Winsorization for Robust Bayesian Neural Networks

With the advent of big data and the popularity of black-box deep learning methods, it is imperative to address the robustness of neural networks to noise and outliers. We propose the use of Winsorization to recover model performances when the data may have outliers and other aberrant observations. W...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Somya Sharma, Snigdhansu Chatterjee
Formato: article
Lenguaje:EN
Publicado: MDPI AG 2021
Materias:
Q
Acceso en línea:https://doaj.org/article/911e03cac136460fa0bfa8c32c4cbcc3
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
id oai:doaj.org-article:911e03cac136460fa0bfa8c32c4cbcc3
record_format dspace
spelling oai:doaj.org-article:911e03cac136460fa0bfa8c32c4cbcc32021-11-25T17:30:52ZWinsorization for Robust Bayesian Neural Networks10.3390/e231115461099-4300https://doaj.org/article/911e03cac136460fa0bfa8c32c4cbcc32021-11-01T00:00:00Zhttps://www.mdpi.com/1099-4300/23/11/1546https://doaj.org/toc/1099-4300With the advent of big data and the popularity of black-box deep learning methods, it is imperative to address the robustness of neural networks to noise and outliers. We propose the use of Winsorization to recover model performances when the data may have outliers and other aberrant observations. We provide a comparative analysis of several probabilistic artificial intelligence and machine learning techniques for supervised learning case studies. Broadly, Winsorization is a versatile technique for accounting for outliers in data. However, different probabilistic machine learning techniques have different levels of efficiency when used on outlier-prone data, with or without Winsorization. We notice that Gaussian processes are extremely vulnerable to outliers, while deep learning techniques in general are more robust.Somya SharmaSnigdhansu ChatterjeeMDPI AGarticleBayesian neural networkuncertainty quantificationvariational Gaussian processWinsorizationconcrete dropoutflipoutScienceQAstrophysicsQB460-466PhysicsQC1-999ENEntropy, Vol 23, Iss 1546, p 1546 (2021)
institution DOAJ
collection DOAJ
language EN
topic Bayesian neural network
uncertainty quantification
variational Gaussian process
Winsorization
concrete dropout
flipout
Science
Q
Astrophysics
QB460-466
Physics
QC1-999
spellingShingle Bayesian neural network
uncertainty quantification
variational Gaussian process
Winsorization
concrete dropout
flipout
Science
Q
Astrophysics
QB460-466
Physics
QC1-999
Somya Sharma
Snigdhansu Chatterjee
Winsorization for Robust Bayesian Neural Networks
description With the advent of big data and the popularity of black-box deep learning methods, it is imperative to address the robustness of neural networks to noise and outliers. We propose the use of Winsorization to recover model performances when the data may have outliers and other aberrant observations. We provide a comparative analysis of several probabilistic artificial intelligence and machine learning techniques for supervised learning case studies. Broadly, Winsorization is a versatile technique for accounting for outliers in data. However, different probabilistic machine learning techniques have different levels of efficiency when used on outlier-prone data, with or without Winsorization. We notice that Gaussian processes are extremely vulnerable to outliers, while deep learning techniques in general are more robust.
format article
author Somya Sharma
Snigdhansu Chatterjee
author_facet Somya Sharma
Snigdhansu Chatterjee
author_sort Somya Sharma
title Winsorization for Robust Bayesian Neural Networks
title_short Winsorization for Robust Bayesian Neural Networks
title_full Winsorization for Robust Bayesian Neural Networks
title_fullStr Winsorization for Robust Bayesian Neural Networks
title_full_unstemmed Winsorization for Robust Bayesian Neural Networks
title_sort winsorization for robust bayesian neural networks
publisher MDPI AG
publishDate 2021
url https://doaj.org/article/911e03cac136460fa0bfa8c32c4cbcc3
work_keys_str_mv AT somyasharma winsorizationforrobustbayesianneuralnetworks
AT snigdhansuchatterjee winsorizationforrobustbayesianneuralnetworks
_version_ 1718412238468939776