What is it about?
Sometimes, we have to use approximate solutions (heuristics) instead of optimal (optimal may be too slow). Approximate solutions may be too far from the optimal in some cases. Xplain proposes a framework to get insights into understanding heuristic underperformance.
Featured Image
Photo by charlesdeluvio on Unsplash
Why is it important?
Approximate solutions (heuristics) are deployed in significant productions, such as large-scale data centers like Amazon and Microsoft Azure. Failures come at huge financial costs and waste of electric power.
Perspectives
Researchers and operators have been developing heuristics for years. The tools used to analyze the heuristics usually provide a few examples that show gaps in heuristic performance. We are interested in broadening our understanding of heuristics, and we propose a framework that is a step towards bringing explainability to heuristic analysis. We're interested in getting inside into these questions: 1. For which inputs the heuristic underperforms compared to the optimal? 2. Why does the heuristic underperform? 3. What can we do about it? How can we fix it? Xplain takes a few steps in answering each of these questions.
Pantea Karimi Babaahmadi
Massachusetts Institute of Technology
Read the Original
This page is a summary of: Towards Safer Heuristics With XPlain, November 2024, ACM (Association for Computing Machinery),
DOI: 10.1145/3696348.3696884.
You can read the full text:
Contributors
The following have contributed to this page
