arXiv research

PAC learning sample complexity is decidable with finite support bounds.

The paper improves SVM margin-based generalization bounds.

KSG mutual information estimator, which is based on the distances of each sample to its k-th nearest neighbor, is widely used to estimate mutual information between two continuous random variables. Existing work has analyzed the convergence rate of this estimator for random variables whose densities are bounded away fr…

The paper analyzes kNN density estimation's convergence rates under different conditions.

SVM and linear regression models coincide in high dimensions.

Bayesian model averaging under predictor redundancy

Theory for deep neural network approximation of score function and its derivatives.

The paper provides tighter error bounds for GPR under bounded support noise.

Kernel-based online learning has often shown state-of-the-art performance for many online learning tasks. It, however, suffers from a major shortcoming, that is, the unbounded number of support vectors, making it non-scalable and unsuitable for applications with large-scale datasets. In this work, we study the problem …

New method calibrates reference distributions for bounded support.

New bounds on continuous random variables' right-tail probabilities.

In this paper we write explicitly the open book decompositions of links of quotient surface singularities supporting the corresponding unique Milnor fillable contact structure. The page-genus of these Milnor open books are minimal among all Milnor open books supporting the same contact structure. We also investigate wh…

Sharp bounds derived for the first two Steklov eigenvalues of exterior domains.

Study improves density estimation for compact domains using $h$-lifted KL divergence.

k Nearest Neighbor (kNN) method is a simple and popular statistical method for classification and regression. For both classification and regression problems, existing works have shown that, if the distribution of the feature vector has bounded support and the probability density function is bounded away from zero in i…

New SVM model balances sparsity and robustness in noisy data.

In this paper, we study the evolution of $L^2$ one forms under Ricci flow with bounded curvature on a non-compact Rimennian manifold. We show on such a manifold that the $L^2$ norm of a smooth one form with compact support is non-increasing along the Ricci flow with bounded curvature. The $L^{\infty}$ norm is showed to…

We propose a Generalized Dantzig Selector (GDS) for linear models, in which any norm encoding the parameter structure can be leveraged for estimation. We investigate both computational and statistical aspects of the GDS. Based on conjugate proximal operator, a flexible inexact ADMM framework is designed for solving GDS…

We investigate the optimality of perturbation based algorithms in the stochastic and adversarial multi-armed bandit problems. For the stochastic case, we provide a unified regret analysis for both sub-Weibull and bounded perturbations when rewards are sub-Gaussian. Our bounds are instance optimal for sub-Weibull pertur…

For binary classification we establish learning rates up to the order of $n^{-1}$ for support vector machines (SVMs) with hinge loss and Gaussian RBF kernels. These rates are in terms of two assumptions on the considered distributions: Tsybakov's noise assumption to establish a small estimation error, and a new geometr…

The Lasso performs well in ultra-sparse linear models with finite support size.

The paper improves support recovery in high-dimensional precision matrix estimation using meta learning.

Study on convex surfaces in Minkowski space, proving completeness and incompleteness conditions.

New inequalities for unbounded functions improve denoising score matching.

PBM mechanism improves privacy and accuracy in federated learning.

We study the existence of D-brane bound states at threshold in Type II string theories. In a number of situations, we can reduce the question of existence to quadrature, and the study of a particular limit of the propagator for the system of D-branes. This involves a derivation of an index theorem for a family of non-F…

Unified learning bound for covariate and concept shifts.

Exact bounds derived for neural network outputs with noisy inputs.

The algorithm given by Akbulut-Ozbagci constructs an explicit open book decomposition on a contact three-manifold described by a contact surgery on a link in the three-sphere. In this article, we will improve this algorithm by using Giroux's contact cell decomposition process. Our algorithm is more economical on choosi…

We derive a novel norm that corresponds to the tightest convex relaxation of sparsity combined with an $\ell_2$ penalty. We show that this new {\em $k$-support norm} provides a tighter relaxation than the elastic net and is thus a good replacement for the Lasso or the elastic net in sparse prediction problems. Through …

Study on consensus formation in manifolds with curvature constraints.

CIVET method provides robustness guarantees for VAEs under adversarial attacks.

Sharp Sobolev inequality derived for Riemannian manifolds with bounded Ricci curvature.

Study improves $H$-consistency bounds for regression analysis.

New Stein identity for q-Gaussians reduces gradient variance in machine learning.

We prove that if a contact manifold $(M,ξ)$ is supported by a planar open book, then Euler characteristic and signature of any Stein filling of $(M,ξ)$ is bounded. We also prove a similar finiteness result for contact manifolds supported by spinal open books with planar pages. Moving beyond the geography of Stein filli…

Nonparametric Thompson Sampling achieves optimal regret for risk-averse bandits with sub-Gaussian rewards.

The geodesic distance vanishes on the group of compactly supported diffeomorphisms of a Riemannian manifold $M$ of bounded geometry, for the right invariant weak Riemannian metric which is induced by the Sobolev metric $H^s$ of order $0\le s<\tfrac12$ on the Lie algebra $\mathfrak X_c(M)$ of vector fields with compact …

The support recovery problem consists of determining a sparse subset of variables that is relevant in generating a set of observations. In this paper, we study the support recovery problem in the phase retrieval model consisting of noisy phaseless measurements, which arises in a diverse range of settings such as optica…

Recent work on follow the perturbed leader (FTPL) algorithms for the adversarial multi-armed bandit problem has highlighted the role of the hazard rate of the distribution generating the perturbations. Assuming that the hazard rate is bounded, it is possible to provide regret analyses for a variety of FTPL algorithms f…

Study provides bounds for estimating intrinsic dimension using Gaussian kernels.

Optimal learning for parametric prophet inequalities with exponential-type distributions

We provide a formulation for Local Support Vector Machines (LSVMs) that generalizes previous formulations, and brings out the explicit connections to local polynomial learning used in nonparametric estimation literature. We investigate the simplest type of LSVMs called Local Linear Support Vector Machines (LLSVMs). For…

Upper bound for max-sliced 2-Wasserstein distance between measures.

Study decay and compact support of solutions to certain nonlinear PDEs.

Study on Wasserstein barycenters with computational hardness and fast algorithm development.

We propose an efficient distributed online learning protocol for low-latency real-time services. It extends a previously presented protocol to kernelized online learners that represent their models by a support vector expansion. While such learners often achieve higher predictive performance than their linear counterpa…

For a convex domain $D$ that is enclosed by the hypersurface $\partial D$ of bounded normal curvature, we prove an angle comparison theorem for angles between $\partial D$ and geodesic rays starting from some fixed point in $D$, and the corresponding angles for hypersurfaces of constant normal curvature. Also, we obtai…