Lecture 7. Graph partitioning algorithms.

Network Science 2021 @ HSE

[pedromoya]

the insights from this lecture are a pearl! thank you

[hasuchObe]

I was able to understand everything except slide 9. When you expand (s_i - s_j)^2 in the first line of math, i get the term on the second line plus an additional term that looks something like \frac{1}{8}\sum_{i, j} A_{ij}(1-\delta_{ij})(s_i^2+ s_j)^2. Not sure how to make this term vanish. Any ideas? It winds up being a constant function of the Laplacian matrix so it doesnt affect the lagrange multiplier procedure but it does affect the actual number of cuts, tho only by a constant and not a function of s.