Why doesn`t min mean cancelling work for submodular flow
We consider how to generalize the Minimum Mean Cycle Canceling algorithm of Goldberg and Tarjan, and its dual counterpart, the Maximum Mean Cut Canceling algorithm of Ervolina and McCormick, to general linear programs. Our aim is to investigate the conditions necessary to get a polynomial bound on the number of iterations of these algorithms. We find that the two vital parameters are the maximum number of nonzero components in any {open_quotes}mean improving direction{close_quotes}, and the maximum size that any component of such a direction can be in a minimal integer scaling. We then apply our generic algorithms to the Submodular Flow problem and discover that in both the primal and dual cases that one of these two parameters is not polynomially bounded.
- OSTI ID:
- 36274
- Report Number(s):
- CONF-9408161--
- Country of Publication:
- United States
- Language:
- English
Similar Records
Submodular optimization problems and greedy strategies: A survey
A capacity scaling algorithm for convex cost submodular flows