Penalized Estimation in Large-Scale Generalized Linear Array Models
Research output: Contribution to journal › Journal article › Research › peer-review
Standard
Penalized Estimation in Large-Scale Generalized Linear Array Models. / Lund, Adam; Vincent, Martin; Hansen, Niels Richard.
In: Journal of Computational and Graphical Statistics, Vol. 26, No. 3, 2017, p. 709-724.Research output: Contribution to journal › Journal article › Research › peer-review
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - JOUR
T1 - Penalized Estimation in Large-Scale Generalized Linear Array Models
AU - Lund, Adam
AU - Vincent, Martin
AU - Hansen, Niels Richard
PY - 2017
Y1 - 2017
N2 - Large-scale generalized linear array models (GLAMs) can be challenging to fit. Computation and storage ofits tensor product design matrix can be impossible due to time and memory constraints, and previously considereddesign matrix free algorithms do not scale well with the dimension of the parameter vector. A newdesign matrix free algorithm is proposed for computing the penalized maximum likelihood estimate forGLAMs, which, in particular, handles nondifferentiable penalty functions. The proposed algorithm is implementedand available via the R package glamlasso. It combines several ideas—previously consideredseparately—to obtain sparse estimates while at the same time efficiently exploiting the GLAM structure.In this article, the convergence of the algorithm is treated and the performance of its implementation isinvestigated and compared to that of glmnet on simulated as well as real data. It is shown that the computationtime for glamlasso scales favorably with the size of the problem when compared to glmnet.Supplementary materials, in the form of R code, data, and visualizations of results, are available online
AB - Large-scale generalized linear array models (GLAMs) can be challenging to fit. Computation and storage ofits tensor product design matrix can be impossible due to time and memory constraints, and previously considereddesign matrix free algorithms do not scale well with the dimension of the parameter vector. A newdesign matrix free algorithm is proposed for computing the penalized maximum likelihood estimate forGLAMs, which, in particular, handles nondifferentiable penalty functions. The proposed algorithm is implementedand available via the R package glamlasso. It combines several ideas—previously consideredseparately—to obtain sparse estimates while at the same time efficiently exploiting the GLAM structure.In this article, the convergence of the algorithm is treated and the performance of its implementation isinvestigated and compared to that of glmnet on simulated as well as real data. It is shown that the computationtime for glamlasso scales favorably with the size of the problem when compared to glmnet.Supplementary materials, in the form of R code, data, and visualizations of results, are available online
KW - Generalized linear array models
KW - Multidimensional smoothing
KW - Penalized estimation
KW - Proximal gradient algorithm
U2 - 10.1080/10618600.2017.1279548
DO - 10.1080/10618600.2017.1279548
M3 - Journal article
VL - 26
SP - 709
EP - 724
JO - Journal of Computational and Graphical Statistics
JF - Journal of Computational and Graphical Statistics
SN - 1061-8600
IS - 3
ER -
ID: 184322927