1: \begin{abstract}

2: Non-linear least squares solvers are used across a broad range of offline and real-time model fitting problems.

3: %Despite innovations in convergence, the efficiency of commonly used solvers has hardly improved in the two centuries since Gauss-Newton was proposed.

4: Most improvements of the basic Gauss-Newton algorithm tackle convergence guarantees or leverage the sparsity of the underlying problem structure for computational speedup.

5: With the success of deep learning methods leveraging large datasets, stochastic optimization methods received recently a lot of attention.

6: Our work borrows ideas from both stochastic machine learning and statistics, and we present an approach for non-linear least-squares that guarantees convergence while at the same time significantly reduces the required amount of computation.

7: Empirical results show that our proposed method achieves competitive convergence rates compared to traditional second-order approaches on common computer vision problems, such as image alignment and essential matrix estimation, with very large numbers of residuals.

8: \end{abstract}

9: