Driver for Multiple runs:

A semisupervised learning model for heterogeneous information networks (HINs)

ujson == 1.35 pytorch >= 1.5.0

(DBLP dataset only) Download Glove word embdding: http://nlp.stanford.edu/data/glove.840B.300d.zip
Generate node embeddings by node2vec or metapath2vec: a. preprocess/yelp.py -> gen_homograph() preprocess/main.py --input ../../data/yelp/homograph.txt --output ../../data/ yelp/RUBK_128.emb --dimensions 128 --workers 56 --walk-length 100 --num-walks 40 --window-size 5 b. preprocess/yelp.py -> gen_walk(path='../data/yelp/', walk_length=100,n_walks=40) preprocess/metapath2vec -train ../../data/yelp/BRKRB.walk -output ../../data/yelp/BRKRB_128.emb -size 128 -threads 40
Fuse edge features: preprocess/yelp.py -> dump_yelp_edge_emb(path='../data/yelp/')
Compute index: preprocess/yelp.py -> gen_edge_adj_random(path='../data/yelp/',edge_dim=130)

run_dblp.py; run_freebase.py; run_yelp.py

Entry for training program:

Conch: train_reg3.py Conch_nc: train_ncreg.py Conch_rd: train_rdreg.py

Conch: train.py Conch_nc: train_nc.py Conch_rd: train_rd.py

MIT

Name		Name	Last commit message	Last commit date
Latest commit History 8 Commits
data		data
model		model
preprocess		preprocess
.gitignore		.gitignore
LG_sage.py		LG_sage.py
LICENSE		LICENSE
README.md		README.md
bipartite.py		bipartite.py
bipartite_random.py		bipartite_random.py
cling.py		cling.py
gpu_mem_track.py		gpu_mem_track.py
helpers.py		helpers.py
line_graph_models.py		line_graph_models.py
logistics.py		logistics.py
lr.py		lr.py
models.py		models.py
nn_modules.py		nn_modules.py
parse_plot.py		parse_plot.py
plot_all.py		plot_all.py
problem.py		problem.py
reg.py		reg.py
reg2.py		reg2.py
run_aminer.py		run_aminer.py
run_dblp.py		run_dblp.py
run_freebase.py		run_freebase.py
run_yelp.py		run_yelp.py
train.py		train.py
train_nc.py		train_nc.py
train_ncreg.py		train_ncreg.py
train_rd.py		train_rd.py
train_rdreg.py		train_rdreg.py
train_reg3.py		train_reg3.py
train_reg_gs.py		train_reg_gs.py