Explore graph embedding techniques and prototype selection methods in this 54-minute guest presentation by Massimo Piccardi from the University of Central Florida. Delve into key concepts including graph matching, graph edit distance, and bipartite graph edit distance. Learn about prototype-based graph embedding and various discriminative prototype selection approaches, such as center, border, repelling, spanning, and targetsphere selections. Examine experimental results comparing discriminative and conventional methods across different datasets, and understand the impact of prototype numbers per class. Gain valuable insights into graph theory and its applications in machine learning and data analysis.
Overview
Syllabus
Intro
Definitions
Main properties
Graph matching
Graph edit distance - more formally
Minimum-cost edit path
Bipartite graph edit distance
Prototype-based graph embedding
Prototype set: example
Supervised prototype selection: example
Discriminative prototype selection
Discriminative center prototype selection
Discriminative border prototype selection
Discriminative repelling prototype selection
Discriminative spanning prototype selection
Discriminative targetsphere prototype selection
Targetsphere selection
Experiments
Datasets
Discriminative vs conventional (cnt'd)
Number of prototypes per class
Conclusion
Taught by
UCF CRCV
