Abstract
In-network (or on-path) inference over programmable data planes allows fast and low-overhead inference in deep neural networks. In this work, we propose an adaptive approach to strike the balance between accuracy and processing cost. To be specific, the confidence score is evaluated at the end of each layer, and an early exit is triggered if the confidence score is sufficiently high. We implement this early-exit scheme over BMv2 software switches and the results demonstrate that the proposed scheme successfully controls the trade-off by making use of the confidence score.
| Original language | English |
|---|---|
| Title of host publication | EuroP4 2023 - Proceedings of the 6th International Workshop on P4 in Europe |
| Publisher | Association for Computing Machinery, Inc |
| Pages | 61-64 |
| Number of pages | 4 |
| ISBN (Electronic) | 9798400704468 |
| DOIs | |
| Publication status | Published - 2023 Dec 8 |
| Event | 6th International Workshop on P4 in Europe, EuroP4 2023, co-located with ACM CoNEXT 2023 - Paris, France Duration: 2023 Dec 8 → … |
Publication series
| Name | EuroP4 2023 - Proceedings of the 6th International Workshop on P4 in Europe |
|---|
Conference
| Conference | 6th International Workshop on P4 in Europe, EuroP4 2023, co-located with ACM CoNEXT 2023 |
|---|---|
| Country/Territory | France |
| City | Paris |
| Period | 23/12/8 → … |
Bibliographical note
Publisher Copyright:© 2023 ACM.
Keywords
- early-exit
- in-network intelligence
- programmable data plane
ASJC Scopus subject areas
- Computer Networks and Communications
- Hardware and Architecture
- Electrical and Electronic Engineering