Incentive Auctions and Spectrum Repacking
summary
Reverse auction feasibility testing must answer whether removing one of ~3,000 TV stations still allows a global repacking under ~3 million interference constraints, about 100,000 checks per simulated auction. Mixed integer programming solves ~50% of instances in 60s, a reduction to SAT (mostly binary clauses) plus algorithm configuration (SMAC) yields a configured clasp portfolio that solves ~85% then a 4‑way portfolio plus superset cache reaches ~99% in 0.2s and 99.6% in 60s on held out instances.
AI-written summary · about
chapters (20)
- 0:00Intro
- 0:52Incentive Auctions
- 3:38How the Auction Works
- 6:57Feasibility Testing
- 10:47Constraints
- 12:54Reusable
- 16:49How comfortable are the policymakers
- 17:59How much time do you have
- 18:48Conditional information
- 19:46Data
- 26:57Satisfiability histogram
- 28:08Algorithm configuration
- 31:53Clasp H2
- 34:38Local Search Solvers
- 34:58Decomposition
- 36:58Algorithms
- 39:28Simulation
- 40:20Caching
- 43:19Cash
- 43:31Cash vs Portfolio
related talks
56mMarket Design and Computer-Assisted Markets: An Economist's Perspective
25mBudget Pacing in Repeated Auctions: Regret and Efficiency without Convergence
10mSynchronized Priority Auctions
32mMaking DeFi Truly Cross-Chain
29mClearing Matching Markets Efficiently: Informative Signals and Match Recommendations
50mDiversity in Matching Markets
36mThe Intractability of Hylland-Zeckhauser and its Aftermath
29mKeynote Playing games in the clouds