A Randomised Subspace Gauss-Newton Method for Nonlinear Least-Squares
This work addresses optimization efficiency for nonlinear least-squares problems, but appears incremental as it builds on existing Gauss-Newton methods with randomization.
The paper tackled nonlinear least-squares optimization by proposing a Randomised Subspace Gauss-Newton (R-SGN) algorithm, which achieved a sublinear global convergence rate matching deterministic methods in accuracy tolerance, with promising results on logistic and nonlinear regression problems.
We propose a Randomised Subspace Gauss-Newton (R-SGN) algorithm for solving nonlinear least-squares optimization problems, that uses a sketched Jacobian of the residual in the variable domain and solves a reduced linear least-squares on each iteration. A sublinear global rate of convergence result is presented for a trust-region variant of R-SGN, with high probability, which matches deterministic counterpart results in the order of the accuracy tolerance. Promising preliminary numerical results are presented for R-SGN on logistic regression and on nonlinear regression problems from the CUTEst collection.