arXiv subjects

Po Chen

Publications and source records attributed to Po Chen.

5 recordsLinked to original sources

arXiv2026

Benign Landscape of Quadratic Programs with Orthogonality Constraints and Its Application to Heteroscedastic Probabilistic PCA

In this work, we study the optimization landscape of homogeneous quadratic programs with orthogonality constraints (QPOC) and apply the resulting theory to heteroscedastic probabilistic PCA (HePPCA). For QPOC, we establish a complete characterization of the benign optimization landscape by showing that every critical point is either a global maximizer or a strict saddle point. Our analysis builds on a closed-form characterization of the critical point set, from which we derive a necessary and sufficient condition for global optimality and show that every non-optimal critical point has a direction of positive curvature. As an application, we show that the population version of HePPCA is a special instance of QPOC and therefore has a benign optimization landscape; moreover, it satisfies local geodesic strong concavity near every global maximizer. We furthermore prove that, when the sample size is sufficiently large, the sample version of HePPCA inherits these favorable properties with high probability. Together with existing theory on avoiding strict saddle points, our results provide a theoretical justification for the observed local linear convergence of retraction-based optimization methods to global solutions for both QPOC and HePPCA. Finally, we present numerical experiments to corroborate our theoretical results.

math.OC↗

arXiv2025

A Complete Loss Landscape Analysis of Regularized Deep Matrix Factorization

Despite its wide range of applications across various domains, the optimization foundations of deep matrix factorization (DMF) remain largely open. In this work, we aim to fill this gap by conducting a comprehensive study of the loss landscape of the regularized DMF problem. Toward this goal, we first provide a closed-form characterization of all critical points of the problem. Building on this, we establish precise conditions under which a critical point is a local minimizer, a global minimizer, a strict saddle point, or a non-strict saddle point. Leveraging these results, we derive a necessary and sufficient condition under which every critical point is either a local minimizer or a strict saddle point. This provides insights into why gradient-based methods almost always converge to a local minimizer of the regularized DMF problem. Finally, we conduct numerical experiments to visualize its loss landscape to support our theory.

math.OC↗

arXiv2025

Error Bound Analysis for the Regularized Loss of Deep Linear Neural Networks

The optimization foundations of deep linear networks have recently received significant attention. However, due to their inherent non-convexity and hierarchical structure, analyzing the loss functions of deep linear networks remains a challenging task. In this work, we study the local geometry of the regularized squared loss of deep linear networks around each critical point. Specifically, we obtain a closed-form characterization of the critical point set building on existing results and establish an error bound for the regularized loss under mild conditions on network width and regularization parameters. Notably, this error bound quantifies the distance from a point to the critical point set in terms of the current gradient norm, which can be used to derive linear convergence of first-order methods. To support our theoretical findings, we conduct numerical experiments and demonstrate that gradient descent converges linearly to a critical point when optimizing the regularized loss of deep linear networks.

math.OC↗

arXiv2023

Ground-state phase diagram of two-component interacting bosons on a two-leg ladder

Using the cluster Gutzwiller mean-field method, we numerically study the ground-state phase diagram of the non-hard-core two-component interacting bosons trapped in a two-leg ladder with and without an artificial magnetic field. There are three quantum phases namely Mott insulator (MI), supercounterfluid (SCF), and superfluid (SF) are found in the phase diagram. Interestingly, several loophole SCF phases are observed at a sufficiently small intra- to inter-leg hopping ratio when the magnetic flux is absent. While if the ratio is not so small, the loophole SCF phase would disappear, but it can still be induced by applying a sufficiently large magnetic flux. Additionally, we also find that the presence of the magnetic flux leads to an enlargement of the MI lobe and the conventional SCF lobe. Moreover, the SF-MI phase boundary is quantitatively consistent with the strong-couping expansion at a weak hopping amplitude.

cond-mat.quant-gas↗

arXiv2023

Ground state properties of a multi-component bosonic mixture: a Gutzwiller mean-field study

Using the single-site Gutzwiller method, we theoretically study the ground state and the interspecies entanglement properties of interexchange symmetric multi-component (two- and three-) bosonic mixtures in an optical lattice, and the results are generalized to an $n$-component ($n=2,3,4,\cdots$) system. We compute the mean-field phase diagram, the interspecies entanglement entropy, and the ground state spectral decomposition. Three phases namely the $n$-component Superfluid state (nSF), the $n$-component Mott insulator state (nMI), and the Super-counter-fluid state (SCF) are observed. Interestingly, we find that there are $n-1$ SCF lobes to separate every two neighboring nMI lobes in the phase diagram. More importantly, we derive the exact general expression of the interspecies entanglement entropy for the SCF phase. In addition, we also investigate the demixing effect of an n-component mixture and demonstrate that the mixing-demixing critical point is independent of n.

cond-mat.quant-gas↗