Module embedding 

Graph embedding via simplified Node2Vec random walks.

Learns low-dimensional vector representations of graph nodes by:

1. Performing random walks from each node
2. Training Skip-gram embeddings with SGD
3. Finding structurally similar node pairs via cosine similarity

Functions

compute_embeddings

Compute node embeddings using random walks + Skip-gram.

find_similar

Find top-N most similar node pairs by cosine similarity of embeddings.