Reserve Price Optimization for First Price Auctions in Display Advertising

Authors: Zhe Feng, Sebastien Lahaie, Jon Schneider, Jinchao Ye

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our approach over synthetic data where bidder values are drawn uniformly, and also over real bid distributions collected from the logs of the major ad exchange. Our experimental results confirm that the combination of variance reduction on both objective components leads to the fastest convergence rate. For the demand component, a simple logistic model works well over the synthetic (i.e., uniform) data, but a flexible neural net is needed over the semi-synthetic data. For the bidding component, we find that quantile truncation is much more robust to assumptions on the bidding model.
Researcher Affiliation Collaboration 1Harvard University, this work was done when the first author was an intern in Google Inc, NYC. 2Google Inc, NYC.
Pseudocode Yes Algorithm 1 Zeroth-order stochastic projected gradient framework for reserve optimization.
Open Source Code No The paper does not contain any explicit statement about providing open-source code or a link to a code repository.
Open Datasets No The paper uses
Dataset Splits No The paper mentions collecting
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory, or cloud instance types) used for running the experiments.
Software Dependencies No The paper mentions using a
Experiment Setup Yes The parameters used in these algorithms are specified in Appendix B. (Appendix B content: Perturbation size βt = δ/rt; The learning rate αt = α0/ √ t; The number of samples nt = N/T; For demand curve training, the simple logistic model uses a learning rate of 0.01 and the neural network uses a learning rate of 0.001.)