Implementation of Ji, S. & Ye, J., An accelerated gradient method for trace norm minimization, Proceedings of the 26th Annual International Conference on Machine Learning, 2009, 457-464 and comparison to Ridge regression.
This regression method is useful for multi-task learning of sparse representations.
Given two matrices 'U' and 'V', the function to minimize is:
where the right term is the sum of the singular values of 'A'.