Convex optimization methods for dimension reduction and coefficient estimation in multivariate linear regression (Q662292)

The paper is one of the many items recently published on the topic of nuclear norm minimisation as a heuristic for rank minimisation. The target application is multivariate linear regression. The authors study a variant of a smoothing method due to Nesterov and show that an appropriately tuned implementation of this algorithm can outperform a general purpose semidefinite programming solver. Numerical experiments reported in the paper are however not reproducible and it is likely that appropriate tuning of the algorithm (which can be potentially difficult and time-consuming) is instrumental to its success.