Node Representation for downstream tasks with Node2Vec

Total time needed: ~2 hours
Learn about the Node2Vec method to generate node embedding for downstream tasks.
Potential Use Cases
Perfect for when you want to use Graph Neural Network for Node Classification, Community Detection or Link Prediction, but have no node features (just edges).
Who is This For ?
INTERMEDIATEIntermediate ML developers with some knowledge of NLP and Graph Theory
Click on each of the following annotated items to see details.
ARTICLE 1. (optional) A word2vec introduction if you don't know it
  • What is word2vec?
  • How does it work and what it can do?
30 minutes
VIDEO 2. A refresher of the CBOW and Skipgram algorithm
  • How does the Word2Vec algorithm work?
20 minutes
VIDEO 3. Explanation of Random Walk
  • What is random walk?
  • How does random walk work programmatically?
10 minutes
VIDEO 4. Deep Walk brings NLP techniques to Graph Theory
  • What is Deep Walk?
  • How does it utilize the ideas of NLP's word2vec?
  • What representation learning for graph it can be used for?
15 minutes
VIDEO 5. Paper Walkthrough and Explanation of Node2Vec
  • How biased random walk is used for sampling of node sequence?
  • How node sequences are used to generate embeddings for nodes?
  • How Node2Vec works?
15 minutes
PAPER 6. Using hyper-parameter p & q to influence walks
  • How Node2Vec parameters are used to influence embedding results?
  • What are the use cases proposed by the author?
15 minutes
ARTICLE 7. A walkthrough of how to implement Node2Vec for feature learning and downstream task
  • How to implement Node2Vec on a downstream task?
  • How to evaluate Node2Vec compared to a simple Deep Walk?
25 minutes

Concepts Covered

0 comment