Monte Carlo Dropout Based BatchEnsemble For Improving Uncertainty Estimation

Jain, Shubham and Srijith, P K (2023) Monte Carlo Dropout Based BatchEnsemble For Improving Uncertainty Estimation. In: 6th ACM India Joint International Conference on Data Science and Management of Data, CODS-COMAD 2023, 4-7 January 2023, Mumbai.

Text
3570991.3571038.pdf - Published Version
Download (374kB)

Abstract

Modelling uncertainty in deep learning is important for several high-risk applications such as autonomous driving and healthcare. Existing techniques for uncertainty modelling in deep learning such as Monte Carlo (MC) Dropout and BatchEnsemble suffer from some drawbacks. MC dropout shares parameters across models resulting in highly correlated predictions while BatchEnsemble requires storing additional parameters for each model in the ensemble. In our work, we aim to bring the best of both worlds by combining MC-dropout in the process of ensemble creation in BatchEnsemble. The proposed approach, Monte-Carlo BatchEnsemble, helps in generating ensembles with less correlation in prediction with the addition of a few parameters. The experimental results show the effectiveness of the proposed technique for image classification.

[error in script]

IITH Creators:

IITH Creators	ORCiD
Srijith, P K	http://www.orcid.org/0000-0002-2820-0835

Item Type:

Conference or Workshop Item (Paper)

Uncontrolled Keywords:

Monte Carlo methods; Risk assessment; Uncertainty analysis;Autonomous driving; Highly-correlated; Images classification; Modeling uncertainties; Uncertainty estimation; Uncertainty models;Deep learning

Subjects:

Computer science
Computer science > Computer programming, programs, data

Divisions:

Department of Computer Science & Engineering

Depositing User:

Mr Nigam Prasad Bisoyi

Date Deposited:

16 Aug 2023 11:16

Last Modified:

16 Aug 2023 11:16

URI:

http://raiithold.iith.ac.in/id/eprint/11548