Graph similarity algorithms based on NetworkX.
BSD Licensed
First, install building tool:
$ yum install -y scons
On Mac OS:
$ brew install scons
Then install graphsim via PyPI:
$ pip install -U graphsim
By default, sudo
is required to give permission to install cpp modules into system /usr/local/{lib,include}
.
If you prefer local installation, following instructions may help you:
export LIBTACSIM_LIB_DIR=~/usr/lib/
export LIBTACSIM_INC_DIR=~/usr/include/
pip install -U graphsim
Make sure that the local directories are aware for C linkers:
export LD_LIBRARY_PATH=~/usr/lib:$LD_LIBRARY_PATH
export C_INCLUDE_PATH=~/usr/include:$C_INCLUDE_PATH
export CPLUS_INCLUDE_PATH=~/usr/include:$CPLUS_INCLUDE_PATH
NOTE: libtacsim
was tested on Ubuntu 12.04, Ubuntu 16.04, CentOS 6.5 and Mac OS 10.11.2, 10.13.2.
>>> import graphsim as gs
gs.ascos
: Asymmetric network Structure COntext Similarity, by Hung-Hsuan Chen et al. [paper]gs.nsim_bvd04
: node-node similarity matrix, by Blondel et al. [paper]gs.hits
: the hub and authority scores for nodes, by Kleinberg. [paper]gs.nsim_hs03
: node-node similarity with mismatch penalty, by Heymans et al. [paper]gs.simrank
: A Measure of Structural-Context Similarity, by Jeh et al. [paper]gs.simrank_bipartite
: SimRank for bipartite graphs, by Jeh et al. [paper]gs.tacsim
: Topology-Attributes Coupling Similarity, by Chen et al. [paper]gs.tacsim_combined
: A combined topology-attributes coupling similarity, by Chen et al. [paper]gs.tacsim_in_C
: an efficient implementation of TACSim in pure C.gs.tacsim_combined_in_C
: an efficient implementation of combined TACSim in pure C.
gs.normalized
: L2 normalization of vectors, matrices or arrays.gs.node_edge_adjacency
: Obtain node-edge adjacency matrices in source and dest directions.
@article{Chen2017,
title = "Discovering and modeling meta-structures in human behavior from city-scale cellular data",
journal = "Pervasive and Mobile Computing ",
year = "2017",
issn = "1574-1192",
doi = "http://dx.doi.org/10.1016/j.pmcj.2017.02.001",
author = "Xiaming Chen and Haiyang Wang and Siwei Qiang and Yongkun Wang and Yaohui Jin"
}
Xiaming Chen [email protected]