- Ubuntu 16.04.1 LTS (Xenial Xerus)
- g++ 7.5.0
- gcc 7.5.0
We provide the dataset threads-stack-overflow in the directory "./dataset/".
The other three datasets used in the paper can be downloaded from:
- bitcoin-temporal: https://www.cs.cornell.edu/~arb/data/temporal-bitcoin/
- indochina-2004: http://data.law.di.unimi.it/webdata/indochina-2004/indochina-2004.graph
- colisten-Spotify: https://www.cs.cornell.edu/~arb/data/colisten-Spotify/
Note that the dataset indochina-2004 downloaded from LAW website needs to be decompressed using the toolkit webgraph provided by the website(http://webgraph.di.unimi.it/).
After downloading the indochina-2004, you need first convert it into motif-based weighted graphs using the MAPPR method given in [Yin et al, KDD2017]. The code of MAPPR is published at http://snap.stanford.edu/mappr/ . Please put the motif-based weighted graph returned by MAPPR in the directory: CoinFlipWalk/dataset/ and rename them as indochina-2004.txt. In addition, we provide a small dataset grqc.txt in ./dataset/grqc_weighted directory for convenience.
We will convert the txt files into four files, .attribute, .outEdges, .outPtr, .outWEdges.
bash query.shNote that you may need to run powermethod first to generate groundtruths for performance evaluations (i.e. max-additive-error, precision@50, conductance). Because powermethod is a time-consuming algorithm, there may be a long time to wait in the beginning.
A command for a method is as follows.
./<algo> -d </path/to/dataset/directory> -f <filelabel> -e <epsilon sequence> -qn <the number of query> -l <random walk step> [-u]-u is optional. If -u is added, it will only run the updating part and exit. If not, it will only run the query.
For example, to run queries with RejectionWalk for threads-stack-overflow, you can run
./RejectionWalk -d ./dataset/threads-stack-overflow_weighted/ -f threads-stack-overflow -e 1e-1 1e-2 1e-3 1e-4 1e-5 -qn 10 -l 10
To get the performance evaluations, please run
bash metric.sh