Improved Algorithms for Contextual Dynamic Pricing

Authors: Matilde Tullii, Solenne Gaucher, Nadav Merlis, Vianney Perchet

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we illustrate some numerical simulations that aim to show the empirical performance of our VAPE algorithm for Linear Valuations.
Researcher Affiliation Collaboration Matilde Tullii Fair Play Team, CREST, ENSAE Solenne Gaucher Fair Play Team, CREST, ENSAE Nadav Merlis Fair Play Team, CREST, ENSAE Vianney Perchet Fair Play Team, CREST, ENSAE Criteo AI Lab
Pseudocode Yes Algorithm 1 VALUATION APPROXIMATION PRICE ELIMINATION (VAPE): General scheme
Open Source Code Yes The code implemented for these simulations is publicly available in the repository: https://github.com/Matilde Tulii1/ Improved-Algorithms-for-Contextual-Dynamic-Pricing
Open Datasets No In order to test our algorithm, we built a dataset of 5 contexts belonging to R3 generated by a canonical gaussian distribution and subsequently normalized.
Dataset Splits No The paper discusses simulation runs over specified time horizons (e.g., T [1000, 10000, 50000, 200000, 500000, 800000]) and uses a 'stochastic' and 'adversarial' case for comparison with previous work, but it does not specify explicit train/validation/test dataset splits or cross-validation methodology.
Hardware Specification Yes All the simulation can be (and were) run on a laptop without gpus.
Software Dependencies No The paper provides a GitHub repository link for the code, but it does not explicitly list specific software dependencies with their version numbers within the paper itself.
Experiment Setup Yes The algorithm has been tested on time horizons T [1000, 10000, 50000, 200000, 500000, 800000], and the hyperparameters α, µ, ϵ are set as in the statement of Theorem 1.