Benefits of Permutation-Equivariance in Auction Mechanisms
Authors: Tian Qin, Fengxiang He, Dingfeng Shi, Wenbing Huang, Dacheng Tao
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments fully support our theory. |
| Researcher Affiliation | Collaboration | 1University of Science and Technology of China 2JD Explore Academy, JD.com Inc. 3Gaoling School of Artificial Intelligence, Renmin University of China 4Beijing Key Laboratory of Big Data Management and Analysis Methods |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | We are associated in a industrial organization. The code is currently under review of our organization. |
| Open Datasets | No | We employ a multivariate uniform distribution U[0, 1]m to model the bidder valuation profiles. In all settings, we sample 640,000 data points for training and 5,000 points for test. |
| Dataset Splits | No | In all settings, we sample 640,000 data points for training and 5,000 points for test. |
| Hardware Specification | Yes | The experiments are conducted on 1 GPU (NVIDIA Tesla V100 16GB) and 10 CPU cores (Intel Xeon Processor E5-2650 v4 @ 2.20GHz). |
| Software Dependencies | No | We optimize the auction mechanism model via solving the following optimization problem, following the standard settings [12, 29, 10, 17]... Lρ(ω, λ) is minimized via Adam with a learning rate of 0.001... |
| Experiment Setup | Yes | The objective function Lρ(ω, λ) is minimized via Adam with a learning rate of 0.001 with respect to the model parameter ω and the Lagrange multiplier λ is updated once in every 100 iterations, until the ex-post regret is smaller than 0.001. The regularization factor ρ is set to 1.0 initially and gradually increased along the training process. In calculating the best bid profile v i of every bidder i, we first randomly initialize the bid profiles once in training and 1,000 times in test, optimize each of them individually via Adam with the same settings, and take the best one as the approximated best bidding. |