Preference-based optimization from noisy pairwise comparisons

Siyi Wang, Zifan Wang, Karl Henrik Johanssson

In interactive systems, feedback is often provided in the form of preference between queried options rather than precise scores, which motivates optimization methods to learn from…