Penalized Principal Component Regression on Graphs for Analysis of Subnetworks
A. Shojaie, G. Michailidis

Random Projection Trees Revisited
A. Dhesi, P. Kar

Size Matters: Metric Visual Search Constraints from Monocular Metadata
M. Fritz, K. Saenko, T. Darrell

Structural epitome: a way to summarize one’s visual experience
N. Jojic, A. Perina, V. Murino

Universal Consistency of Multi-Class Support Vector Classification
T. Glasmachers






Comments