Results and Findings Task 1a: Learning Complex Models


    We demonstrated both theoretically and empirically that our graph learning framework requires fewer calls to BestSubgraph than a naive greedy approach, O(N 3) as compared to O(N 4) for exact greedy search, and O(N log N) as compared to O(N 2) for approximate greedy search, resulting in 2 to 3 orders of magnitude speedup in practice. All five methods were… (More)


    1 Figure or Table