Currently, the new improved version is under the peer-review process...

The aim of this work is to propose a dynamic network embedding method for better global topology preserving of a dynamic network at each time step. Unlike all previous works that mainly consider the most affected regions of a network, the idea of this work, motivated by divide and conquer, is to partition a network into smaller sub-networks such that we can diversely consider the topological changes over a network.
The motivation of this work is that most real-world networks have some inactivate regions/sub-networks which would receive accumulated topological changes propagated via the high-order proximity. See the figure below for illustration. Therefore, in order to better preserve the global topology, we also need to consider the accumulated topological changes in the inactivate regions/sub-networks. However, previous works did not consider this issue.

Fig. a) A change (new edge in red) affects all nodes in the connected network via high-order proximity. The proximity of nodes 1-6 becomes $1^{st}$ order from $5^{th}$ order, nodes 2-6 becomes $2^{nd}$ order from $4^{th}$ order, etc. Besides, the proximity of any node in sub-network 1 to any node in sub-network 2 is reduced by $5^{th}$ order. b-d) The real-world dynamic networks have some inactive sub-networks (e.g., defined as no change occurs lasting for at least 5 time steps). The x-axis indicates the number of consecutive time steps that no change occurs in a sub-network. The y-axis gives the counts of each case in x-axis. The sub-networks, in average 50 nodes per sub-network, are obtained by applying METIS algorithm [Karypis and Kumar 1998] on the largest snapshot of a dynamic network. The details of the three dynamic networks are described in Section 5.

More details to be CONT'D...

If you find this work is useful, please use the following citation.

@article{hou2019dynwalks,
    title={DynWalks: Global Topology and Recent Changes Awareness Dynamic Network Embedding},
    author={Chengbin Hou and Han Zhang and Ke Tang and Shan He},
    journal={arXiv preprint arXiv:1907.11968},
    year={2019}
}

conda create -n GloDyNE python=3.6.8
source activate GloDyNE
cd GloDyNE
pip install -r requirements.txt

You may also need to install METIS package from the source.
Note, Python 3.6.6 or above is required due to the new print(f' ') feature.

cd GloDyNE
python src/main.py --method DynWalks --task gr --graph data/AS733/AS733_new.pkl --label data/AS733/AS733_label.pkl --emb-file output/AS733_DynWalks.pkl --num-walks 10 --walk-length 80 --window 10 --limit 0.1 --scheme 4 --seed 2019 --emb-dim 128 --workers 32

cd GloDyNE
python src/main.py --method DynWalks --task save --graph data/AS733/AS733_new.pkl --label data/AS733/AS733_label.pkl --emb-file output/AS733_DynWalks.pkl --num-walks 10 --walk-length 80 --window 10 --limit 0.1 --scheme 4 --seed 2019 --emb-dim 128 --workers 32
python src/eval.py --task all --graph data/AS733/AS733_new.pkl --emb-file output/AS733_DynWalks.pkl --label data/AS733/AS733_label.pkl --seed 2019

cd GloDyNE
bash bash/ALL_small.sh

Please see the README.md under the data folder.
If you would like to use your own dataset, the input dynamic network can be prepared as follows:

1. create an empty dynamic network as an empty python list, called DynG;
   
2. use Networkx to build the graph based on the edge steams from time step t-k to t, called Gt;
   
3. append the current snapshot Gt to the dynamic network DynG;
   
4. repeat 2) and 3) as time goes on...
   
5. finally, use Pickle to store DynG as a .pkl file

We are happy to answer any questions about the code and paper.