Exposes

Articles

• Simple Local Models for Complex Dynamical Systems, Erik Talvitie, Satinder Singh, NIPS 2008
  • Rémi Durand
• Curriculum Learning Yoshua Bengio, Jerome Louradour, Ronan Collobert and Jason Weston, ICML 2009
  • Humans and animals learn much better when the examples are not randomly presented but organized in a meaningful order...
  • Wang Zhe
• Supervised Learning from Multiple Experts: Whom to trust when everyone lies a bit, Vikas Raykar, Shipeng Yu, Linda Zhao, Anna Jerebko, Charles Florin, Gerardo Valadez, Luca Bogoni and Linda Moy, paper ID: 96
  • What to do when we have multiple experts/annotators providing (possibly noisy) labels but no absolute gold standard..
  • ZHANG Fan
• Learning the Difference between Partially Observable Dynamical Systems Laviolette et al. ECML 2009
  • We propose a new approach for estimating the difference between two partially observable dynamical systems. We assume that one can interact with the systems
• Neighbourhood Components Analysis, J Goldberger, S Roweis, G Hinton, Salakhutdinov, NIPS
  • The point is to learn a distance by optimizing the number of misclassified examples through a k-nearest neighbors...
  • Asma MEHIAOUI
• Multiagent Reinforcement Learning for Urban Traffic Control Using Coordination Graphs, 2008
  • Amina Kacem (proposé).
• Random Classification Noise Defeats All Convex Potential Boosters, Philip M. Long and Rocco A. Servedio, ICML 2008
  • We show that for some loss functions there exists a training set which can be efficiently learned by a boosting algorithm iff there is no training noise...
• Boosting the area under the ROC curve P. M. Long and R. A. Servedio, NIPS 2007.
  • Same principles as for boosting can be applied when considering another learning criterion, the Area under the ROC curve...
  • Djalel Benbouzid
• The Tradeoffs of Large Scale LearningLéon Bottou and Olivier Bousquet, NIPS 07:
  • There is the bias and the variance; what about the error of optimization ?...
• Distance Metric Learning for Large Margin Nearest Neighbor, K. Weinberger, NIPS 2005
  • Learning a distance by optimizing the classification accuracy of a k-nearest neighbor classifier...
  • Yahia Dahbia
• Efficient MultiClass Maximum Margin Clustering Bin Zhao, Fei Wang, and Changshui Zhang, ICML 2008.
  • We are iteratively building clustering with increasingly large margin between clusters, using a cutting plane approach...
• A Support Vector Method for Multivariate Performance Measures, T. Joachims, ICML 2005
  • We show how to extend Support Vector Machines to other loss functions, e.g. the Area Under the ROC Curve...
  • Pierre Allegraud
• Adaptive Cluster on Ensemble Selection, Javad Azimi and Xiaoli Fern, IJCAI 2009.
  • In ensemble learning, the diversity of the classifiers is useful. We show that ensemble clustering should use moderately diverse partitions...
  • Wuhan Soy
• Learning to learn implicit queries from gaze patterns Kai Puolamakki, ICML 2008
  • The goal is to learn user's interests from implicit signals, obtained by eye tracking...
  • Mouna SELMI
• Bayes Optimal Classification for Decision Trees, Siegfried Nijssen, ICML 2008
  • This paper bridges the gap between building decision trees and frequent itemset mining...
  • Dhouha BOUAMOR
• The Many Faces of Optimism: a Unifying Approach, Istvan Szita and Andras Lorincz, ICML 2008
  • Reinforcement learning aims at learning an optimal decision policy, for which it must explore the search space. This paper examines the exploration task...
• Multi-Task Learning for HIV Therapy Screening, Steffen Bickel, Jasmina Bogojeska, Thomas Lengauer, and Tobias Scheffer, ICML 2008
  • When the dataset is extracted from different distributions (e.g. people in different countries have been infected by slightly different viruses, have received different antibiotics), how to exploit these data...
• Random projection in dimensionality reduction Applications to image and text data, E Bingham, H Mannila
  • Random projection is compared to standard dimensionality reduction on image and text-related applications. We show that sparse random projections have particular properties...
  • Ghazal JABER
• EM algorithms for PCA and SPCA, Sam Roweiss
  • Standard Principal Component Analysis is computationally very expensive. This paper shows how Expectation Maximization can be used for the fast extraction of the highest eigenvalues/vectors...
  • MBENGUE Cheikh
• Model selection via the Area Under the Roc Curve Saharon Rosset, ICML 2004.
  • ''A learning criterion is the Area Under the Roc Curve. This criterion is compared to the rate of misclassified examples, examining its bias, variance,..."
  • Rouabhia Salih
• Structured Sparse Principal Component Analysis, Jenatton et al.
  • ''The goal is to learn a sparse dictionary, e.g. for face recognition or identification of proteins schemes, when the dictionary elements should satisfy some constraints..."

Si quelqu'un veut trouver un autre article / in case someone would prefer finding a paper:

• List of papers ICML 2009
• List of papers ICML 2008

Tutoriels

You also have the possibility to hear, summarize and comment a tutorial:

• Massive Security, High dimensional, Leon Bottou, Video Lectures MLSS 2007
• Energy-based learning, Yann Le Cun, NIPS 2006

Examen 2 decembre: Questions de cours