Sketching for Large-Scale Learning of Mixture Models

Abstract: Learning parameters from voluminous data can be prohibitive in terms of memory and computational requirements. Given a data collection, it is however a classical assumption that it was drawn from an underlying probability distribution which typically has low intrinsic complexity. This assumption makes it conceivable to apply Compressive Sensing paradigms, in which low-dimensional objects are recovered from fewer measurements than the ambient dimension, directly to probability distributions. A data collection can then be compressed in a few measurements of the underlying probability distribution, referred to as “sketch”, which theoretically encodes all information about this probability distribution.
Using this basis principle, is it possible to derive practical methods to perform learning tasks on large databases ? Can theoretical guarantees from Compressive Sensing be adapted to this framework, and is it possible to link these guarantees to more classical learning theory ?

Biography: He graduated from Ecole polytechnique (Palaiseau, France), and obtained the « Mathématiques, Vision, Apprentissage » (MVA) Master’s degree from Ecole Normale Supérieure de Cachan in 2014. He is currently a PhD student at IRISA, Rennes, France, under the supervision of Rémi Gribonval.

Research Interests

• Compressive sensing
• Dimensionality reduction
• Learning, big data

His PhD thesis consists in applying compressive sensing paradigms to generalized objects (such as probability distributions) for deriving efficient learning methods on big databases.