Section: New Results

Highlight 1 High-dimensional Adaptive Ranking with PAC-Bayesian Bounds

Participant : Benjamin Guedj.

The quasi-Bayesian perspective has been extended to the popular setting high-dimensional ranking. This is a pivotal problem in machine learning and is at the core of several applications in industry (recommender systems, active learning, ...). An original estimator of the scoring function is proposed, and we have shown its minimax optimal properties. Our procedure is adaptive to the unknown sparsity level of the data. This work is published in Journal of Statistical Planning and Inference.

It a joint work with Sylvain Robbiano from University College London.