@article{Xue_Zhang_2016, title={Learning Unknown Structure in CRFs via Adaptive Gradient Projection Method}, volume={4}, url={http://www.iapress.org/index.php/soic/article/view/20160904}, DOI={10.19139/soic.v4i3.228}, abstractNote={<p>We study the problem of fitting probabilistic graphical models to the given data when the structure is not known. More specifically, we focus on learning unknown structure in conditional random fields, especially learning both the structure and parameters of a conditional random field model simultaneously. To do this, we first formulate the learning problem as a convex minimization problem by adding an l_2-regularization to the node parameters and a group l_1-regularization to the edge parameters, and then a gradient-based projection method is proposed to solve it which combines an adaptive stepsize selection strategy with a nonmonotone line search. Extensive simulation experiments are presented to show the performance of our approach in solving unknown structure learning problems.</p>}, number={3}, journal={Statistics, Optimization & Information Computing}, author={Xue, Wei and Zhang, Wensheng}, year={2016}, month={Aug.}, pages={243-251} }