Hyperparameter estimation in maximum a posteriori regression using group sparsity with an application to brain imaging. Bekhti, Y., Badeau, R., & Gramfort, A. In 2017 25th European Signal Processing Conference (EUSIPCO), pages 246-250, Aug, 2017. Paper doi abstract bibtex Hyperparameter estimation is a recurrent problem in the signal and statistics literature. Popular strategies are cross-validation or Bayesian inference, yet it remains an active topic of research in order to offer better or faster algorithms. The models considered here are sparse regression models with convex or non-convex group-Lasso-like penalties. Following the recent work of Pereyra et al. [1] we study the fixed point iteration algorithm they propose and show that, while it may be suitable for an analysis prior, it suffers from limitations when using high-dimensional sparse synthesis models. The first contribution of this paper is to show how to overcome this issue. Secondly, we demonstrate how one can extend the model to estimate a vector of regularization parameters. We illustrate this on models with group sparsity reporting improved support recovery and reduced amplitude bias on the estimated coefficients. This approach is compared with an alternative method that uses a single parameter but a non-convex penalty. Results are presented on simulations and an inverse problem relevant for neuroscience which is the localization of brain activations using magneto/electroencephalography.
@InProceedings{8081206,
author = {Y. Bekhti and R. Badeau and A. Gramfort},
booktitle = {2017 25th European Signal Processing Conference (EUSIPCO)},
title = {Hyperparameter estimation in maximum a posteriori regression using group sparsity with an application to brain imaging},
year = {2017},
pages = {246-250},
abstract = {Hyperparameter estimation is a recurrent problem in the signal and statistics literature. Popular strategies are cross-validation or Bayesian inference, yet it remains an active topic of research in order to offer better or faster algorithms. The models considered here are sparse regression models with convex or non-convex group-Lasso-like penalties. Following the recent work of Pereyra et al. [1] we study the fixed point iteration algorithm they propose and show that, while it may be suitable for an analysis prior, it suffers from limitations when using high-dimensional sparse synthesis models. The first contribution of this paper is to show how to overcome this issue. Secondly, we demonstrate how one can extend the model to estimate a vector of regularization parameters. We illustrate this on models with group sparsity reporting improved support recovery and reduced amplitude bias on the estimated coefficients. This approach is compared with an alternative method that uses a single parameter but a non-convex penalty. Results are presented on simulations and an inverse problem relevant for neuroscience which is the localization of brain activations using magneto/electroencephalography.},
keywords = {Bayes methods;brain;convex programming;electroencephalography;inverse problems;iterative methods;maximum likelihood estimation;medical image processing;regression analysis;hyperparameter estimation;brain imaging;recurrent problem;faster algorithms;sparse regression models;high-dimensional sparse synthesis models;nonconvex penalty;inverse problem;brain activations;posteriori regression;Bayesian inference;Brain modeling;Estimation;Bayes methods;Signal processing;Inverse problems;Sensors;Europe},
doi = {10.23919/EUSIPCO.2017.8081206},
issn = {2076-1465},
month = {Aug},
url = {https://www.eurasip.org/proceedings/eusipco/eusipco2017/papers/1570347015.pdf},
}
Downloads: 0
{"_id":"cpeMMZuctxBXgTSAp","bibbaseid":"bekhti-badeau-gramfort-hyperparameterestimationinmaximumaposterioriregressionusinggroupsparsitywithanapplicationtobrainimaging-2017","authorIDs":[],"author_short":["Bekhti, Y.","Badeau, R.","Gramfort, A."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"firstnames":["Y."],"propositions":[],"lastnames":["Bekhti"],"suffixes":[]},{"firstnames":["R."],"propositions":[],"lastnames":["Badeau"],"suffixes":[]},{"firstnames":["A."],"propositions":[],"lastnames":["Gramfort"],"suffixes":[]}],"booktitle":"2017 25th European Signal Processing Conference (EUSIPCO)","title":"Hyperparameter estimation in maximum a posteriori regression using group sparsity with an application to brain imaging","year":"2017","pages":"246-250","abstract":"Hyperparameter estimation is a recurrent problem in the signal and statistics literature. Popular strategies are cross-validation or Bayesian inference, yet it remains an active topic of research in order to offer better or faster algorithms. The models considered here are sparse regression models with convex or non-convex group-Lasso-like penalties. Following the recent work of Pereyra et al. [1] we study the fixed point iteration algorithm they propose and show that, while it may be suitable for an analysis prior, it suffers from limitations when using high-dimensional sparse synthesis models. The first contribution of this paper is to show how to overcome this issue. Secondly, we demonstrate how one can extend the model to estimate a vector of regularization parameters. We illustrate this on models with group sparsity reporting improved support recovery and reduced amplitude bias on the estimated coefficients. This approach is compared with an alternative method that uses a single parameter but a non-convex penalty. Results are presented on simulations and an inverse problem relevant for neuroscience which is the localization of brain activations using magneto/electroencephalography.","keywords":"Bayes methods;brain;convex programming;electroencephalography;inverse problems;iterative methods;maximum likelihood estimation;medical image processing;regression analysis;hyperparameter estimation;brain imaging;recurrent problem;faster algorithms;sparse regression models;high-dimensional sparse synthesis models;nonconvex penalty;inverse problem;brain activations;posteriori regression;Bayesian inference;Brain modeling;Estimation;Bayes methods;Signal processing;Inverse problems;Sensors;Europe","doi":"10.23919/EUSIPCO.2017.8081206","issn":"2076-1465","month":"Aug","url":"https://www.eurasip.org/proceedings/eusipco/eusipco2017/papers/1570347015.pdf","bibtex":"@InProceedings{8081206,\n author = {Y. Bekhti and R. Badeau and A. Gramfort},\n booktitle = {2017 25th European Signal Processing Conference (EUSIPCO)},\n title = {Hyperparameter estimation in maximum a posteriori regression using group sparsity with an application to brain imaging},\n year = {2017},\n pages = {246-250},\n abstract = {Hyperparameter estimation is a recurrent problem in the signal and statistics literature. Popular strategies are cross-validation or Bayesian inference, yet it remains an active topic of research in order to offer better or faster algorithms. The models considered here are sparse regression models with convex or non-convex group-Lasso-like penalties. Following the recent work of Pereyra et al. [1] we study the fixed point iteration algorithm they propose and show that, while it may be suitable for an analysis prior, it suffers from limitations when using high-dimensional sparse synthesis models. The first contribution of this paper is to show how to overcome this issue. Secondly, we demonstrate how one can extend the model to estimate a vector of regularization parameters. We illustrate this on models with group sparsity reporting improved support recovery and reduced amplitude bias on the estimated coefficients. This approach is compared with an alternative method that uses a single parameter but a non-convex penalty. Results are presented on simulations and an inverse problem relevant for neuroscience which is the localization of brain activations using magneto/electroencephalography.},\n keywords = {Bayes methods;brain;convex programming;electroencephalography;inverse problems;iterative methods;maximum likelihood estimation;medical image processing;regression analysis;hyperparameter estimation;brain imaging;recurrent problem;faster algorithms;sparse regression models;high-dimensional sparse synthesis models;nonconvex penalty;inverse problem;brain activations;posteriori regression;Bayesian inference;Brain modeling;Estimation;Bayes methods;Signal processing;Inverse problems;Sensors;Europe},\n doi = {10.23919/EUSIPCO.2017.8081206},\n issn = {2076-1465},\n month = {Aug},\n url = {https://www.eurasip.org/proceedings/eusipco/eusipco2017/papers/1570347015.pdf},\n}\n\n","author_short":["Bekhti, Y.","Badeau, R.","Gramfort, A."],"key":"8081206","id":"8081206","bibbaseid":"bekhti-badeau-gramfort-hyperparameterestimationinmaximumaposterioriregressionusinggroupsparsitywithanapplicationtobrainimaging-2017","role":"author","urls":{"Paper":"https://www.eurasip.org/proceedings/eusipco/eusipco2017/papers/1570347015.pdf"},"keyword":["Bayes methods;brain;convex programming;electroencephalography;inverse problems;iterative methods;maximum likelihood estimation;medical image processing;regression analysis;hyperparameter estimation;brain imaging;recurrent problem;faster algorithms;sparse regression models;high-dimensional sparse synthesis models;nonconvex penalty;inverse problem;brain activations;posteriori regression;Bayesian inference;Brain modeling;Estimation;Bayes methods;Signal processing;Inverse problems;Sensors;Europe"],"metadata":{"authorlinks":{}},"downloads":0},"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/Roznn/EUSIPCO/main/eusipco2017url.bib","creationDate":"2021-02-13T16:38:25.518Z","downloads":0,"keywords":["bayes methods;brain;convex programming;electroencephalography;inverse problems;iterative methods;maximum likelihood estimation;medical image processing;regression analysis;hyperparameter estimation;brain imaging;recurrent problem;faster algorithms;sparse regression models;high-dimensional sparse synthesis models;nonconvex penalty;inverse problem;brain activations;posteriori regression;bayesian inference;brain modeling;estimation;bayes methods;signal processing;inverse problems;sensors;europe"],"search_terms":["hyperparameter","estimation","maximum","posteriori","regression","using","group","sparsity","application","brain","imaging","bekhti","badeau","gramfort"],"title":"Hyperparameter estimation in maximum a posteriori regression using group sparsity with an application to brain imaging","year":2017,"dataSources":["2MNbFYjMYTD6z7ExY","uP2aT6Qs8sfZJ6s8b"]}