AI Paper Atlas

Research map / Bandits and online learning / Online convex optimization and games

Optimal-er Auctions through Attention

Dmitry Ivanov, Iskander Safiulin, Igor Filippov, Ksenia Balabaeva

NeurIPS 2022 · Bandits and online learning › Online convex optimization and games · approach: Learning for pricing and auctions

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Replace RegretNet's fully connected architecture with attention-based RegretFormerauction mechanism learning · 2x2, 2x3, 2x5, and 3x10; regret budget 10^-3revenue: 0.889 -> 0.908; 1.317 -> 1.416; 2.339 -> 2.453; 5.59 -> 6.121Improved
Replace RegretNet's fully connected architecture with attention-based RegretFormerauction mechanism learning · 1x2, regret budget 10^-3revenue: 0.572 -> 0.571Hurt
Replace RegretNet or EquivariantNet with RegretFormerauction mechanism learning · 1x2 through 3x10; regret budget 10^-4revenue: 0.551 -> 0.556; 0.825 -> 0.861; 1.249 -> 1.327; 2.121 -> 2.339; 5.02 -> 5.745Improved
Add positional encoding to RegretFormer (ablation)RegretFormer · asymmetric one bidder, two itemsrevenue: 9.967 -> 10.056Improved
Use RegretFormer rather than RegretNet in multi-setting trainingauction mechanism learning · multi-setting averagerevenue: 2.573 -> 2.703Improved
Use RegretFormer rather than RegretNet for out-of-multi-setting generalizationauction mechanism learning · out-of-multi-setting averageregret: 0.0662 -> 0.00478Improved
Tighten the regret budget from 10^-3 to 10^-4RegretFormer · 1x2revenue: 0.571 -> 0.556Hurt
Tighten the regret budget from 10^-3 to 10^-4RegretFormer · 2x2revenue: 0.908 -> 0.861Hurt
Tighten the regret budget from 10^-3 to 10^-4RegretFormer · 2x3revenue: 1.416 -> 1.327Hurt
Tighten the regret budget from 10^-3 to 10^-4RegretFormer · 2x5revenue: 2.453 -> 2.339Hurt
Tighten the regret budget from 10^-3 to 10^-4RegretFormer · 3x10revenue: 6.121 -> 5.745Hurt
Use EquivariantNet rather than RegretFormerauction mechanism learning · multi-settingregret: 0.00391 -> 0.00989Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy