Improved Algorithms for Contextual Dynamic Pricing

In contextual dynamic pricing, a seller sequentially prices goods based on contextual information. Buyers will purchase products only if the prices are below their valuations. The goal of the seller is to design a pricing strategy that collects as much revenue as possible. We focus on two different valuation models. The first assumes that valuations linearly depend on the context and are further distorted by noise. Under minor regularity assumptions, our algorithm achieves an optimal regret bound of $\tilde{\mathcal{O}}(T^{2/3})$, improving the existing results. The second model removes the linearity assumption, requiring only that the expected buyer valuation is $\beta$-H\"older in the context. For this model, our algorithm obtains a regret $\tilde{\mathcal{O}}(T^{d+2\beta/d+3\beta})$, where $d$ is the dimension of the context space.

Paper

References (33)

Scroll for more · 21 remaining

Similar papers

Peer review

Reviewer iWzp6/10 · confidence 3/52024-06-20

Summary

This paper expands the line of work on contextual dynamic pricing where the buyer's valuation may be noisy. The authors give a simple way to obtain an unbiased estimate for the true valuation function, and propose an exploration - exploitation type algorithm that learn simultaneously the true valuation function and the noise distribution. Based on this observation, algorithms are proposed for when the valuation function is a linear model or when it is Holder continuous.

Strengths

The paper is well-written and I found it easy to follow. The necessary intuitions are adequately given. The main idea behind the algorithm is conceptually simple. The results in this work expands the line of work on contextual dynamic pricing, though I did not check the proof, I believe the results are new and correct.

Weaknesses

I think for me, the most limiting assumption is that the noise is i.i.d. across all contexts, which the authors also mentioned. It may not be possible to add experiments now, but I think it will be helpful for this paper.

Questions

In assumption 2, it is assumed the noise is bounded. Do all prior work assume that the noise is bounded? Is there any work that does not? This is minor, but I am not sure what the green / red color signify in table 1. Also you made the assumption that the noise is bounded, but this did not seem reflected in the rows where you presented your results. The table also has margin issues.

Rating

6

Confidence

3

Soundness

3

Presentation

3

Contribution

3

Limitations

Yes.

Reviewer B94q7/10 · confidence 3/52024-07-08

Summary

The paper studies the contextual dynamic pricing problem with binary demands and an unknown noise distribution. It presents a general framework for a pricing algorithm and proves its regret bounds for two specific types of valuation functions: linear and non-parametric. In the linear case, its results improve existing results and match the lower bound.

Strengths

The paper is well-written and introduces a novel framework for solving and analyzing the contextual dynamic pricing problem with unknown noise distribution. This new framework can be applied in various settings and leads to algorithms with improved regret upper bounds in the linear case.

Weaknesses

One potential concern is the paper's similarity with [1] in the exploration part and its contribution in this part. Line 186 mentions, “this method (using random pricing to construct unbiased estimates of $g(x_t)$) appears to have never been used in dynamic pricing” but as mentioned in line 195, if I understand correctly, [1] (where $g(x_t)$ is a linear function) also uses random pricing to construct such unbiased estimates. Although this paper’s main algorithm and analyses are different from [1]’s, a detailed discussion on the difference in using random prices as an exploration policy would be helpful to better understand its contribution. Another minor weakness is the lack of experimental results. Numerical experiments could help validate the efficiency of the algorithms and demonstrate their potential advantages over existing methods. [1]Fan, Jianqing, Yongyi Guo, and Mengxin Yu. "Policy optimization using semiparametric models for dynamic pricing." Journal of the American Statistical Association119.545 (2024): 552-564.

Questions

Please check point (1) in weaknesses. Additionally, compared to [1], why do the proposed algorithms achieve better regret bounds with milder assumptions (e.g., Assumption 1 does not require i.i.d. contexts as in [1])? Is this improvement due to using bandit techniques (e.g., price/action eliminations) or other reasons? More explanation would be appreciated.

Rating

7

Confidence

3

Soundness

3

Presentation

3

Contribution

3

Limitations

See weaknesses.

Reviewer EQLG6/10 · confidence 3/52024-07-10

Summary

The paper addresses the problem of dynamic pricing using contextual information, aiming to maximize a seller's revenue by setting prices based on the context and buyer valuations. Buyers purchase products if the prices are lower than their valuations. The authors propose an algorithm called VALUATION APPROXIMATION - PRICE ELIMINATION (VAPE) and analyze its performance under two valuation models: the linear valuation model and the non-parametric valuation model.

Strengths

- The paper is well-written and easy to follow. - The work is well-motivated from the practical perspective. - The theoretical results seem to be correct, even if I did not check the proofs carefully. - The related works are properly discussed.

Weaknesses

- A paper with such practical motivation would benefit greatly from being accompanied by a thorough experimental campaign. - See questions below.

Questions

- In the definition of regret (between lines 96 and 97), it seems that the optimal price (i.e., the one in the max) is unique for every product. I expect that every context presents an optimal price. Do the authors agree? - Lines 277-278: can the authors provide some more comments on the results and compare the assumptions of the other works to the one in this paper?

Rating

6

Confidence

3

Soundness

3

Presentation

3

Contribution

3

Limitations

No limitations.

Reviewer beqc8/10 · confidence 4/52024-07-12

Summary

This paper studies the problem of online contextual pricing under a linear noisy valuation model, with the noise distribution **unknown** to the seller. This work proposes an "exploration-then-elimination" algorithm that achieves $O(T^{2/3})$ **optimal** regret under the assumptions of (1) stochastic feature sequence, (2) Lipschitz noise CDF, and (3) bounded feature norm and noise range. Besides, this work also studies the non-parametric $\beta$-Holder valuation model, and propose an algorithm that achieves $O(T^{d+2\beta/d+3\beta})$ regret.

Strengths

1, The problem of contextual pricing problem has been extensively studied over the past few years. However, the minimax regret of linear noisy valuation model with unknown noise distribution is a long-existing open problem. Existing works leave a gap between $O(T^{3/4})$ and $\Omega(T^{2/3})$ under the assumption of Lipschitz noise CDF. This work closes the gap by showing that $O(T^{2/3})$ is the correct minimax regret, under certain assumptions. 2, This work also considers the non-parametric $\beta$-Holder valuation model, and propose an algorithm that achieves $O(T^{d+2\beta/d+3\beta})$ regret. This problem is relatively new to the community but it is important as well. In most of the cases, we cannot rely on the correctness of a linear model (although it is causally sound), and therefore this non-parametric model is of significance.

Weaknesses

1, The algorithm, VAPE, proposed in this work is quite similar to the "Explore-then-UCB" algorithm in Luo et al. (2022, Neurips). Seems like the authors are unaware of this work at all. The only difference is that VAPE adopts a policy-elimination design in the second stage after uniformly pure exploration, while Luo et al. (2022) uses a UCB-style strategy. From the perspective of regret analysis, I do not think there are much difference between those two methods. I then went back to the previous work and looked into their regret analysis, and noticed that they were using a linear bandit model instead of a multi-armed bandit (which is also applicable), which brings an extra $\sqrt{\frac1{\Delta}}$ multiplier on the regret of the second stage, where $\Delta$ is the discretization grid size. In other words, by slightly modifying the "Explore-then-UCB" algorithm, it is able to achieve $O(T^{2/3})$ regret instead of $O(T^{3/4})$ as they claimed. Since both of your works share the same idea of uniformly-pure-exploration stage for an unbiased estimate of $x^\top\theta^*$, which actually originates from Fan et al. (2024, JASA), I am skeptical to the novelty of this work, especially from the perspective of methodology. The authors are encouraged to discuss how their methods and analysis are different from Luo et al. (2022) in the rebuttal. This is the major concern that I (and other reviewers as I expect) would have. 2, The authors claim that it is required to assume Lipschitzness on the noise CDF, which is commonly applied except for Xu and Wang (2022, Aistats). It is worth noting that Xu and Wang (2022) have proposed an insight on the "Half-Lipschitz" nature of a pricing problem, indicating that the regret rate would not be different with or without Lipschitz assumption. If this is not applicable, the authors are encouraged to explain in the rebuttal how the "Half-Lipschitz" nature does not work. References: Xu, Jianyu, and Yu-Xiang Wang. "Towards agnostic feature-based dynamic pricing: Linear policies vs linear valuation with unknown noise." International Conference on Artificial Intelligence and Statistics. PMLR, 2022. Luo, Yiyun, Will Wei Sun, and Yufeng Liu. "Contextual dynamic pricing with unknown noise: Explore-then-ucb strategy and improved regrets." Advances in Neural Information Processing Systems 35 (2022): 37445-37457. Fan, Jianqing, Yongyi Guo, and Mengxin Yu. "Policy optimization using semiparametric models for dynamic pricing." Journal of the American Statistical Association 119.545 (2024): 552-564.

Questions

The authors are encouraged to respond to the following questions at a secondary priority (i.e. after kindly answering my questions in "Weaknesses"). 1, Regarding Weakness 1, is there any necessity to apply a policy eliminiation strategy in stage 2 of VAPE? Can it be replaced by a (version of) UCB without impairing the regret rate? 2, On Page 5 Line 196, Please kindly explain the difference between your work and Fan et al. (2024) for valuation estimation methods in details.

Rating

8

Confidence

4

Soundness

4

Presentation

4

Contribution

3

Limitations

The authors have discussed the limitations in their Conclusions. Overall, this is a theoretic work and there are not much societal impact involved. However, the authors are encouraged to discuss more on the societal aspect of the work, e.g. personalized pricing and unfairnesses.

Reviewer EQLG2024-08-09

Thank you for the clarifications

Thank you for the clarifications. I have no further questions. I increased my score to 6.

Reviewer B94q2024-08-11

Thank you for your response. It addresses my concerns and I have raised the score.

Reviewer beqc2024-08-12

Thanks for your response

The reviewer thanks the authors for your insightful and exhaustive reply. For the comparison with [1], I somewhat agree that it is suboptimal to have a linear bandit algorithm, while yours are optimal. However, as I asked in Q1, by replacing your policy-elimination stage with a UCB, the algorithm is on the one hand optimal enough and on the other hand closer to their algorithmic design. However, since you are anyways the first to achieve optimal, and given that [1] had been accepted for Neurips 2022, I would like to support the acceptance of your paper. For the difference of assumptions comparing to [2], I am still skeptical if their half-lipschitzness is not applicable. In other words, I am not sure whether you really need $F(\delta) \approx F(\delta \pm \epsilon)$ which is stronger than it appears for a $T^{2/3}$ regret bound. For the estimates of $F$ and its difference with [3], I tend to believe that there are no substantial difference between (poly) kernels and discretizations. But I am happy to leave them for discussions beyond this paper.

Program Chairsdecision2024-09-25

Decision

Accept (poster)

© 2026 NYSGPT2525 LLC