Robust Subspace-Constrained Quadratic Models for Low-Dimensional Structure Learning

In three linesNew SCQM method for learning low-dimensional structure from high-dimensional data. Generalizes SQMF framework to handle diverse noise distributions (generalized Gaussian, radial Laplace). Gradient-based algorithm with backtracking line-search ensures stable convergence. Numerical experiments demonstrate superior robustness and reconstruction accuracy versus existing methods.

Summary generated by Claude — human-verified