Contextual Search via Intrinsic Volumes

Abstract

We study the problem of contextual search, a multidimensional generalization of binary search that captures many problems in contextual decision-making. In contextual search, a learner is trying to learn the value of a hidden vector. Every round the learner is provided an adversarially-chosen context vector, submits a guess for the inner product of the context vector with the hidden vector, learns whether their guess was too high or too low, and incurs some loss based on the quality of their guess. The learner's goal is to minimize their total loss over the course of some number of rounds. We present an algorithm for the contextual search problem for the symmetric loss function that achieves constant total loss. We present a new algorithm for the dynamic pricing problem (which can be realized as a special case of the contextual search problem) that achieves doubly-logarithmic total loss, improving exponentially on the previous best known upper bounds and matching the known lower bounds (up to a polynomial dependence on dimension). Both algorithms make significant use of ideas from the field of integral geometry, most notably the notion of intrinsic volumes of a convex set. To the best of our knowledge this is the first application of intrinsic volumes to algorithm design.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 26 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.