Research map / Learning theory and algorithms / Submodular and game algorithms
Learning to Clear the Market
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use the clearing loss instead of the benchmark pricing losses | Linear pricing model · Google Ad Exchange test set | revenue at a fixed match rate | Helped |
| Use the clearing loss instead of the benchmark pricing losses | Linear pricing model · Google Ad Exchange test set | revenue at a fixed social welfare: over 10% revenue improvement with less than 2% impact on buyer welfare | Helped |
| Set the clearing-loss supply parameter to λ = 0.25 (ablation) | Clearing loss · Google Ad Exchange test set | revenue improvement: 20% | Helped |
| Set the clearing-loss supply parameter to λ = 0.25 (ablation) | Clearing loss · Google Ad Exchange test set | match rate: 30% | Hurt |
| Vary λ to control supply and price aggressiveness (ablation) | Clearing loss · Google Ad Exchange test set | match rate | Helped |
| Use the default clearing-loss parameter λ = 1 (ablation) | Clearing loss · Google Ad Exchange test set | match rate: 1 - 1/e ≈ 0.63 | Mixed |
| Use the stated match-rate formula to choose λ (ablation) | Clearing loss · Google Ad Exchange test set | match rate: λ = log(1/(1-{sf MR})) | Helped |
| Fit prices by least-squares regression on the highest bid rather than clearing loss | Linear pricing model · Google Ad Exchange test set | social welfare: almost 5% | Hurt |
| Use revenue surrogate loss instead of clearing loss | Linear pricing model · Google Ad Exchange test set | convergence speed: 400K iterations | Hurt |
| Vary γ across {0.25, 0.5, 0.75, 1} and select γ = 0.75 (ablation) | Revenue surrogate · Google Ad Exchange test set | revenue | Helped |
| Use least-squares regression on the highest bid without regularization | Linear pricing model · Google Ad Exchange test set | revenue | Helped |
| Use least-squares regression on the highest bid without regularization | Linear pricing model · Google Ad Exchange test set | match rate | Hurt |
Closest papers by meaning
- Revenue Optimization with Approximate Bid Predictions (NeurIPS 2017 · similarity 0.89)
- Real-Time Optimisation for Online Learning in Auctions (ICML 2020 · similarity 0.88)
- Improved Online Learning Algorithms for CTR Prediction in Ad Auctions (ICML 2023 · similarity 0.87)
- Learning Auctions with Robust Incentive Guarantees (NeurIPS 2019 · similarity 0.87)
- Leveraging the Hints: Adaptive Bidding in Repeated First-Price Auctions (NeurIPS 2022 · similarity 0.87)
- Increasing Revenue in Efficient Combinatorial Auctions by Learning to Generate Artificial Competition (AAAI 2025 · similarity 0.87)
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.