Chen, Yurong
(2020)
Graph-embedding Enhanced Attention Adversarial Autoencoder.
Master's Thesis, University of Pittsburgh.
(Unpublished)
Abstract
When dealing with the graph data in real problems, only part of the nodes in the graph are labeled and the rest are not. A core problem is how to use this information to extend the labeling so that all nodes are assigned a label (or labels). Intuitively we can learn the patterns (or extract some representations) from those labeled nodes and then apply the patterns to determine the membership for those unknown nodes. A majority of previous related studies focus on extracting the local information representations and may suffer from lack of additional constraints which are necessary for improving the robustness of representation. In this work, we presented Graph- embedding enhanced attention Adversarial Autoencoder Networks (Great AAN), a new scalable generalized framework for graph-structured data representation learning and node classification. In our framework, we firstly introduce the attention layers and provide insights on the self-attention mechanism with multi-heads. Moreover, the shortest path length between nodes is incorporated into the self-attention mechanism to enhance the embedding of the node’s structural spatial information. Then a generative adversarial autoencoder is proposed to encode both global and local information and enhance the robustness of the embedded data distribution. Due to the scalability of our approach, it has efficient and various applications, including node classification, a recommendation system, and graph link prediction. We applied this Great AAN on multiple datasets (including PPI, Cora, Citeseer, Pubmed and Alipay) from social science and biomedical science. The experimental results demonstrated that our new framework significantly outperforms several popular methods.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
27 September 2020 |
| Date Type: |
Publication |
| Defense Date: |
31 March 2020 |
| Approval Date: |
27 September 2020 |
| Submission Date: |
1 April 2020 |
| Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
| Number of Pages: |
37 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
Swanson School of Engineering > Electrical and Computer Engineering |
| Degree: |
MS - Master of Science |
| Thesis Type: |
Master's Thesis |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
Graph Embedding, Adversarial Autoencoder, Shortest Path Length Attention |
| Date Deposited: |
27 Sep 2020 22:16 |
| Last Modified: |
27 Sep 2020 22:16 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/38525 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}