Adversarial Attack for Uncertainty Estimation: Identifying Critical Regions in Neural Networks

Authors: Alarab, I. and Prakoonwit, S.

Journal: Neural Processing Letters

Volume: 54

Issue: 3

Pages: 1805-1821

eISSN: 1573-773X

ISSN: 1370-4621

DOI: 10.1007/s11063-021-10707-3

Abstract:

We propose a novel method to capture data points near decision boundary in neural network that are often referred to a specific type of uncertainty. In our approach, we sought to perform uncertainty estimation based on the idea of adversarial attack method. In this paper, uncertainty estimates are derived from the input perturbations, unlike previous studies that provide perturbations on the model's parameters as in Bayesian approach. We are able to produce uncertainty with couple of perturbations on the inputs. Interestingly, we apply the proposed method to datasets derived from blockchain. We compare the performance of model uncertainty with the most recent uncertainty methods. We show that the proposed method has revealed a significant outperformance over other methods and provided less risk to capture model uncertainty in machine learning.

https://eprints.bournemouth.ac.uk/36455/

Source: Scopus

Adversarial Attack for Uncertainty Estimation: Identifying Critical Regions in Neural Networks

Authors: Alarab, I. and Prakoonwit, S.

Journal: NEURAL PROCESSING LETTERS

Volume: 54

Issue: 3

Pages: 1805-1821

eISSN: 1573-773X

ISSN: 1370-4621

DOI: 10.1007/s11063-021-10707-3

https://eprints.bournemouth.ac.uk/36455/

Source: Web of Science (Lite)

Adversarial Attack for Uncertainty Estimation: Identifying Critical Regions in Neural Networks

Authors: Alarab, I. and Prakoonwit, S.

Journal: Neural Processing Letters

eISSN: 1573-773X

ISSN: 1370-4621

DOI: 10.1007/s11063-021-10707-3

Abstract:

We propose a novel method to capture data points near decision boundary in neural network that are often referred to a specific type of uncertainty. In our approach, we sought to perform uncertainty estimation based on the idea of adversarial attack method. In this paper, uncertainty estimates are derived from the input perturbations, unlike previous studies that provide perturbations on the model's parameters as in Bayesian approach. We are able to produce uncertainty with couple of perturbations on the inputs. Interestingly, we apply the proposed method to datasets derived from blockchain. We compare the performance of model uncertainty with the most recent uncertainty methods. We show that the proposed method has revealed a significant outperformance over other methods and provided less risk to capture model uncertainty in machine learning.

https://eprints.bournemouth.ac.uk/36455/

Source: Manual

Preferred by: Simant Prakoonwit

Adversarial Attack for Uncertainty Estimation: Identifying Critical Regions in Neural Networks

Authors: Alarab, I. and Prakoonwit, S.

Abstract:

We propose a novel method to capture data points near decision boundary in neural network that are often referred to a specific type of uncertainty. In our approach, we sought to perform uncertainty estimation based on the idea of adversarial attack method. In this paper, uncertainty estimates are derived from the input perturbations, unlike previous studies that provide perturbations on the model's parameters as in Bayesian approach. We are able to produce uncertainty with couple of perturbations on the inputs. Interestingly, we apply the proposed method to datasets derived from blockchain. We compare the performance of model uncertainty with the most recent uncertainty methods.

We show that the proposed method has revealed a significant outperformance over other methods and provided less risk to capture model uncertainty in machine learning.

https://eprints.bournemouth.ac.uk/36455/

http://dx.doi.org/10.1007/s11063-021-10707-3

Source: arXiv

Adversarial Attack for Uncertainty Estimation: Identifying Critical Regions in Neural Networks

Authors: Alarab, I. and Prakoonwit, S.

Journal: Neural Processing Letters

Volume: 54

Pages: 1805-1821

ISSN: 1370-4621

Abstract:

We propose a novel method to capture data points near decision boundary in neural network that are often referred to a specific type of uncertainty. In our approach, we sought to perform uncertainty estimation based on the idea of adversarial attack method. In this paper, uncertainty estimates are derived from the input perturbations, unlike previous studies that provide perturbations on the model’s parameters as in Bayesian approach. We are able to produce uncertainty with couple of perturbations on the inputs. Interestingly, we apply the proposed method to datasets derived from blockchain. We compare the performance of model uncertainty with the most recent uncertainty methods. We show that the proposed method has revealed a significant outperformance over other methods and provided less risk to capture model uncertainty in machine learning

https://eprints.bournemouth.ac.uk/36455/

Source: BURO EPrints