arXiv research

Gradient flow preserves speed for integral Menger curvature curves.

We study polygonal analogues of several moving boundary problems and their time discretization which preserves the constant area speed property. We establish various polygonal analogues of geometric formulas for moving boundaries and make use of the geometric formulas for our numerical scheme and its analysis of genera…

The paper studies how convex hypersurfaces in hyperbolic space evolve under a specific curvature flow.

We consider a convex Euclidean hypersurface that evolves by a volume or area preserving flow with speed given by a general nonhomogeneous function of the mean curvature. For a broad class of possible speed functions, we show that any closed convex hypersurface converges to a round sphere. The proof is based on the mono…

New flow method solves Christoffel-Minkowski problem.

Convexity preserved in curved surfaces moving at concave speeds.

We consider the evolution of a closed convex hypersurface under a volume preserving curvature flow. The speed is given by a power of the m-th mean curvature plus a volume preserving term, including the case of powers of the mean curvature or of the Gauss curvature. We prove that if the initial hypersurface satisfies a …

General area-preserving motion of polygonal curves is formulated as a system of ODEs. Solution polygonal curves belong to a prescribed polygonal class, which is similar to the admissible class used in the crystalline curvature flow. The ODEs are discretized implicitly in time keeping a given constant area speed while s…

TTRP method preserves distances in high-dimensional data with reduced storage and speed.

Study improves privacy-preserving online prediction from experts with speed-ups.

GenFormer uses deep learning to generate complex stochastic data.

New neural network captures spatial correlations in wind speed predictions.

In this paper we consider a class of weighted-volume preserving curvature flows acting on hypersurfaces that are trapped within two parallel hyperplanes and satisfy an orthogonal boundary condition. In the author's thesis the stability of cylinders under the flows was considered; it was found that they are stable provi…

DistillKac generates images quickly using damped wave equations.

We study a volume preserving curvature flow of convex hypersurfaces, driven by a power of the $k$-th elementary symmetric polynomial in the principal curvatures. Unlike most of the previous works on related problems, we do not require assumptions on the curvature pinching of the initial datum. We prove that the solutio…

One of the difficulties of training deep neural networks is caused by improper scaling between layers. Scaling issues introduce exploding / gradient problems, and have typically been addressed by careful scale-preserving initialization. We investigate the value of preserving scale, or isometry, beyond the initial weigh…

Study shows how curved surfaces evolve smoothly to spherical shapes.

PASCO speeds up graph clustering for large graphs.

We study a volume/area preserving curvature flow of hypersurfaces that are convex by horospheres in the hyperbolic space, with velocity given by a generic positive, increasing function of the mean curvature, not necessarly homogeneous. For this class of speeds we prove the exponential convergence to a geodesic sphere. …

Many reinforcement learning applications involve the use of data that is sensitive, such as medical records of patients or financial information. However, most current reinforcement learning methods can leak information contained within the (possibly sensitive) data on which they are trained. To address this problem, w…

We consider the quermassintegral preserving flow of closed \emph{h-convex} hypersurfaces in hyperbolic space with the speed given by any positive power of a smooth symmetric, strictly increasing, and homogeneous of degree one function $f$ of the principal curvatures which is inverse concave and has dual $f_*$ approachi…

NoTMF forecasts sparse urban road movement speeds with nonstationary temporal matrix factorization.

Efficient privacy-preserving machine learning framework using random transformations.

Convex hypersurfaces in hyperbolic space evolve to geodesic spheres.

Curvature flows in hyperbolic space preserve positive sectional curvature and contract to a point.

In this paper we discuss the stability of geodesic spheres in $\mathbb{S}^{n+1}$ under constrained curvature flows. We prove that under some standard assumptions on the speed and weight functions, the spheres are stable under perturbations that preserve a volume type quantity. This extends results by Escher and Simonet…

Gradient sparsification enhances privacy-preserving machine learning models.

AdaScale SGD adapts learning rates for large-batch training efficiently.

It is inevitable to train large deep learning models on a large-scale cluster equipped with accelerators system. Deep gradient compression would highly increase the bandwidth utilization and speed up the training process but hard to implement on ring structure. In this paper, we find that redundant gradient and gradien…

This paper concerns the evolution of a closed hypersurface of dimension $n(\geq 2)$ in the Euclidean space ${\mathbb{R}}^{n+1}$ under a mixed volume preserving flow. The speed equals a power $β(\geq 1)$ of homogeneous, either convex or concave, curvature functions of degree one plus a mixed volume preserving term, incl…

This paper speeds up SVC clustering by compressing data while preserving key properties.

Improved kernel herding algorithm for faster quadrature rule convergence.

Powered by machine learning services in the cloud, numerous learning-driven mobile applications are gaining popularity in the market. As deep learning tasks are mostly computation-intensive, it has become a trend to process raw data on devices and send the deep neural network (DNN) features to the cloud, where the feat…

Extends driving model to control agent behavior in simulations.

The paper studies a flow of surfaces in spacetime with a focus on curvature evolution.

Over a compact oriented manifold, the space of Riemannian metrics and normalised positive volume forms admits a natural pseudo-Riemannian metric $G$, which is useful for the study of Perelman's $\mathcal{W}$ functional. We show that if the initial speed of a $G$-geodesic is $G$-orthogonal to the tangent space to the or…

SharedMF uses secret sharing to protect privacy in distributed recommendation systems.

This paper studies the problem of error-runtime trade-off, typically encountered in decentralized training based on stochastic gradient descent (SGD) using a given network. While a denser (sparser) network topology results in faster (slower) error convergence in terms of iterations, it incurs more (less) communication …

We study flows of hypersurfaces in Riemannian manifolds with specific curvature speeds.

The paper studies curvature measures and volume-preserving flows on convex bodies.

This paper concerns closed hypersurfaces of dimension $n(\geq 2)$ in the hyperbolic space ${\mathbb{H}}_κ^{n+1}$ of constant sectional curvature $κ$ evolving in direction of its normal vector, where the speed is given by a power $β(\geq 1/m)$ of the $m$th mean curvature plus a volume preserving term, including the case…

NCVis speeds up data visualization for large datasets.

We consider the flow of closed convex hypersurfaces in Euclidean space $\mathbb{R}^{n+1}$ with speed given by a power of the $k$-th mean curvature $E_k$ plus a global term chosen to impose a constraint involving the enclosed volume $V_{n+1}$ and the mixed volume $V_{n+1-k}$ of the evolving hypersurface. We prove that i…

Quantum algorithm speeds up Gibbs partition function estimation.

New optimizers improve stock market forecasting accuracy.

Deep networks with orthogonal weights show stable fluctuations, improving generalization and training speed.

In this work, we define a collaborative and privacy-preserving machine teaching paradigm with multiple distributed teachers. We focus on consensus super teaching. It aims at organizing distributed teachers to jointly select a compact while informative training subset from data hosted by the teachers to make a learner l…

Framework preserves emergent physics in non-equilibrium systems from particle trajectories.