Maxime Vono

Welcome

I am a Ph.D. student since October, 2017 in the beautiful city of Toulouse, France. I work under the supervision of Pierre Chainais and Nicolas Dobigeon, within the SC group of the IRIT laboratory.

Prior to that, I graduated from Ecole Centrale de Lille majoring in data science. I also hold a M.Sc. degree in applied mathematics from University of Lille.

My research interests are in new methods and algorithms to solve challenging Bayesian inference problems encountered in machine learning & signal processing. In particular, I am interested in making more efficient Markov chain Monte Carlo (MCMC) algorithms that scale in high-dimension and/or big data settings by exploiting the connections between optimization and simulation-based methods.

To this purpose, I recently proposed a unifying statistical framework, namely asymptotically exact data augmentation (AXDA), which circumvents the art of finding an exact data augmentation scheme. AXDA is strongly related to approximate Bayesian computation (ABC) and encompasses well-established (e.g., mixture and robust modeling) but also recently proposed (e.g., variable splitting-based) augmented models. Algorithms derived from AXDA have important benefits such as simpler inference steps which can be distributed over multiple machines or kernels and a non-asymptotic control of the approximation under some mild assumptions.

News/Events:
March-April 2019 I will visit Arnaud Doucet's research group at University of Oxford.
February 2019 Submitted paper on asymptotically exact data augmentation (AXDA).
February 2019 2 papers accepted at ICASSP 2019.
January 2019 Our paper on split-and-augmented Gibbs sampler (aka ADMM-inspired sampling) has been accepted for publication in IEEE Transactions on Signal Processing.
(starting) Sept., 2018 Data science/analytics consultancy missions for a large retailer.
Sept. 17-20, 2018 I will present our paper in the sparse learning session of MLSP'18 conference.

Research

Working Papers

  1. M. Vono, N. Dobigeon, and P. Chainais, “Asymptotically exact data augmentation: models, properties and algorithms,” submitted, February 2019.

    Data augmentation, by the introduction of auxiliary variables, has become an ubiquitous technique to improve mixing convergence properties, simplify the implementation or reduce the computational time of inference methods such as Markov chain Monte Carlo. Nonetheless, introducing appropriate auxiliary variables while preserving the initial target probability distribution cannot be conducted in a systematic way but highly depends on the considered problem. To deal with such issues, this paper draws a unified framework, namely asymptotically exact data augmentation (AXDA), which encompasses several well-established but also more recent approximate augmented models. Benefiting from a much more general perspective, it delivers some additional qualitative and quantitative insights concerning these schemes. In particular, general properties of AXDA along with non-asymptotic theoretical results on the approximation that is made are stated. Close connections to existing Bayesian methods (e.g. mixture modeling, robust Bayesian models and approximate Bayesian computation) are also drawn. All the results are illustrated with examples and applied to standard statistical learning problems.

            @article{Vono2019_AXDA,
            author = {Vono, M. and Dobigeon, N., and Chainais, P.},
            year = {2019},
            title = {Asymptotically exact data augmentation: models, properties and algorithms},
            journal = {submitted},
            volume = {},
            number = {},
            pages = {}
            }
          

Journal Articles

  1. M. Vono, N. Dobigeon, and P. Chainais, “Split-and-augmented Gibbs sampler - Application to large-scale inference problems,” IEEE Transactions on Signal Processing, vol. 67, no. 6, pp. 1648-1661, March 2019.

    This paper derives two new optimization-driven Monte Carlo algorithms inspired from variable splitting and data augmentation. In particular, the formulation of one of the proposed approaches is closely related to the alternating direction method of multipliers (ADMM) main steps. The proposed framework enables to derive faster and more efficient sampling schemes than the current state-of-the-art methods and can embed the latter. By sampling efficiently the parameter to infer as well as the hyperparameters of the problem, the generated samples can be used to approximate Bayesian estimators of the parameters to infer. Additionally, the proposed approach brings confidence intervals at a low cost contrary to optimization methods. Simulations on two often-studied signal processing problems illustrate the performance of the two proposed samplers. All results are compared to those obtained by recent state-of-the-art optimization and MCMC algorithms used to solve these problems.

            @article{Vono2019,
            author = {Vono, M. and Dobigeon, N., and Chainais, P.},
            year = {2019},
            title = {Split-and-augmented {G}ibbs sampler - {A}pplication to large-scale inference problems},
            journal = {IEEE Transactions on Signal Processing},
            volume = {67},
            number = {6},
            pages = {1648--1661}
            }
          

Conference Articles

  1. M. Vono, N. Dobigeon, and P. Chainais, "Bayesian image restoration under Poisson noise and log-concave prior," in Proc. IEEE Int. Conf. Acoust., Speech, and Signal Processing (ICASSP), Brighton, U.K., May 2019.

    In recent years, much research has been devoted to the restoration of Poissonian images using optimization-based methods. On the other hand, the derivation of efficient and general fully Bayesian approaches is still an active area of research and especially if standard regularization functions are used, e.g. the total variation (TV) norm. This paper proposes to use the recent split-and-augmented Gibbs sampler (SPA) to sample efficiently from an approximation of the initial target distribution when log-concave prior distributions are used. SPA embeds proximal Markov chain Monte Carlo (MCMC) algorithms to sample from possibly non-smooth log-concave full conditionals. The benefit of the proposed approach is illustrated on several experiments including different regularizers, intensity levels and with both analysis and synthesis approaches.

            @Inproceedings{Vono_IEEE_ICASSP_2019b,
            author       = {M. Vono and N. Dobigeon and P. Chainais},
            title        = {Bayesian image restoration under {P}oisson noise and log-concave prior},
            booktitle    = {Proc. IEEE Int. Conf. Acoust., Speech, and Signal Processing (ICASSP)},
            address      = {Brighton, U.K.},
            month        = {May},
            year         = {2019},
            pages        = {},
    }
          
  2. M. Vono, N. Dobigeon, and P. Chainais, "Efficient sampling through variable splitting-inspired Bayesian hierarchical models," in Proc. IEEE Int. Conf. Acoust., Speech, and Signal Processing (ICASSP), Brighton, U.K., May 2019.

    Markov chain Monte Carlo (MCMC) methods are an important class of computation techniques to solve Bayesian inference problems. Much recent research has been dedicated to scale these algorithms in high-dimensional settings by relying on powerful optimization tools such as gradient information or proximity operators. In a similar vein, this paper proposes a new Bayesian hierarchical model to solve large scale inference problems by taking inspiration from variable splitting methods. Similarly to the latter, the derived Gibbs sampler permits to divide the initial sampling task into simpler ones. As a result, the proposed Bayesian framework can lead to a faster sampling scheme than state-of-the-art methods by embedding them. The strength of the proposed methodology is illustrated on two often-studied image processing problems.

            @Inproceedings{Vono_IEEE_ICASSP_2019a,
            author       = {M. Vono and N. Dobigeon and P. Chainais},
            title        = {Efficient sampling through variable splitting-inspired {B}ayesian hierarchical models},
            booktitle    = {Proc. IEEE Int. Conf. Acoust., Speech, and Signal Processing (ICASSP)},
            address      = {Brighton, U.K.},
            month        = {May},
            year         = {2019},
            pages        = {},
    }
          
  3. M. Vono, N. Dobigeon, and P. Chainais, “Sparse Bayesian binary logistic regression using the split-and-augmented Gibbs sampler,” in Proc. IEEE Int. Workshop Machine Learning for Signal Processing (MLSP), Aalborg, Denmark, 2018.
    Finalist for the Best Student Paper Awards.

    Logistic regression has been extensively used to perform classification in machine learning and signal/image processing. Bayesian formulations of this model with sparsity-inducing priors are particularly relevant when one is interested in drawing credibility intervals with few active coefficients. Along these lines, the derivation of efficient simulation-based methods is still an active research area because of the analytically challenging form of the binomial likelihood. This paper tackles the sparse Bayesian binary logistic regression problem by relying on the recent split-and-augmented Gibbs sampler (SPA). Contrary to usual data augmentation strategies, this Markov chain Monte Carlo (MCMC) algorithm scales in high dimension and divides the initial sampling problem into simpler ones. These sampling steps are then addressed with efficient state-of-the-art methods, namely proximal MCMC algorithms that can benefit from the recent closed-form expression of the proximal operator of the logistic cost function. SPA appears to be faster than efficient proximal MCMC algorithms and presents a reasonable computational cost compared to optimization-based methods with the advantage of producing credibility intervals. Experiments on handwritten digits classification problems illustrate the performances of the proposed approach.

            @inproceedings{Vono_MLSP18,
            author = {Vono, Maxime and Dobigeon, Nicolas and Chainais, Pierre},
            title = {Sparse {B}ayesian binary logistic regression using the split-and-augmented {G}ibbs sampler},
            year = {2018},
            booktitle = {Proc. IEEE Int. Workshop Machine Learning for Signal Processing (MLSP), 2018, Aalborg, Denmark}
            }
          

Talks

  1. May 2018 - Invited talk organized by team SigMA (CRIStAL laboratory, Lille, France)
    Split-and-augmented Gibbs sampler - A divide & conquer approach to solve large-scale inference problems

    Recently, a new class of Markov chain Monte Carlo (MCMC) algorithms took advantage of convex optimization to build efficient and fast sampling schemes from high-dimensional distributions. Variable splitting methods have become classical in optimization to divide difficult problems into simpler ones and have proven their efficiency in solving high-dimensional inference problems encountered in machine learning and signal processing. This paper derives two new optimization-driven sampling schemes inspired from variable splitting and data augmentation. In particular, the formulation of one of the proposed approaches is closely related to the alternating direction method of multipliers (ADMM) main steps. The proposed framework enables to derive faster and more efficient sampling schemes than the current state-of-the-art methods and can embed the latter. By sampling efficiently the parameter to infer as well as the hyperparameters of the problem, the generated samples can be used to approximate maximum a posteriori (MAP) and minimum mean square error (MMSE) estimators of the parameters to infer. Additionally, the proposed approach brings credibility intervals at a low cost contrary to optimization methods. Simulations on two often-studied signal processing problems illustrate the performance of the two proposed samplers. All results are compared to those obtained by recent state-of-the-art optimization and MCMC algorithms used to solve these problems.

Presentations

  1. Sept. 2018 - Poster presentation at Optimization and Learning workshop (Toulouse, France).
    Split-and-augmented Gibbs sampler - A divide & conquer approach to solve large-scale inference problems
  2. July 2018 - Poster presentation at LMS/CRiSM summer school on computational statistics (Univ. Warwick, UK).
    Split-and-augmented Gibbs sampler - A divide & conquer approach to solve large-scale inference problems
  3. July 2018 - Poster presentation at BNPSI 2018 workshop (Bordeaux, France).
    Split-and-augmented Gibbs sampler - A divide & conquer approach to solve large-scale inference problems

Consulting

LMlogo ITMlogo

Starting from September, 2018 I will work, as a data science/analytics consultant, for the Marketing & Strategy direction of an international supermarket chain called Intermarché Alimentaire International (ITM AI). These consultancy missions will be carried out in parallel of my Ph.D. studies and will be the opportunity to apply my previous experiences, knowledge and work to concrete and important issues for retailers, namely sales forecasting, pricing strategy and promotional events.

Some of the above issues were tackled in my previous internships and Master's thesis (in applied mathematics) where I was particularly interested in optimal pricing policies to apply in clearance and/or promotional events. To this purpose, I did my Master's thesis in partnership with the Pricing direction of a French home-improvement and gardening retailer called Leroy Merlin France (LMF) working on pricing policy optimization for clearance events. The optimal pricing strategy I derived during this work met the requirements and constraints of the Pricing direction of LMF (limited number of price changes, modeling of the buying process uncertainty, etc.), was tested on their past transactional data and was later proposed to some French brick and mortar stores.

CV

ToulouseUnivlogo LilleUnivlogo CentraleLillelogo

2017 - currently Ph.D. - Statistics
University of Toulouse, France

Optimization-driven Monte Carlo algorithms

2016 - 2017 M.Sc. - Applied Mathematics (Probability & Statistics)
University of Lille, France

Obtained with honors

2013 - 2017 M.Sc. - Engineering (Data Science)
Ecole Centrale de Lille, France

Including a gap year
Head of the class (rank: 1)

Find me

On the Web

Mailing address

INP - ENSEEIHT Toulouse
2, rue Charles Camichel
B.P. 7122
31071 Toulouse Cedex 7
France



View larger map
CRISTALlogo IRITlogo CNRSlogo Toulouselogo ENSEEIHTlogo

© 2017. All rights reserved.

Powered by a customized version of Hydejack v7.5.1