An example of implementing USPORF on anomaly detection

[ShumingXu]

Description
Detecting the anomalies in a time series of graphs is an important and interesting task. In this example, I plan to compare the performance of two Euclidean embedding algorithms which are multiple adjacency spectral embedding (MASE) and omnibus embedding (OMNI), with random forest embedding and USPORF. The simulation will be evolving time series of graphs generated with RDPG (random dot product graph) from the paper Anomaly Detection in Time Series of Graphs.
Plan

• Read and summarize the paper and recreate the RDPG simulation (currently here)
• Figure out how USPORF can be applied to the simulation
• Run USPORF directly on the generated data and see how it performs
• Complete the distance metrics issue The similarity metrics used in USPORF may need modification #7 of SPORF (introducing depth of nearest common ancestor and length of shortest path)
• Update the distance metrics in USPORF and rerun the simulations to see if new metrics boosts the performance of USPORF.