Open michaelweylandt opened 5 years ago
In PR #112, @dansenglund implemented the generalized ADMM from the paper above. It may be worth comparing it to the exact ADMM proposed in the same paper. The generalized ADMM avoids a Sylvester equation and so was easier to implemented, but the exact ADMM has faster per iteration convergence and fewer tuning parameters, so may be more suitable for an algorithmic regularization approach.
This may require additional features in Eigen or a move to Armadillo which already supports Sylvester equations.
Test and Implement the new CBASS based on https://arxiv.org/abs/1901.06075