first commit

README.md

@@ -0,0 +1,46 @@
+# MVE
+This is an implementation of the MVE model for multi-view network embedding [(An Attention-based Collaboration Framework for Multi-View Network Representation Learning)](https://arxiv.org/abs/1709.06636). 
+
+## Install
+Our codes rely on two external packages, which are the Eigen package and the GSL package.
+
+#### Eigen
+The [Eigen](http://eigen.tuxfamily.org/index.php?title=Main_Page) package is used for matrix operations. To run our codes, users need to download the Eigen package and modify the package path in the makefile.
+
+#### GSL
+The [GSL](https://www.gnu.org/software/gsl/) package is used to generate random numbers. After installing the package, users also need to modify the package path in the makefile. 
+
+## Compile
+After installing the two packages and modifying the package paths, users may go to the "MVE" folder and use the makefile to compile the codes.
+
+## Data
+The MVE model receives a multi-view network and a set of labeled nodes as input. 
+
+Each view of the multi-view network is described by a single file. The files of different views should have the same prefix, and the indices should start from 0 to K-1, where K is the number of views. For example, users may describe a multi-view network with three files, "view_0", "view_1", "view_2", where the prefix is "view_" and the number of views is 3. Each view file contains several lines, with each line representing an edge in that view. The format of each line is "<u> <v> <w>", meaning that there is an edge from node <u> to node <v> and the weight is <w>.
+
+The labeled nodes are listed in another file. This file contains several lines, where each line gives the labels of a node. The format of a line is "<node name> <label name 1> <label name 2>...", which starts with the name of the node, followed by the names of different labels the node has.
+
+A toy dataset is provided in the "data/toy/" folder, and we will upload the dataset used in the paper later.
+
+## Running
+To run the MVE model, users may directly use the example script (run.sh) we provide. 
+
+## Contact: 
+If you have any questions about the codes and data, please feel free to contact us.
+```
+Meng Qu, qumn123@gmail.com
+```
+
+## Citation
+```
+@inproceedings{qu2018curriculum,
+title={Curriculum Learning for Heterogeneous Star Network Embedding via Deep Reinforcement Learning},
+author={Qu, Meng and Tang, Jian and Han, Jiawei},
+booktitle={Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining},
+pages={468--476},
+year={2018},
+organization={ACM}
+}
+}
+```
+

data/toy/label.txt

@@ -0,0 +1,2 @@
+a	label_0	label_1
+b	label_2

data/toy/view_0

@@ -0,0 +1,4 @@
+a	b	3
+a	c	5
+a	d	4
+b	c	2

data/toy/view_1

@@ -0,0 +1,4 @@
+b	c	3
+b	d	5
+b	a	4
+c	a	2

data/toy/view_2

@@ -0,0 +1,4 @@
+d	a	3
+d	b	5
+d	c	4
+d	d	2

mve/makefile

@@ -0,0 +1,8 @@
+CC = g++
+CFLAGS = -lm -pthread -Ofast -march=native -Wall -funroll-loops -Wno-unused-result -lgsl -lm -lgslcblas
+INCLUDES = -I<gsl path> -I<eigen path>
+LIBS = -L<gsl path>
+
+mve : model.cpp
+	$(CC) $(CFLAGS) model.cpp -o mve $(INCLUDES) $(LIBS) -lgsl -lm -lgslcblas
+