Matching of Matching-Graphs - A Novel Approach for Graph Classification
- Resource Type
- Conference
- Authors
- Fuchs, Mathias; Riesen, Kaspar
- Source
- 2020 25th International Conference on Pattern Recognition (ICPR) Pattern Recognition (ICPR), 2020 25th International Conference on. :6570-6576 Jan, 2021
- Subject
- Computing and Processing
Signal Processing and Analysis
Data acquisition
Data structures
Encoding
Pattern matching
Graph Matching
Matching-Graphs
Graph Edit Distance
- Language
Due to fast developments in data acquisition, we observe rapidly increasing amounts of data available in diverse areas. Simultaneously, we observe that in many applications the underlying data is inherently complex, making graphs a very useful and adequate data structure for formal representation. A large amount of graph based methods for pattern recognition have been proposed. Many of these methods actually rely on graph matching. In the present paper a novel encoding of graph matching information is proposed. The idea of this encoding is to formalize the stable cores of specific classes by means of graphs. In an empirical evaluation we show that it can be highly beneficial to focus on these stable parts of graphs during graph classification.