Understanding Over-parametrization Through Matrix Sensing
1,636 views · Published 25 January 2018 · 1:17:49 · Indexed 26 October 2026
Channel: Microsoft Research · 2018 · Science & Technology
We study the problem of recovering a low-rank matrix from linear measurements using an over-parameterized model. We show that the gradient descent process on the square loss function, starting from a small initialization, can converge to the ground truth matrix. Although the total number of observations is much smaller than the total number of parameters. See more at https://www.microsoft.com/en-us/research/video/understanding-over-parametrization-through-matrix-sensing/