A Closer Look at Branch Classifiers of Multi-exit Architectures

04/28/2022
by   Shaohui Lin, et al.
15

Multi-exit architectures consist of a backbone and branch classifiers that offer shortened inference pathways to reduce the run-time of deep neural networks. In this paper, we analyze different branching patterns that vary in their allocation of computational complexity for the branch classifiers. Constant-complexity branching keeps all branches the same, while complexity-increasing and complexity-decreasing branching place more complex branches later or earlier in the backbone respectively. Through extensive experimentation on multiple backbones and datasets, we find that complexity-decreasing branches are more effective than constant-complexity or complexity-increasing branches, which achieve the best accuracy-cost trade-off. We investigate a cause by using knowledge consistency to probe the effect of adding branches onto a backbone. Our findings show that complexity-decreasing branching yields the least disruption to the feature abstraction hierarchy of the backbone, which explains the effectiveness of the branching patterns.

READ FULL TEXT

page 7

page 9

page 10

page 11

research
03/17/2022

TMS: A Temporal Multi-scale Backbone Design for Speaker Embedding

Speaker embedding is an important front-end module to explore discrimina...
research
10/03/2019

Regularizing Neural Networks via Stochastic Branch Layers

We introduce a novel stochastic regularization technique for deep neural...
research
08/04/2021

FPB: Feature Pyramid Branch for Person Re-Identification

High performance person Re-Identification (Re-ID) requires the model to ...
research
09/27/2017

Connectivity Learning in Multi-Branch Networks

While much of the work in the design of convolutional networks over the ...
research
07/13/2015

Unconstrained Facial Landmark Localization with Backbone-Branches Fully-Convolutional Networks

This paper investigates how to rapidly and accurately localize facial la...
research
10/15/2021

Receptive Field Broadening and Boosting for Salient Object Detection

Salient object detection requires a comprehensive and scalable receptive...
research
08/08/2017

Cherry-Picking of Code Commits in Long-Running, Multi-release Software

This paper presents Tartarian, a tool that supports maintenance of softw...

Please sign up or login with your details

Forgot password? Click here to reset