Optimality of the Plug-in Estimator for Differential Entropy Estimation under Gaussian Convolutions
© 2019 IEEE. This paper establishes the optimality of the plugin estimator for the problem of differential entropy estimation under Gaussian convolutions. Specifically, we consider the estimation of the differential entropy h(X + Z), where X and Z are independent d-dimensional random variables with...
Main Authors: | , , , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
Institute of Electrical and Electronics Engineers (IEEE)
2021
|
Online Access: | https://hdl.handle.net/1721.1/137042 |
_version_ | 1811071971422633984 |
---|---|
author | Goldfeld, Ziv Greenewald, Kristjan Weed, Jonathan Polyanskiy, Yury |
author2 | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory |
author_facet | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory Goldfeld, Ziv Greenewald, Kristjan Weed, Jonathan Polyanskiy, Yury |
author_sort | Goldfeld, Ziv |
collection | MIT |
description | © 2019 IEEE. This paper establishes the optimality of the plugin estimator for the problem of differential entropy estimation under Gaussian convolutions. Specifically, we consider the estimation of the differential entropy h(X + Z), where X and Z are independent d-dimensional random variables with Z{\sim}\mathcal{N}( {0,{σ ^2}{{\text{I}}-d}} ). The distribution of X is unknown and belongs to some nonparametric class, but n independently and identically distributed samples from it are available. We first show that despite the regularizing effect of noise, any good estimator (within an additive gap) for this problem must have an exponential in d sample complexity. We then analyze the absolute-error risk of the plug-in estimator and show that it converges as frac{{{c^d}}}{{n }}, thus attaining the parametric estimation rate. This implies the optimality of the plug-in estimator for the considered problem. We provide numerical results comparing the performance of the plug-in estimator to general-purpose (unstructured) differential entropy estimators (based on kernel density estimation (KDE) or k nearest neighbors (kNN) techniques) applied to samples of X + Z. These results reveal a significant empirical superiority of the plug-in to state-of-the-art KDE- and kNN-based methods. |
first_indexed | 2024-09-23T08:58:51Z |
format | Article |
id | mit-1721.1/137042 |
institution | Massachusetts Institute of Technology |
language | English |
last_indexed | 2024-09-23T08:58:51Z |
publishDate | 2021 |
publisher | Institute of Electrical and Electronics Engineers (IEEE) |
record_format | dspace |
spelling | mit-1721.1/1370422022-09-30T12:35:30Z Optimality of the Plug-in Estimator for Differential Entropy Estimation under Gaussian Convolutions Goldfeld, Ziv Greenewald, Kristjan Weed, Jonathan Polyanskiy, Yury Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science © 2019 IEEE. This paper establishes the optimality of the plugin estimator for the problem of differential entropy estimation under Gaussian convolutions. Specifically, we consider the estimation of the differential entropy h(X + Z), where X and Z are independent d-dimensional random variables with Z{\sim}\mathcal{N}( {0,{σ ^2}{{\text{I}}-d}} ). The distribution of X is unknown and belongs to some nonparametric class, but n independently and identically distributed samples from it are available. We first show that despite the regularizing effect of noise, any good estimator (within an additive gap) for this problem must have an exponential in d sample complexity. We then analyze the absolute-error risk of the plug-in estimator and show that it converges as frac{{{c^d}}}{{n }}, thus attaining the parametric estimation rate. This implies the optimality of the plug-in estimator for the considered problem. We provide numerical results comparing the performance of the plug-in estimator to general-purpose (unstructured) differential entropy estimators (based on kernel density estimation (KDE) or k nearest neighbors (kNN) techniques) applied to samples of X + Z. These results reveal a significant empirical superiority of the plug-in to state-of-the-art KDE- and kNN-based methods. 2021-11-01T18:45:11Z 2021-11-01T18:45:11Z 2019-09 2021-04-15T16:05:35Z Article http://purl.org/eprint/type/ConferencePaper https://hdl.handle.net/1721.1/137042 Goldfeld, Ziv, Greenewald, Kristjan, Weed, Jonathan and Polyanskiy, Yury. 2019. "Optimality of the Plug-in Estimator for Differential Entropy Estimation under Gaussian Convolutions." IEEE International Symposium on Information Theory - Proceedings, 2019-July. en http://dx.doi.org/10.1109/ISIT.2019.8849414 IEEE International Symposium on Information Theory - Proceedings Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf Institute of Electrical and Electronics Engineers (IEEE) MIT web domain |
spellingShingle | Goldfeld, Ziv Greenewald, Kristjan Weed, Jonathan Polyanskiy, Yury Optimality of the Plug-in Estimator for Differential Entropy Estimation under Gaussian Convolutions |
title | Optimality of the Plug-in Estimator for Differential Entropy Estimation under Gaussian Convolutions |
title_full | Optimality of the Plug-in Estimator for Differential Entropy Estimation under Gaussian Convolutions |
title_fullStr | Optimality of the Plug-in Estimator for Differential Entropy Estimation under Gaussian Convolutions |
title_full_unstemmed | Optimality of the Plug-in Estimator for Differential Entropy Estimation under Gaussian Convolutions |
title_short | Optimality of the Plug-in Estimator for Differential Entropy Estimation under Gaussian Convolutions |
title_sort | optimality of the plug in estimator for differential entropy estimation under gaussian convolutions |
url | https://hdl.handle.net/1721.1/137042 |
work_keys_str_mv | AT goldfeldziv optimalityofthepluginestimatorfordifferentialentropyestimationundergaussianconvolutions AT greenewaldkristjan optimalityofthepluginestimatorfordifferentialentropyestimationundergaussianconvolutions AT weedjonathan optimalityofthepluginestimatorfordifferentialentropyestimationundergaussianconvolutions AT polyanskiyyury optimalityofthepluginestimatorfordifferentialentropyestimationundergaussianconvolutions |