Výsledky vyhledávání - "Williams, Elliott"

Report

Probing the Transition to Dataset-Level Privacy in ML Models Using an Output-Specific and Data-Resolved Privacy Profile

Autor: LeBlond, Tyler, Munoz, Joseph, Lu, Fred, Fuchs, Maya, Zaresky-Williams, Elliott, Raff, Edward, Testa, Brian

Differential privacy (DP) is the prevailing technique for protecting user data in machine learning models. However, deficits to this framework include a lack of clarity for selecting the privacy budget $\epsilon$ and a lack of quantification for the

Externí odkaz: http://arxiv.org/abs/2306.15790

Zobrazit plný text záznamu

Report

A General Framework for Auditing Differentially Private Machine Learning

Autor: Lu, Fred, Munoz, Joseph, Fuchs, Maya, LeBlond, Tyler, Zaresky-Williams, Elliott, Raff, Edward, Ferraro, Francis, Testa, Brian

We present a framework to statistically audit the privacy guarantee conferred by a differentially private machine learner in practice. While previous works have taken steps toward evaluating privacy loss through poisoning attacks or membership infere

Externí odkaz: http://arxiv.org/abs/2210.08643

Zobrazit plný text záznamu

Elektronická kniha

Transported to Botany Bay : Class, National Identity, and the Literary Figure of the Australian Convict

Autor: Dorice Williams Elliott

Literary representations of British convicts exiled to Australia were the most likely way that the typical English reader would learn about the new colonies there. In Transported to Botany Bay, Dorice Williams Elliott examines how writers—from cano

Zobrazit plný text záznamu

Report

An Algorithm for Approximating Continuous Functions on Compact Subsets with a Neural Network with one Hidden Layer

Autor: Zaresky-Williams, Elliott

George Cybenko's landmark 1989 paper showed that there exists a feedforward neural network, with exactly one hidden layer (and a finite number of neurons), that can arbitrarily approximate a given continuous function $f$ on the unit hypercube. The pa

Externí odkaz: http://arxiv.org/abs/1902.03638

Zobrazit plný text záznamu

Report

Non-trivial Poincare Constant in any Subset of $\mathbb{R}^n$

Autor: Zaresky-Williams, Elliott

The Poincare Inequality is an extremely useful tool in the analysis of PDEs. A significant amount of literature has dealt with finding the optimal constant $C(p,\Omega)$, depending only the domain $\Omega$, and the $L^p$ norm in question. For convex

Externí odkaz: http://arxiv.org/abs/1811.07470

Zobrazit plný text záznamu