Skip to content

A fast combinatoric reduction algorithm for jet hypos

Peter Sherwood requested to merge peter/athena:gnfn3 into master

An exploration has been made of the use of a maxflow algorithm to produce a hypo decision for complex jet hypos. Maxflow can be used to find the optimum matching in a bipartite graph. When applied to jet hypos, this approach works well for simple jet- hypo condition matching. However, when multiple jets are are used to satisfy a condition, as occurs in more complex hypos (for example, any condition on diets requires the consideration of two jets) the flow in the network no longer corresponds to a jet-Condition matching. The maxflow algorithm directs flow through the compound jet objects to the conditions, but without necessarily saturating the multi jet. Maxflow is achieved, but matching is not.

In this MR, the code that was used to generate the flow network has been reused to provide a fast reduction of the number of combinatorial possibilities that need be considered. Although it was hoped that maxflow would solve the final jet- last condition (where the last conditions are the final conditions in the hypo), fast reduction is competitive due to the rapid decrease in combinations that are considered as the algorithm proceeds.

@khoo @jbossios @sschramm

Merge request reports

Loading