Graphs and recommendation Mathematics
[Graph theory] Graph diffusion and Personalized PageRank (Rewritten from an earlier blog post)
Explain graph diffusion as a weighted sum of random walks of different lengths. Derive Personalized PageRank (PPR) from walks with restarts and compare a three-vertex hand calculation with a sparse Python implementation. Examine restart probabilities and starting distributions, then compare diffusion weighted by geometric and Poisson distributions.
