Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Electronic and Information Engineering | en_US |
| dc.contributor.advisor | Mao, Yuyi (EIE) | en_US |
| dc.creator | Chen, Jizhou | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/12047 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | A pruning method for multi-exit neural networks based on the genetic algorithm | en_US |
| dcterms.abstract | Deep neural network is a type of atificial neural network that is made up of many hidden layers. It is widely applied in audio, image, video analytics. Especially with the development of the Internet of Things and edge artificial intelligence technologies, deep neural networks are also expected to be deployed in a great variety of intelligent terminal devices. The terminal devices are typically with low storage capacity and poor computing capability, which makes it very difficult to host deep neural networks. | en_US |
| dcterms.abstract | To solve this problem, model compression technologies and multi-exit neural networks are proposed. Specifically, model compression technologies remove the network parameters to save calculations to a certain extent, introducing the smallest possible impact on the network accuracy. Multi-exit neural network refers to inserting early exits into a backbone network structure, so that the processing of some easy input instances can be completed in advance. Both technologies can effectively reduce the computation overhead, and thereby help to realize fast inference. | en_US |
| dcterms.abstract | However, when inferring, the multi-exit network needs to be executed exit-by-exit. Similarly, for the front exit pruning, the accuracy and computational complexity of all subsequent exits will also be reduced. As a result, we cannot measure the impact of pruning a channel on the whole network. This dissertation proposes a fast pruning method based on genetic algorithm to make the multi-exit network prune more emphatically. | en_US |
| dcterms.extent | 39 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2022 | en_US |
| dcterms.educationalLevel | M.Sc. | en_US |
| dcterms.educationalLevel | All Master | en_US |
| dcterms.LCSH | Neural networks (Computer science) | en_US |
| dcterms.LCSH | Genetic algorithms | en_US |
| dcterms.LCSH | Hong Kong Polytechnic University -- Dissertations | en_US |
| dcterms.accessRights | restricted access | en_US |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| 6511.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 1.52 MB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/12047