Last semester, I developed an algorithm to compute the maximum likelihood low dimensional representation of an ER SBM. Optimizing the objective of this algorithm is very slow, and thus, it is intractable for large problems. This semester, I would like to improve on this algorithm, and in particular:
Prove that blocks only merge when ER SBMs are reduced (I am almost sure this is true)
Use the above fact to decrease the complexity of the algorithm from O(2^n) to O(n log (n) )
Run the algorithm on the HBN data to cluster patients by reduction tree distance
(REACH) Run the algorithm on Coyote/Sea Lion dataset to investigate properties of many/one nomination
Last semester, I developed an algorithm to compute the maximum likelihood low dimensional representation of an ER SBM. Optimizing the objective of this algorithm is very slow, and thus, it is intractable for large problems. This semester, I would like to improve on this algorithm, and in particular: