MATLAB demo codes for fast von Neumann graph entropy computation method published at ICML 2019
Paper link: https://arxiv.org/abs/1805.11769
Authors: Pin-Yu Chen, Lingfei Wu, Sijia Liu, Indika Rajapakse
Environment: tested on MATLAB R2016b
- Run main_FINGER_demo.m
- The demo code first generates a Erdos-Renyi graph with n nodes and n*p average degree. Its connectivity pattern is characterized by the adjacency matrix A
- GraEntExact(A) computes the exact VNGE of A and reports the computation time
- VNGE_FINGER(A) uses FINGER (FAST) to compute the approximate VNGE of A and report the computation time
- The demo code also shows the scaled appromimation error (SAE) and computation time reduction ratio (CTRR)
Computing exact VNGE takes 37.6566 seconds
Computing VNGE using FINGER (FAST) takes 0.35195 seconds
The computation time reduction ratio is 99.0654%
The scaled appromimation error is 0.039722