DANG Hong-Phuong
Assistant Professor of Machine Learning - Co-Head of the Data Engineering and Machine Learning program Research interests- Apprentissage automatique (supervisé et non supervisé)
- Apprentissage statistique
- Apprentissage de dictionnaire, Représentation parcimonieuse
- Statistique Bayésienne, Bayésien non paramétrique
- Traitement du signal et de l'image
Campus de Ker Lann
51 Rue Blaise Pascal
BP 37203
35172 BRUZ Cedex
I work with Bayesian Models, particularly non-parametric ones. These models are linked to Stochastic Processes such as the Dirichet Process, the Chinese Restaurant Process, or the Indian Buffet Process. The interest of such models is not fixing the number of degrees of freedom in advance. The result is models which enrich themselves are the number of observations increase. I use these models for problems of factorizing matrices, particularly dictionary learning for parsimonious representation which is well known for resolving inverse problems. This makes it possible to not limit the size of the dictionary in advance. Regarding applications, I'm interested in images (denoising, deconvolution, inpainting, or image segmentation) and compressed acquisition.
I am currently studying interactions between Bayesian Methods and Optimization to improve the numerical complexity of algorithms by proposing approximation methods of a posteriori law via variational approaches and asymptotic small variance.
Key Words:
automatic learning, dictionary learning, factorization of matrices, classification, Bayesian Non-Parametrics, Stochastic Processes, Monte Carlo, Markov, parcimonious representations, inverse problems, interaction between Bayesian Methods and Optimization.
My website:
http://www.hongphuong-dang.com/index.html