Awesome

Heterogeneous-Deep-Graph-Infomax

Parameter Setting:

HDGI-A: <br>

Node-level dimension: 16<br> Attention head: 4<br> Semantic-level attention vector: 8<br> learning rate: 0.02<br>

HDGI-C: <br>

Node-level dimension: 64<br> Semantic-level attention vector: 8<br> learning rate: 0.02<br>

GAT:<br>

Node-level dimension: 16<br> Attention head: 4<br> learning rate: 0.005<br> Drop out ratio: 0.6<br>

GCN:<br>

Hidden-unit dimension: 64<br> learning rate: 0.01<br> Drop out ratio: 0.5<br>

RGCN:<br>

Hidden-unit dimension: 16<br> learning rate: 0.01<br> Drop out ratio: 0<br>

Metapath2Vec:<br>

Embedding dimension: 100<br> learning rate: 0.01<br> negative-samples: 5<br> Window: 1<br>

HAN:<br>

Node-level dimension: 16<br> Attention head: 4<br> Semantic-level dimension: 8<br> learning rate: 0.005<br> Node-level Drop out ratio: 0.6<br> Semantic-level Drop out ratio: 0.6<br>

Deepwalk:<br>

Number of walks per node: 10<br> Dimensions of word embeddings: 128<br> Length of random walk: 30<br> Window size for skipgram: 5<br>

DGI:<br>

Hidden-unit dimension: 64<br> learning rate: 0.001<br> Drop out ratio: 0<br>