Details

Autor(en) / Beteiligte

• Dragomir, Radu-Alexandru
• d’Aspremont, Alexandre
• Bolte, Jérôme

Titel

Quartic First-Order Methods for Low-Rank Minimization

Ist Teil von

• Journal of optimization theory and applications, 2021-05, Vol.189 (2), p.341-363

Ort / Verlag

New York: Springer US

Erscheinungsjahr

2021

Link zum Volltext

Quelle

SpringerLink

Beschreibungen/Notizen

• We study a general nonconvex formulation for low-rank minimization problems. We use recent results on non-Euclidean first-order methods to provide efficient and scalable algorithms. Our approach uses the geometry induced by the Bregman divergence of well-chosen kernel functions; for unconstrained problems, we introduce a novel family of Gram quartic kernels that improve numerical performance. Numerical experiments on Euclidean distance matrix completion and symmetric nonnegative matrix factorization show that our algorithms scale well and reach state-of-the-art performance when compared to specialized methods.