Please use this identifier to cite or link to this item:
https://scholarbank.nus.edu.sg/handle/10635/212687
Title: | DYNAMIC NEURAL ARCHITECTURES FOR IMPROVED INFERENCE | Authors: | CAI SHAOFENG | ORCID iD: | orcid.org/0000-0001-8605-076X | Keywords: | Dynamic Neural Architecture, Neural Networks, Inference Effectiveness and Efficiency | Issue Date: | 29-Jul-2021 | Citation: | CAI SHAOFENG (2021-07-29). DYNAMIC NEURAL ARCHITECTURES FOR IMPROVED INFERENCE. ScholarBank@NUS Repository. | Abstract: | Deep neural networks have achieved super-human prediction performance for many data types. However, besides effectiveness, the deployment of DNNs in real-world applications has also highlighted the need for efficiency and interoperability. In this thesis, we propose dynamic neural architectures, which customize their architectures or adapt model weights conditioned on the input at runtime to improve model inference. We will formulate dynamic neural architectures in a unified framework and then, identify the challenges and limitations of existing architectures and approaches in terms of model inference. Next, we will propose general and novel dynamic architecture techniques, in particular, dynamic routing, model slicing and adaptive relation modeling, for improved inference. With these techniques, DNNs will be able to support more efficient, effective and interpretable model inference. | URI: | https://scholarbank.nus.edu.sg/handle/10635/212687 |
Appears in Collections: | Ph.D Theses (Open) |
Show full item record
Files in This Item:
File | Description | Size | Format | Access Settings | Version | |
---|---|---|---|---|---|---|
CaiSF.pdf | 14.37 MB | Adobe PDF | OPEN | None | View/Download |
Google ScholarTM
Check
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.