Saved in:
Bibliographic Details
Main Authors: Hor, Soheil, Qian, Ying, Pilanci, Mert, Arbabian, Amin
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2402.04359
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929236013481984
author Hor, Soheil
Qian, Ying
Pilanci, Mert
Arbabian, Amin
author_facet Hor, Soheil
Qian, Ying
Pilanci, Mert
Arbabian, Amin
contents This paper introduces the first theoretical framework for quantifying the efficiency and performance gain opportunity size of adaptive inference algorithms. We provide new approximate and exact bounds for the achievable efficiency and performance gains, supported by empirical evidence demonstrating the potential for 10-100x efficiency improvements in both Computer Vision and Natural Language Processing tasks without incurring any performance penalties. Additionally, we offer insights on improving achievable efficiency gains through the optimal selection and design of adaptive inference state spaces.
format Preprint
id arxiv_https___arxiv_org_abs_2402_04359
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Adaptive Inference: Theoretical Limits and Unexplored Opportunities
Hor, Soheil
Qian, Ying
Pilanci, Mert
Arbabian, Amin
Machine Learning
This paper introduces the first theoretical framework for quantifying the efficiency and performance gain opportunity size of adaptive inference algorithms. We provide new approximate and exact bounds for the achievable efficiency and performance gains, supported by empirical evidence demonstrating the potential for 10-100x efficiency improvements in both Computer Vision and Natural Language Processing tasks without incurring any performance penalties. Additionally, we offer insights on improving achievable efficiency gains through the optimal selection and design of adaptive inference state spaces.
title Adaptive Inference: Theoretical Limits and Unexplored Opportunities
topic Machine Learning
url https://arxiv.org/abs/2402.04359