Node2Vec node features initialization.md

Node2Vec Features Initialization

Using Node2Vec model to preprocessing node features for non-feature's node types

Node2vec is a method for learning low-dimensional representations of nodes in a graph. It is designed to work with a single type of node, but it can be applied to different types of nodes by treating each node type as a separate graph, and learning embeddings for each one separately. In this way, multiple node types can be incorporated into a single graph representation, by concatenating the embeddings of each node type. This will allow to the model to learn a multi-typed node structure. It is worth mentioning that this works well if the multiple node types are connected to each other in a meaningful way. Otherwise, it would be better to treat them as separate graph.

Example graph:

	session	aid
0	0	0
1	0	1
2	0	2
3	1	1
4	1	0
5	2	0

session nodes

Create graph of connected session that interacted with the same aid

	session	pair_session
0	0	1
1	0	2
2	1	2
3	2	1

>>> edge_index
tensor([[0, 0, 1, 2],
        [1, 2, 2, 1]])
        
>>> edge_index.shape
torch.Size([2, 4]) #[2, num_edges]

aid nodes

	session	aid	prev_aid
0	0	0	0
1	0	1	0
2	0	2	1
3	1	1	1
4	1	0	1
5	2	0	0

>>> edge_index
tensor([[0, 1, 2, 1, 0, 0],
        [0, 0, 1, 1, 1, 0]])
        
>>> edge_index.shape
torch.Size([2, 6]) #[2, num_edges]

>>> data.num_nodes
3 #number of `aid` unique node [0, 1, 2]

model = Node2Vec(edge_index=data.edge_index,
                 embedding_dim=..., 
                 walk_length=...,
                 context_size=...,
                 walks_per_node=...,
                 num_negative_samples=..., 
                 p=0.2, q=0.4, sparse=True).to(device)