The least favorable noise

03/17/2021
by   Philip A. Ernst, et al.
0

Suppose that a random variable X of interest is observed perturbed by independent additive noise Y. This paper concerns the "the least favorable perturbation" Ŷ_, which maximizes the prediction error E(X-E(X|X+Y))^2 in the class of Y with (Y)≤. We find a characterization of the answer to this question, and show by example that it can be surprisingly complicated. However, in the special case where X is infinitely divisible, the solution is complete and simple. We also explore the conjecture that noisier Y makes prediction worse.

READ FULL TEXT

Authors

page 1

page 2

page 3

page 4

07/29/2017

Refuting Feder, Kinne and Rafiey

I give an example showing that the recent claimed solution by Feder, Kin...
09/13/2018

PhD Dissertation: Generalized Independent Components Analysis Over Finite Alphabets

Independent component analysis (ICA) is a statistical method for transfo...
09/13/2018

Generalized Independent Components Analysis Over Finite Alphabets

Independent component analysis (ICA) is a statistical method for transfo...
05/21/2018

Learning Maximum-A-Posteriori Perturbation Models for Structured Prediction in Polynomial Time

MAP perturbation models have emerged as a powerful framework for inferen...
05/09/2012

Identifying confounders using additive noise models

We propose a method for inferring the existence of a latent common cause...
10/30/2019

Dynamically Protecting Privacy, under Uncertainty

We propose and analyze the ε-Noisy Goal Prediction Game to study a funda...
This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.