Commit a098f6fb authored by Megha Khosla's avatar Megha Khosla

Update readme.md

parent 30e9a6c0
The repository contains refernce implementation of NERD proposed in
**Node Representation Learning for Directed Graphs.**
M. Khosla, J. Leonhardt, W. Nejdl, A. Anand.
In ECML-PKDD 2019.
================================================================================
>***Basic Usage***
**Required**: gsl library for random number generation
**To Compile and Link** : g++ -lgsl -lgslcblas -lm -pthread *.cpp -o NERD
**For Large Graphs** : Increase the size of hash table in NERD.h (look for **hash_table_size**) accordingly.
> **For Large Graphs** : Increase the size of hash table in NERD.h (look for **hash_table_size**) accordingly.
The algorithm expects a **directed weighted edgelist**. If your graph is unweighted, please add 1 as weight for all edges. For undirected graphs, please add 2 edges ( in both directions) corresponding to each edge.
For bipartite undirected graphs, one can treat edges directed from the left set to the right set and doubling of edges is not required.
......@@ -45,9 +51,9 @@ Parameters for training:
-inputvertex : if set 0 Use the first vertex in the walk sample as the input word otherwise use the middle vertex in the sample walk (default is 0)
**Evaluation**: The embeddings are evaluated for three tasks : Node Classification, Link Prediction and Graph Reconstruction. Respective evaluation scripts are evaluation/ multilabel_class_cv.py,
>**Evaluation**: The embeddings are evaluated for three tasks : Node Classification, Link Prediction and Graph Reconstruction. Respective evaluation scripts are evaluation/ multilabel_class_cv.py,
evaluation/graph_reconstruction_sigmoid.py , evaluation/link_pred.py
**Data**: Links to data and train-test psplits can be found under data/
> **Data**: Links to data and train-test psplits can be found under data/
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment