asherniu / cs-249-project

CS-249-Project

Download the adjacency matrix and feature vector from here https://drive.google.com/drive/folders/1ZdROyFY3KTh7ARqleMKwfYFL8BoWBsLD?usp=sharing or run the generation codes in model 2 or 3 (about 10-15 min to generate)

Model 1 Naive Bayes:

You can find the codes in the NaiveBayes_Jupyter.ipynb. It will require author_feature.pickle and author_features.pickle. Both can be generated, downloaded and came with the repo. You can also run NaiveBayes_Python.py if you prefer python codes.

Model 2 Pairwise Conditional Random Field:

The codes are in the ModelPairwiseCRF directory. Implemented with PyStruct

Prerequisite

1. numpy
2. pandas
3. pystruct (note that pystruct is only supported on Python2, Python3.6 or less, we use Python2.7 to test the code)
4. The model requires af_py2.pickle which stores the author feature matrix with pickle protocol 2 (since we use Python2). We provide the pickle file in the directory, or you can re-generate the pickle file, run gen_author_feature_py2_pickle.py.

Train and Evaluate

python crf.py

Experiment Results

We have run the model with python crf.py > crf.log, you can either rerun the model or directly check our results in crf.log

Model3 Graph Convolutional Networks in PyTorch:

PyTorch implementation of Graph Convolutional Networks (GCNs) for semi-supervised classification [1].

For a high-level introduction to GCNs, see:

Thomas Kipf, Graph Convolutional Networks (2016)

Requirements

• PyTorch 0.4 or 0.5
• Python 2.7 or 3.6

Usage

python train.py

Dataset:

gcn/data/DBLP_four_area/ and gcn/data/four_area/ 
(note: we directly read dataset from author_matrix.pickle and author_feature.pickle, which were generated from the dataset through function "generate_adj_feature()" in gcn/pygcn/utils.py)
You can generate the pickle files or download them, and modify the path of reading function "load_data()" in gcn/pygcn/utils.py.

Model Design:

Data structure: authors as the nodes in the graph, Adjacency matrix between authors, features of authors, labels of authors
Generate our own dataset: "load_data()" in gcn/pygcn/utils.py.

References

[1] Kipf & Welling, Semi-Supervised Classification with Graph Convolutional Networks, 2016

[2] Sen et al., Collective Classification in Network Data, AI Magazine 2008