Minding the gap between Fast Heuristics and their Optimal Counterparts
- Pooria Namyar ,
- Behnaz Arzani ,
- Ryan Beckett ,
- Santiago Segarra ,
- Srikanth Kandula
Hot Topics in Networking |
Organized by acm
Production systems use heuristics because they are faster or scale better than the corresponding optimal algorithms. Yet, practitioners are often unaware of how worse off a heuristic’s solution may be with respect to the optimum in realistic scenarios. Leveraging two-stage games and convex optimization, we present a provable framework that unveils settings where a given heuristic underperforms.