Models of Noise and Robust Estimates
Given n noisy observations g; of the same quantity f, it is common use to give an estimate of f by minimizing the function Eni=1(gi-f)2. From a statistical point of view this corresponds to computing the Maximum likelihood estimate, under the assumption of Gaussian noise. However, it is well known t...
Main Author: | |
---|---|
Language: | en_US |
Published: |
2004
|
Online Access: | http://hdl.handle.net/1721.1/6564 |
_version_ | 1811073406407278592 |
---|---|
author | Girosi, Federico |
author_facet | Girosi, Federico |
author_sort | Girosi, Federico |
collection | MIT |
description | Given n noisy observations g; of the same quantity f, it is common use to give an estimate of f by minimizing the function Eni=1(gi-f)2. From a statistical point of view this corresponds to computing the Maximum likelihood estimate, under the assumption of Gaussian noise. However, it is well known that this choice leads to results that are very sensitive to the presence of outliers in the data. For this reason it has been proposed to minimize the functions of the form Eni=1V(gi-f), where V is a function that increases less rapidly than the square. Several choices for V have been proposed and successfully used to obtain "robust" estimates. In this paper we show that, for a class of functions V, using these robust estimators corresponds to assuming that data are corrupted by Gaussian noise whose variance fluctuates according to some given probability distribution, that uniquely determines the shape of V. |
first_indexed | 2024-09-23T09:32:34Z |
id | mit-1721.1/6564 |
institution | Massachusetts Institute of Technology |
language | en_US |
last_indexed | 2024-09-23T09:32:34Z |
publishDate | 2004 |
record_format | dspace |
spelling | mit-1721.1/65642019-04-11T02:52:27Z Models of Noise and Robust Estimates Girosi, Federico Given n noisy observations g; of the same quantity f, it is common use to give an estimate of f by minimizing the function Eni=1(gi-f)2. From a statistical point of view this corresponds to computing the Maximum likelihood estimate, under the assumption of Gaussian noise. However, it is well known that this choice leads to results that are very sensitive to the presence of outliers in the data. For this reason it has been proposed to minimize the functions of the form Eni=1V(gi-f), where V is a function that increases less rapidly than the square. Several choices for V have been proposed and successfully used to obtain "robust" estimates. In this paper we show that, for a class of functions V, using these robust estimators corresponds to assuming that data are corrupted by Gaussian noise whose variance fluctuates according to some given probability distribution, that uniquely determines the shape of V. 2004-10-04T15:31:30Z 2004-10-04T15:31:30Z 1991-11-01 AIM-1287 http://hdl.handle.net/1721.1/6564 en_US AIM-1287 112191 bytes 361984 bytes application/octet-stream application/pdf application/octet-stream application/pdf |
spellingShingle | Girosi, Federico Models of Noise and Robust Estimates |
title | Models of Noise and Robust Estimates |
title_full | Models of Noise and Robust Estimates |
title_fullStr | Models of Noise and Robust Estimates |
title_full_unstemmed | Models of Noise and Robust Estimates |
title_short | Models of Noise and Robust Estimates |
title_sort | models of noise and robust estimates |
url | http://hdl.handle.net/1721.1/6564 |
work_keys_str_mv | AT girosifederico modelsofnoiseandrobustestimates |