Content

The graph mining course starts with describing fundamentals of graph theory to the users along with demonstrations using tiny examples (in python). Slowly it enters into algorithm designs based on various graph heuristics and network principles. The next level of the course explains how machine learning models interpret/represent graphs when provided as input in adjacency and what patterns/knowledge they can infer from them . The end part of the course focuses on explaining graph convolutional networks, a popular deep learning architecture used in multiple domains e.g. NLP, computer vision, biomedical and many more to solve prediction and classification tasks.

• Graph mining: preliminaries Slides Reference
  • What are graphs? Different types?
  • Similarity graphs
  • Eigenvalues & eigenvectors
  • Adjacency, diagonal, laplacian
• Cheatsheet on Networkx Slides Reference
• Link analysis: Degree, Betweenness, Eigenvector centrality Slides Reference
• Neuman kernels Slides Reference
• Pagerank, HITS, Spectral clustering Slides Reference
• Encoder-decoder
• Loss functions
• Performance metrics Reference

If you wish, we are happy to receive contributions from past and ongoing registered class students to make these notes better. Contact at vinti.agarwal@pilani.bits-pilani.ac.in