Active embedding search via noisy paired comparisons
This work addresses the challenge of costly and noisy queries in preference estimation for applications like recommender systems and psychology experiments, offering incremental improvements over existing methods.
The paper tackles the problem of actively selecting paired comparisons to estimate a user's preference vector from noisy responses, proposing two novel strategies that maximize lower bounds on information gain. The strategies achieve similar performance to greedy information maximization and outperform state-of-the-art selection methods and random queries in preference estimation.
Suppose that we wish to estimate a user's preference vector $w$ from paired comparisons of the form "does user $w$ prefer item $p$ or item $q$?," where both the user and items are embedded in a low-dimensional Euclidean space with distances that reflect user and item similarities. Such observations arise in numerous settings, including psychometrics and psychology experiments, search tasks, advertising, and recommender systems. In such tasks, queries can be extremely costly and subject to varying levels of response noise; thus, we aim to actively choose pairs that are most informative given the results of previous comparisons. We provide new theoretical insights into the benefits and challenges of greedy information maximization in this setting, and develop two novel strategies that maximize lower bounds on information gain and are simpler to analyze and compute respectively. We use simulated responses from a real-world dataset to validate our strategies through their similar performance to greedy information maximization, and their superior preference estimation over state-of-the-art selection methods as well as random queries.