Accelerated Graph Integration with Neumann Approximation
- 주제(키워드) Graph-based Machine Learning , Graph Integration
- 주제(DDC) 006.31
- 발행기관 아주대학교 일반대학원
- 지도교수 신현정
- 발행년도 2024
- 학위수여년월 2024. 2
- 학위명 석사
- 학과 및 전공 일반대학원 인공지능학과
- 실제URI http://www.dcollection.net/handler/ajou/000000033653
- 본문언어 한국어
- 저작권 아주대학교 논문은 저작권에 의해 보호받습니다.
초록/요약
When a data source contains relationship information that can be irregular and complex, representing it as a graph consisting of nodes and edges is an inevitable choice. With the rapid growth in data availability, it becomes more important to utilize multiple data sources containing different but complementary information for a given task. Using multiple graphs can technically be interpreted as finding the optimal combination of each graph. There have been various approaches for graph integration or graph fusion, but most of them have suffered from scalability issues due to long computation times. To overcome this difficulty, we propose a novel graph integration method that can be performed quickly and simply even for large-sized networks by using the approximation technique of Neumann expansion in the process of maximum likelihood estimation. As a result of various experiments conducted on several datasets, the node classification performance of the proposed method was competitive compared to existing methods, and the computation speed was extremely faster. In particular, the proposed method showed very fast speed even when the number of nodes in the network increased, which proves that the proposed method has very high scalability.
more목차
1. Introduction 1
2. Fundamentals 5
2.1 Graph-based Semi-supervised Learning 6
2.2 Optimization of Parameter 7
3. Proposed Method: Accelerated Graph Integration 8
3.1 Combining Coefficient Estimation 9
3.2 Remarks on Complexity 13
4. Related Works 14
5. Experiments 18
5.1 Datasets 19
5.2 Experimental Setting 22
5.3 Single vs. Integrated, Exact vs. Approximate 22
5.4 Comparison with Existing Methods 24
5.5 Combining Coefficient Estimation 27
5.6 Empirical Time Complexity 28
6. Conclusion 30
References 32