Post-Hoc Selection of Pareto-Optimal Solutions in Search and Recommendation

By V. Paparella et al

Published on June 21, 2023

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Multi-Objective Optimization
3. Background
3.1 Selection Strategies
3.1.1 Knee Point
3.1.2 Hypervolume
3.1.3 Other Techniques
3.2 Related Work on MOO for IR and RS
4. Population Distance from Utopia
5. Experimental Evaluation
6. Conclusion
References

Summary

The document discusses post-hoc selection of Pareto-optimal solutions in Information Retrieval (IR) and Recommender Systems (RS). It introduces a novel technique named Population Distance from Utopia (PDU) for selecting the best Pareto-optimal solution. PDU analyzes the distribution of points on the Pareto frontier and proposes a method to identify the one-best solution. The paper highlights the importance of selecting a single optimal solution from the Pareto frontier in IR and RS tasks. It compares PDU against existing strategies and shows experimental results supporting its effectiveness. The paper also provides a GitHub repository for implementation and reproducibility of results.