Source author record

Luca Amodei

Explore this author’s publications, collaborators and research topics. This catalog record has not been claimed by its author.

ResearcherUnclaimed source record

math.OC math.DG math.NA

Catalog footprint

Research overview

2works

3topics

2close collaborators

Actions

Is this your work?

Open graph Browse papers

preprint2022arXiv

Revisiting second-order optimality conditions for equality-contrained minimization problem

The aim of this note is to give a geometric insight into the classical second order optimality conditions for equality-constrained minimization problem. We show that the Hessian's positivity of the Lagrangian function associated to the problem at a local minimum point $x^*$ corresponds to inequalities between the respective algebraic curvatures at point $x^*$ of the hypersurface $\mathcal{M}_{f, x^*}=\{ x \in \R^n \, | \, f(x) = f(x^*)\}$ defined by the objective function $f$ and the submanifold $\mathcal{M}_g = \{ x \in \R^n \, | \, g(x)= 0 \}$ defining the contraints. These inequalities highlight a geometric evidence on how, in order to guarantee the optimality, the submanifold $\mathcal{M}_g$ has to be locally included in the half space $\mathcal{M}_{f, x^*}^+ = \{ x \in \R^n \, | \, f(x) \geq f(x^*)\}$ limited by the hypersurface $\mathcal{M}_{f, x^*}.$ This presentation can be used for educational purposes and help to a better understanding of this property.

preprint2012arXiv

Two Newton methods on the manifold of fixed-rank matrices endowed with Riemannian quotient geometries

We consider two Riemannian geometries for the manifold $\mathcal{M}(p,m\times n)$ of all $m\times n$ matrices of rank $p$ . The geometries are induced on $\mathcal{M}(p,m\times n)$ by viewing it as the base manifold of the submersion $π:(M,N)\mapsto MN^T$ , selecting an adequate Riemannian metric on the total space, and turning $π$ into a Riemannian submersion. The theory of Riemannian submersions, an important tool in Riemannian geometry, makes it possible to obtain expressions for fundamental geometric objects on $\mathcal{M}(p,m\times n)$ and to formulate the Riemannian Newton methods on $\mathcal{M}(p,m\times n)$ induced by these two geometries. The Riemannian Newton methods admit a stronger and more streamlined convergence analysis than the Euclidean counterpart, and the computational overhead due to the Riemannian geometric machinery is shown to be mild. Potential applications include low-rank matrix completion and other low-rank matrix approximation problems.

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.