Global Optimality of Local Search for Low Rank Matrix Recovery
نویسندگان
چکیده
We show that there are no spurious local minima in the non-convex factorized parametrization of low-rank matrixrecovery from incoherent linear measurements. With noisy measurements we show all local minima are very close to aglobal optimum. Together with a curvature bound at saddle points, this yields a polynomial time global convergenceguarantee for stochastic gradient descent from random initialization.
منابع مشابه
Structured Low-Rank Matrix Factorization: Global Optimality, Algorithms, and Applications
Recently, convex formulations of low-rank matrix factorization problems have received considerable attention in machine learning. However, such formulations often require solving for a matrix of the size of the data matrix, making it challenging to apply them to large scale datasets. Moreover, in many applications the data can display structures beyond simply being low-rank, e.g., images and vi...
متن کاملStructured Low-Rank Matrix Factorization: Optimality, Algorithm, and Applications to Image Processing
Recently, convex solutions to low-rank matrix factorization problems have received increasing attention in machine learning. However, in many applications the data can display other structures beyond simply being low-rank. For example, images and videos present complex spatio-temporal structures, which are largely ignored by current low-rank methods. In this paper we explore a matrix factorizat...
متن کاملOn Rank-Ordered Nested Multinomial Logit Model and D-Optimal Design for this Model
In contrast to the classical discrete choice experiment, the respondent in a rank-order discrete choice experiment, is asked to rank a number of alternatives instead of the preferred one. In this paper, we study the information matrix of a rank order nested multinomial logit model (RO.NMNL) and introduce local D-optimality criterion, then we obtain Locally D-optimal design for RO.NMNL models in...
متن کاملThe Global Optimization Geometry of Low-Rank Matrix Optimization
In this paper we characterize the optimization geometry of a matrix factorization problem where we aim to find n×r and m×r matrices U and V such that UV T approximates a given matrixX. We show that the objective function of the matrix factorization problem has no spurious local minima and obeys the strict saddle property not only for the exact-parameterization case where rank(X) = r, but also f...
متن کاملThe detection of 11th of March 2011 Tohoku's TEC seismo-ionospheric anomalies using the Singular Value Thresholding (SVT) method
The Total Electron Content (TEC) measured by the Global Positioning System (GPS) is useful for registering the pre-earthquake ionospheric anomalies appearing before a large earthquake. In this paper the TEC value was predicted using the singular value thresholding (SVT) method. Also, the anomaly is detected utilizing this predicted value and the definition of the threshold value, leading to the...
متن کامل