Out of Distribution Detection, Generalization, and Robustness Triangle with Maximum Probability Theorem

Amir Emad Marvasti, Ehsan Emad Marvasti, Ulas Bagci

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Maximum Probability Framework, powered by Maximum Probability Theorem, is a recent theoretical development in artificial intelligence, aiming to formally define probabilistic models, guiding development of objective functions, and regularization of probabilistic models. MPT uses the probability distribution that the models assume on random variables to provide an upper bound on the probability of the model. We apply MPT to challenging out-of-distribution (OOD) detection problems in computer vision by incorporating MPT as a regularization scheme in the training of CNNs and their energy-based variants. We demonstrate the effectiveness of the proposed method on 1080 trained models, with varying hyperparameters, and conclude that the MPT-based regularization strategy stabilizes and improves the generalization and robustness of base models in addition to enhanced OOD performance on CIFAR10, CIFAR100, and MNIST datasets.

Original languageEnglish (US)
Title of host publicationInternational Conference on Electrical, Computer, Communications and Mechatronics Engineering, ICECCME 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665470957
DOIs
StatePublished - 2022
Event2022 International Conference on Electrical, Computer, Communications and Mechatronics Engineering, ICECCME 2022 - Male, Maldives
Duration: Nov 16 2022Nov 18 2022

Publication series

NameInternational Conference on Electrical, Computer, Communications and Mechatronics Engineering, ICECCME 2022

Conference

Conference2022 International Conference on Electrical, Computer, Communications and Mechatronics Engineering, ICECCME 2022
Country/TerritoryMaldives
CityMale
Period11/16/2211/18/22

Keywords

  • Out of distribution detection
  • deep learning
  • maximum probability theorem
  • regularization
  • robustness

ASJC Scopus subject areas

  • Automotive Engineering
  • Electrical and Electronic Engineering
  • Mechanical Engineering
  • Artificial Intelligence
  • Computer Networks and Communications
  • Computer Science Applications
  • Hardware and Architecture
  • Renewable Energy, Sustainability and the Environment

Fingerprint

Dive into the research topics of 'Out of Distribution Detection, Generalization, and Robustness Triangle with Maximum Probability Theorem'. Together they form a unique fingerprint.

Cite this