Neural Dynamic Pricing: Provable and Practical Efficiency

Hyungkyu Kang, Seunghoon Kim, Young-Geun Choi, Min-hwan Oh

Keywords: dynamic pricing, neural networks

Abstract:

Despite theoretical guarantees of existing dynamic pricing (DP) methods, their strong model assumptions may not reflect real-world conditions and are often unverifiable. This poses major challenges in practice since the performance of an algorithm may significantly degrade if the assumptions are not satisfied. Moreover, many DP algorithms show unfavorable empirical performance due to the lack of data efficiency.
To address these challenges, we design a practical contextual DP algorithm that utilizes regression oracles. Our proposed algorithm assumes only Lipschitz continuity on the true conditional probability of purchase.
We prove $\tilde{\mathcal{O}}(T^{\frac{2}{3}}\text{regret}_R(T)^{\frac{1}{3}})$ regret upper bound where $T$ is the horizon and $\text{regret}_R(T)$ is the regret of the oracle. The bound is nearly minimax optimal in the canonical case of finite function class, and our analysis generically applies to other function approximators including neural networks. To the best of our knowledge, our work is the first algorithm to utilize the powerful generalization capability of neural networks with provable guarantees in dynamic pricing literature.
Extensive numerical experiments show that our algorithm outperforms existing state-of-the-art dynamic pricing algorithms in various settings, which demonstrates both provable efficiency and practicality.

Primary Area: other topics in machine learning (i.e., none of the above)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 10031