Document Type

Presentation

Publication Date

11-2023

Publication Title

SC23 Posters

Pages

1-2

Publisher Name

ACM/IEEE

Publisher Location

USA

Abstract

Deep Learning (DL) methods have shown substantial efficacy in computer vision (CV) and natural language processing (NLP). Despite their proficiency, the inconsistency in input data distributions can compromise prediction reliability. This study mitigates this issue by introducing uncertainty evaluations in DL models, thereby enhancing dependability through a distribution of predictions. Our focus lies on the Vision Transformer (ViT), a DL model that harmonizes both local and global behavior. We conduct extensive experiments on the ImageNet-1K dataset, a vast resource with over a million images across 1,000 categories. ViTs, while competitive, are vulnerable to adversarial attacks, making uncertainty estimation crucial for robust predictions.

Our research advances the field by integrating uncertainty evaluations into ViTs, comparing two significant uncertainty estimation methodologies, and expediting uncertainty computations on high-performance computing (HPC) architectures, such as the Cerebras CS-2, SambaNova DataScale, and the Polaris supercomputer, utilizing the MPI4PY package for efficient distributed training.

Comments

Author Posting © 2023 Association for Computing Machinery. This poster is posted here with permission from the ACM for personal use, not for redistribution. This post was presented at SC ’23, November 12-17, 2023, Denver, CO. https://sc23.supercomputing.org/proceedings/tech_poster/tech_poster_pages/rpost141.html

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License.

Download

Share

COinS