arXiv research

Study reveals flatness of Hessian metrics with non-negative Ricci curvature on foliation leaves.

The paper introduces a new system of equations for Hessian-cscK metrics.

The loss function of deep networks is known to be non-convex but the precise nature of this nonconvexity is still an active area of research. In this work, we study the loss landscape of deep networks through the eigendecompositions of their Hessian matrix. In particular, we examine how important the negative eigenvalu…

In this paper, complex Hessian equation over Kähler manifold was studied. Under the condition that the underline Kähler manifold has non-negative holomorphic bisectional curvature, the existence and regularity of the solution was proved.

Classifies special homogeneous surfaces with unique properties.

New Bethe-Hessian method improves community detection in sparse networks.

Study classifies 3D Hessian manifolds, proving their topology.

Study of constant curvature hypersurfaces in hyperbolic space.

A new unbiased Hessian estimator for expectation-based objectives.

We study the long time behavior of the Hesse-Koszul flow on compact Hessian manifolds. When the first affine Chern class is negative, we prove that the flow converges to the unique Hesse-Einstein metric. We also derive a convergence result for a twisted Hesse-Koszul flow on any compact Hessian manifold. These results g…

Finite-sum optimization problems are ubiquitous in machine learning, and are commonly solved using first-order methods which rely on gradient computations. Recently, there has been growing interest in \emph{second-order} methods, which rely on both gradients and Hessians. In principle, second-order methods can require …

Let $(X,ω)$ be a compact Kähler manifold of dimension $n$, and fix $1\leq m\leq n.$ We prove that the complex Hessian equation $(ω+dd^c\varphi)^m\wedge ω^{n-m}=fω^n$, with $0<f\in \mathcal{C}^{\infty}(X)$ has a smooth admissible solution $\varphi\in \mathcal{C}^{\infty}(X)$. This was previously known to hold when $(X,…

The paper finds convex hypersurfaces with specific curvature properties.

Following P. M. H. Wilson's paper on sectional curvatures of Kahler moduli, we consider a natural Riemannian metric on a hypersurface f=1 in a real vector space, defined using the Hessian of a homogeneous polynomial f. We give examples to answer a question by Wilson about when this metric has nonpositive curvature. Als…

The paper describes a fine representation of the Ricci tensor and Hessian on RCD spaces.

Learning RBMs using standard algorithms such as CD(k) involves gradient descent on the negative log-likelihood. One of the terms in the gradient, which involves expectation w.r.t. the model distribution, is intractable and is obtained through an MCMC estimate. In this work we show that the Hessian of the log-likelihood…

The Ricci Calabi functional is a functional on the space of Kähler metrics of Fano manifolds. Its critical points are called generalized Kähler Einstein metrics. In this article, we show that the Hessian of the Ricci Calabi functional is non-negative at generalized Kähler Einstein metrics. As its application, we give a…

Approximate Newton methods are a standard optimization tool which aim to maintain the benefits of Newton's method, such as a fast rate of convergence, whilst alleviating its drawbacks, such as computationally expensive calculation or estimation of the inverse Hessian. In this work we investigate approximate Newton meth…

Constructs flows on manifolds with small curvature, proving Euclidean topology.

Noise injection regularizes Hessian, improving neural network training and generalization.

Proves smoothness and estimates for special Lagrangian solutions with semi-convexity.

The Hessian-vector product has been utilized to find a second-order stationary solution with strong complexity guarantee (e.g., almost linear time complexity in the problem's dimensionality). In this paper, we propose to further reduce the number of Hessian-vector products for faster non-convex optimization. Previous a…

Negative step sizes improve second-order methods for neural networks.

This work analyzes Adam's preconditioning effect on quadratic functions and quantifies its impact on condition number.

Let L be an ample bundle over a compact complex manifold X. Fix a Hermitian metric in L whose curvature defines a Kähler metric on X. The Hessian of Mabuchi energy is a fourth-order elliptic operator D on functions which arises in the study of scalar curvature. We quantise D by the Hessian E(k) of balancing energy, a f…

Let $P \subset \R^3$ be a polyhedron. It was conjectured that if $P$ is weakly convex (i. e. its vertices lie on the boundary of a strictly convex domain) and decomposable (i. e. $P$ can be triangulated without adding new vertices), then it is infinitesimally rigid. We prove this conjecture under a weak additional assu…

Researchers solve a complex equation to embed graphs with negative curvature.

Study bounds derivatives of solutions to a specific equation on domains.

We examine the squared error loss landscape of shallow linear neural networks. We show---with significantly milder assumptions than previous works---that the corresponding optimization problems have benign geometric properties: there are no spurious local minima and the Hessian at every saddle point has at least one ne…

Let (M,g) be a compact oriented Einstein 4-manifold. Write R-plus for the part of the curvature operator of g which acts on self-dual 2-forms. We prove that if R-plus is negative definite then g is locally rigid: any other Einstein metric near to g is isometric to it. This is a chiral generalisation of Koiso's Theorem,…

The Hessian of neural networks can be decomposed into a sum of two matrices: (i) the positive semidefinite generalized Gauss-Newton matrix G, and (ii) the matrix H containing negative eigenvalues. We observe that for wider networks, minimizing the loss with the gradient descent optimization maneuvers through surfaces o…

The expression (-1/u) times the Hessian of u transforms as a symmetric (0,2) tensor under projective coordinate transformations, so long as u transforms as a section of a certain line bundle. On a locally projectively flat manifold M, the section u can be regarded as a metric potential analogous to the local potential …

In this paper, we introduce a new energy density function $\mathscr Y$ on the projective bundle $\mathbb{P}(T_M)\>M$ for a smooth map $f:(M,h)\>(N,g)$ between Riemannian manifolds $\mathscr Y=g_{ij}f^i_αf^j_β\frac{W^αW^β}{\sum h_{γδ} W^γW^δ}.$ We get new Hessian estimates to this energy density and obtain various new…

In this paper, we study stochastic non-convex optimization with non-convex random functions. Recent studies on non-convex optimization revolve around establishing second-order convergence, i.e., converging to a nearly second-order optimal stationary points. However, existing results on stochastic non-convex optimizatio…

A new quasi-Newton method uses cubic regularization to avoid saddle points in deep learning.

Paper solves complex equations on noncompact manifolds.

Let $M$ be a closed Riemann surface, $N$ a Riemannian manifold of Hermitian non-positive curvature, $f:M\to N$ a continuous map, and $E$ the function on the Teichmüller space of $M$ that assigns to a complex structure on $M$ the energy of the harmonic map homotopic to $f$. We show that $E$ is a plurisubharmonic functio…

The study proves that certain noncompact Hessian manifolds are diffeomorphic to R^n.

New metric for probability measures connects physics and geometry.

Investigates regularity of solutions to complex Hessian equation.

Abstract: Survey on quadratic Hessian equations, their properties, and open problems.

Paper solves Hessian equations on Kähler manifolds.

New Hessian estimates for heat equations on manifolds.

In this work, we study the stability of Hopf vector fields on Lorentzian Berger spheres as critical points of the energy, the volume and the generalized energy. In order to do so, we construct a family of vector fields using the simultaneous eigenfunctions of the Laplacian and of the vertical Laplacian of the sphere. T…

Study calculates Hessian of Busemann function on Damek-Ricci spaces.

A new mutual information optimization method using self-supervised binary contrastive learning.

The paper shows infinitely many components in Floer Hessians space.

The paper describes flat Hessian metrics on surfaces and their potentials.