Michael I. Jordan

TL;DR: This paper presents simulations that show that systems that learn under the influence of an architectural bias have a number of desirable properties, including a tendency to decompose tasks into subtasks, to decouple the dynamics of recurrent subsystems, and to develop location-sensitive internal representations.

TL;DR: This work provides a new proof of the linear convergence of the alternating direction method of multipliers when one of the objective terms is strongly convex, and demonstrates that minimizing the derived bound on the convergence rate provides a practical approach to selecting algorithm parameters for particular ADMM instances.

TL;DR: In this paper, a new stochastic L-BFGS algorithm was proposed and proved to have a linear convergence rate for strongly convex and smooth functions, and the algorithm was shown to perform well for a wide range of step sizes.

TL;DR: This dissertation summarizes the work advancing the state of the art in all three of these areas of research in Warriors for Bayesian nonparametric latent feature models, presenting a non-exchangeable framework for generalizing and extending the original priors and introducing four concrete generalizations applicable when the authors have prior knowledge about object relationships that can be captured either via a tree or chain.

TL;DR: Working in the setting of kernel linear regression and kernel logistic regression, it is shown empirically that the effect of the block 1-norm regularization differs notably from the (non-block) 1- norm regularization commonly used for variable selection, and that the regularization path is of particular value in the block case.