Fair and Welfare-Efficient Constrained Multi-matchings under Uncertainty

We study fair allocation of constrained resources, where a market designer optimizes overall welfare while maintaining group fairness. In many large-scale settings, utilities are not known in advance, but are instead observed after realizing the allocation. We therefore estimate agent utilities using machine learning. Optimizing over estimates requires trading-off between mean utilities and their predictive variances. We discuss these trade-offs under two paradigms for preference modeling -- in the stochastic optimization regime, the market designer has access to a probability distribution over utilities, and in the robust optimization regime they have access to an uncertainty set containing the true utilities with high probability. We discuss utilitarian and egalitarian welfare objectives, and we explore how to optimize for them under stochastic and robust paradigms. We demonstrate the efficacy of our approaches on three publicly available conference reviewer assignment datasets. The approaches presented enable scalable constrained resource allocation under uncertainty for many combinations of objectives and preference models.

Paper

References (59)

Scroll for more · 38 remaining

Similar papers

Peer review

Reviewer UQiJ7/10 · confidence 4/52024-06-30

Summary

The authors propose group-fair multi-matching algorithms in the presence of uncertainty about the valuations of agent-item pairs. The algorithms fall under two families: Conditional Value at Risk (CVaR), and robust optimization. The types of utilities they consider are also two-fold: normal utilitarian, and egalitarian group-fairness. The latter seeks to maximize the utility of the group with the minimum utility. The authors begin with the robust optimization approach for overall utility maximization, where they derive max-min guarantees for the worst possible valuation within a particular valuation uncertainty set. Here, they propose solving a natural nested concave-convex program. Due to computational feasibility, they would like to utilize the structure of the specific problem in order to achieve a more efficient problem framing. To do this, they demonstrate that the nested program (eq 1) can be re-written as a single concave cubic program. Furthermore, the optimal assignment can be found by solving a particular system of equations with the optimal dual variables (Proposition 3.1). They also show that under some simplified assumptions on the uncertainty set containing the true valuation matrix, the optimal solution can be found by simply solving a LP or QP (Prop. 3.2, 3.3). The authors then move to robust optimization for egalitarian welfare. Here, the problem is more complex since the objective function loses its smoothness, and is now a nested max min min optimization problem. They make a key assumption on the independence of the uncertainty set w.r.t. the data groups, which helps them swap the minimum over groups and minimum valuation over uncertainty sets. By another dual derivation, one can solve the resulting problem with iterated quadratic programming (Prop. 3.4). Furthermore, the problem has the nice property that under simplified uncertainty sets, it still reduces to solving a single LP (Prop. 3.5). The authors also note that the problem can be decomposed in the case of monotonic welfare functions. The next topic covered is w.r.t. the CVaR approach, which assumes that the distribution over agent-item valuations is known to the mechanism designer. We begin with the optimal utilitarian assignment. Here, the optimal solution can be computed, but may be computationally infeasible for arbitrary distribution over valuation matrices $D$. Therefore, the authors adopt and approach where they sample many valuation matrices from the distribution, and solve a particular linear program which computes an empirical estimate of the \alpha-CVaR. Prior work shows that this is a consistent estimator as the number of samples goes to infinity, but they provide a concrete sample complexity bound for subgaussian distributions (Prop. 4.3), under assumption 4.2. When the distribution is normal, only the mean and covariance matter, and a particular QP can be solved (Prop. 4.4). Similar results are discussed for the group egalitarian assignment, in section 4.2. The authors close with a series of experiments on real AAMAS reviewer bid data, where they are able to demonstrate the effectiveness and scalability of their method to a modest sample size. Here, uncertainty is defined by a model fit to predict reviewer bids on unseen examples.

Strengths

The problem considered by the authors, namely, understanding the role of uncertainty in matching algorithms, is important and well-motivated. Furthermore, their setup is extremely practical and testable. I enjoyed reading about the optimization tricks and problem specific structure used to solve each of the instances. I found the fact that the optimal assignment was given by a particular system of linear equations with the optimal dual variables very neat. Overall, I do not think that any other work in the literature has explored and proposed feasible algorithms for matching in the presence of uncertainty in such a variety of settings (robust optimization under both utility optimal and egalitarian matchings, similarly for CVaR). I also appreciate that the code for running all programs was included. It can be quite difficult to implement programs of this complexity in practice (which a skim of the implemented gurobi programs can verify). I encourage the authors to consider cleaning up and releasing the code as a package, there is certainly a dearth of available code for efficient matching algorithms online currently!

Weaknesses

I believe that there are numerous opportunities to improve the quality of presentation in the main paper. If addressed, I plan to raise my score. Most importantly, Propositions 3.4, 3.3, 3.1 should — in my opinion — be stated informally in the main text. I don’t believe the reader will gain much by staring at the (impressive!) equations. Having the informal version retain the key takeaway — we can reduce to solving a system of equations with the optimal dual variables — would make it much easier to parse, and also emphasize the important (and cool) part of the proposition. In lines 25-30, you specify where uncertainty may arise from. I would also consider mentioning how you capture this uncertainty: namely, by modeling uncertainty over agent-item pair valuations. These valuations completely capture the “preferences” you discuss later in the paragraph. However, when I think of matchings and preferences, concepts like stability come to mind, which is a much harder constraint to satisfy than utility maximization. I detail more areas for improving clarity and preciseness of writing in the questions section of my response. Most are minor besides simplifying the main statement of the propositions. If accepted, I also recommend utilizing the additional space to discuss the experiments in further detail. In particular, is table 2 reporting normalized utilities? Why are most of the results 1.0? Is this to be expected?

Questions

Q1. 113 “always the case that the constraints define a finite set such that |A| ∈ N.” Do you include the possibility that |A| = 0? Q2. 116 “Let u : A × $\mathcal{V}$ × G → R be an affine function mapping from allocations to utilities for each group.” I don't think $\mathcal{V}$ was defined before this statement, but am assuming that it refers to the set of all valuation matrices. However, in 131, you define $\mathcal{V}$ to be the uncertainty set containing the true valuation matrix with high probability. There is probably a cleaner way to get around this without overloading, since it seems the uncertainty set definition is more important for your discussion throughout. Perhaps simply define the utility function to map from an arbitrary collection of valuations to R, instead of the set of all valuations? Q3. 133 Not sure whether $W({u}(a))$ formally type checks properly, shouldn’t $u$ be a function of $V$ as well as the assignment $a$ (and also groups, as defined above?) Q4. 134 $D_v$ is a distribution over valuation matrices? This seems to be defined in the next paragraph 136-145. Is it possible to switch the order of these two paragraphs? Q5. 184 can probably do without the $\forall G \in \mathcal{G}$ in the sentence immediately following $w_G$. Also, I believe that the presentation would be enhanced by considering numbers to be not-bold (e.g. $w_G$, $u_{G_1}$, etc.), and to keep vectors bold (which the authors have already done). This would certainly help parse the notation, since there is quite a bit. Q6. 224: valuation matrices of each group are independent of each other. What are the practical implications of this assumption? What kind of problems are we not capturing by making this assumption? This should be stated clearly, and perhaps the assumption should be stated more formally. Q7. Equation (9), should the LHS term have an overset of $\tilde{v} \sim D_v$ for the CVaR_\alpha term? My understanding is that CVaR is w.r.t. a particular distribution $D$, not a particular draw of the random variable. Are you sweeping this into the CVaR term somehow? Q8. 326: “Section 5.1”, should this be table 2? Q9. How is adversarial welfare defined? The worst welfare within the uncertainty set? Q10. I do not believe that $W_{USW}$ or $W_{GESW}$ appear in the main paper, and hence, you may be able to remove these acronyms from section 2.1. Furthermore, throughout the paper, you seem to not use the acronyms until presenting the experimental results (which makes sense). Q11. 339: “Although the CVaR approach is less important at low noise levels, the CVaR of welfare decreases for both welfare measures as noise increases”. As someone not super familiar with CVaR, why is it less important at lower noise levels? Do we expect the noise to mainly impact the right tail or something like that? Q12. 178-180: rounding algorithms in the case of fractional solutions. Do you think you can expand a bit on how these randomized rounding algorithms apply to your specific experiments? How much of a difference do they make? Is there a Birkhoff-von-Neumann decomposition-type result for randomized-multi-matchings which can be used to sample?

Rating

7

Confidence

4

Soundness

4

Presentation

3

Contribution

3

Limitations

The limitations are discussed in the main body of the paper, the most salient of which is that that CVaR requires solving a large number of linear programs to obtain the desired guarantees, and hence can be slow in practice. I don't see this as that big of a problem, however, especially since in the proposed setting of reviewer assignment, only a single matching is constructed and utilized.

Reviewer 69cR7/10 · confidence 2/52024-07-10

Summary

This paper considers a new fair allocation problem. In the classical fair allocation problem, each item can only be matched with at most one agent, and the utilities are known prior. This paper considers a variant where each item is required to be matched with some number of agents. Agents can also be matched with some number of items. The utilities are not known prior. The problem is motivated by the paper-reviewer matching system. Each agent represents a reviewer and each item is paper. A paper is required to be assigned more than one reviewer. Each reviewer can be matched to more than one paper. In addition, each reviewer has a capacity that indicates the maximum number of papers assigned to her. The utilities are not known prior because the review quality is only known after the assignment. The main contribution of this work is that they designed an efficient method to optimize both the utilitarian and egalitarian objectives. The main technical tools come from the robust optimization area. In particular, they transform the problem into a pure linear programming problem via a series of propositions.

Strengths

1. The paper is well-motivated. I like the model studied in this paper. The reviewer-paper matching algorithm is important. I expect that the proposed algorithmic idea should have the impact of a positive practice. 2. The paper is well-written. I am a theoretical person, I appreciate that the theory part of the paper is carefully organized. The whole proof idea is clear to me.

Weaknesses

I have to say that I am usually working on approximation algorithms and I am not in the right position to judge the technical novelty of these continuous methods. To me, the downside is that the running time of the proposed algorithm is high, especially since the algorithm is required to solve LP. This limits the application of algorithms.

Questions

I don't have any specific questions.

Rating

7

Confidence

2

Soundness

3

Presentation

3

Contribution

3

Limitations

The societal impact is not available.

Reviewer 5Awd7/10 · confidence 4/52024-07-12

Summary

This paper looks at the problem of computing resource-agent matchings under uncertainty in a group setting, where the actual valuations of agent-resource matchings are unknown. When the distribution of the valuation uncertainty is known, they look at stochastic optimization using Conditional Value at Risk (CVaR), and when a set of candidate valuations is known, they look at robust optimization through max-min optimization for worst-case valuation. They present methods to optimize for two objectives: utilitarian, where each individual is equally weighted, and egalitarian, where they look at only the group with the minimum utility. They consider uncertainty sets which are either linear or elliptical, and provide derivations of the optimization algorithm for each combination, reducing the solution to an LP in many cases, and with quadratic programming or sub-gradient ascent in others. They also show some empirical results on a reviewer assignment dataset.

Strengths

The paper formalizes a set of methods to optimize the important problem of fairness and group-aware resource allocation under uncertainty. The paper is well-written, and sufficient description is given for readers to comfortably follow the theory developed. The solutions to the optimization problems are reduced to computationally tractable instances that are popularly solvable, and the analysis for both linear and quadratic uncertainty sets is a welcome addition. The empirical results also provide context on how well certain solution approaches scale. The approach seems to be novel, and encompasses a range of general use cases.

Weaknesses

1. One of the limitations I found in this paper was that the authors did not compare the uncertainty-aware solutions to other prior approaches at solving these problems: what benefit does this approach give us? There is no discussion regarding this. From the results, it seems the base methods (USW and GESW) are better than the Rob and CVaR variants. 2. The results do not show any indication of where the GESW optimization may be beneficial. Instead, it appears that the USW optimization supersedes the GESW optimization. If that is the case, the importance of the 'fairness' part of the paper is greatly diminished. The authors should preferably select an evaluation dataset where the approach can show its benefits, if any, to justify its inclusion. 3. As the authors state, the CVaR optimization is much slower. It also seems to require more information than the robust variant of the problem. Without a sufficient justification for its use, what is the merit of including it?

Questions

1. What is the benefit of using the CVaR and Robust variations of the optimization when the base optimization performs better in most cases? Except adversarial welfare for the robust optimization, there seems to be no meaningful benefit of considering uncertainty. 1.1. Unless I am mistaken, the baseline USW and GESW methods are supposed to be improved by considering uncertainty. If this is not the case, this needs to be mentioned in the paper and clarified. 2. Why would we need to select the GESW objective when USW performs better in almost all cases? What is the significance of the fairness aspect of the paper in this case? 3. Where can this be applied? Can you give some examples of allocation problems that are admissible based on the assumptions and limitations of the approaches (e.g. linear/quadratic valuations), and cases where it would not hold?

Rating

7

Confidence

4

Soundness

3

Presentation

3

Contribution

3

Limitations

The paper addresses some limitations in the text. The authors could expand on the situations in which the proposed methods offer an advantage. The authors should also mention the possible negative side effects of using methods that do not lead to intuitive solutions like using CVaR (in Example 2.1, for example). This may lead to affected users perceiving unfairness or inefficiency when the metric is not easy to understand.

Authorsrebuttal2024-08-09

Use-cases of CVaR and the negligible difference between CVaR and USW in Figure 1

Thank you for asking these important questions. The utility of CVaR is higher when welfare distributions exhibit a left fat-tail, meaning a greater probability mass is concentrated in the left tail. Unlike the robust (minimax) approach, the CVaR method—particularly at higher $\alpha$ values—balances between extreme pessimism and optimizing for the average performance. This is advantageous in allocation problems with high uncertainty, where worst-case optimization can lead to overly pessimistic and inefficient outcomes. For example, in public housing assignments, where families have uncertain preferences for various housing units, worst-case optimization might result in inefficient allocations, leaving many families dissatisfied. By focusing on the worst $\alpha$-percentile preferences through CVaR, allocations can be made more efficiently (while also being robust to uncertainty with high probability), thereby improving overall satisfaction. However, it's crucial to differentiate between when to apply CVaR and when the robust approach is more appropriate. The robust approach is better suited to scenarios with extremely high stakes, where any failure (no matter how small the probability of occurrence) is unacceptable—such as life-or-death situations (e.g., allocating medical supplies after a disaster). It is also effective in low-uncertainty contexts where optimizing for the worst case is reasonable and doesn't significantly reduce efficiency. There are several reasons why CVaR USW behaves very similarly to the uncertainty-unaware USW maximal in Figure 1. 1) When the valuations are sampled from independent Gaussian distributions, the USW is just the mean of independent Gaussian variables. The variance of the utilitarian welfare is $\\frac{\\sum_a\\sum_i\\sigma_{a,i}^2}{(nm)^2}$ where $\sigma_{a,i}$ is the standard deviation of the valuation of item $i$ according to agent $a$. Due to this the variance of the utilitarian measure is fairly low, and USW is a more stable measure compared to GESW. 2) We were sampling valuations from symmetrical Gaussian distributions and so the noise in the valuations was (mostly) getting averaged out. 3) We also had a large number of items with very small variance. In AAMAS 2015 and 2016, around 8-9% of the entries have variance less than 0.005. We verified that when we model valuations using a negatively-skewed Gaussian distribution with the same means and variances [1], we see increasing robustness of CVaR relative to uncertainty-unaware USW. The difference is sharper as the skew parameter gets more negative. For example, on the AAMAS 2015 dataset we tried the following experiment where we sample valuations from a skewed-Gaussian distribution with varying skew parameter. We optimize and evaluate for CVaR$_{0.3}$. | Skew | CVaR USW | USW | |:----:|:--------:|:----:| | -0.5 | 1.64 | 1.56 | | -1 | 1.45 | 1.21 | | -2 | 1.33 | 0.96 | | -5 | 1.29 | 0.84 | | -10 | 1.28 | 0.82 | [1] https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.skewnorm.html Note that in practice, if the valuation matrix is not Gaussian-distributed, we can generate samples from its posterior distribution using Markov Chain Monte Carlo (MCMC) or Variational Inference (VI).

Authorsrebuttal2024-08-09

Example where considering uncertainty (via CVaR approach) helps over the uncertainty-unaware approaches

The following toy example further demonstrates the robustness of CVaR when the welfare distribution has a fat-tail. Consider a scenario with two agents and four items, where each agent must be assigned one item. Each agent's valuation for an item is independent and follows a skewed Gaussian distribution. The mean valuations for the items are represented by the following 2D array: $$ \\begin{bmatrix} 0.39 & 0.49 & 0.51 & 0.53 \\\\ 0.52 & 0.51 & 0.53 & 0.54 \\end{bmatrix} $$ The standard deviations for the Gaussian distributions of the four items are $[0.01, 0.04, 0.05, 0.09]$, with a skewness factor of 5 across all distributions [1]. We aim to optimize the utilitarian welfare, using CVaR$_{0.04}$ as our evaluation metric. This choice maximizes the expected utilitarian welfare over the worst $\alpha = 0.04$ quantile. To achieve this, we sampled 20,000 valuation matrices from the aforementioned distribution of valuations. We then applied three different optimization approaches: the CVaR approach at $\alpha = 0.04$, the Robust approach, and the Naïve approach that optimizes against the mean valuation. The following results were observed on a test set of 20,000 valuations sampled from the same distribution: **CVaR approach:** *Allocation:* $$ \\begin{bmatrix} 0 & 0 & 1 & 0 \\\\ 1 & 0 & 0 & 0 \\end{bmatrix} $$ *Test CVaR$_{0.04}$ utilitarian welfare:* 0.985 **Robust approach:** *Allocation:* $$ \\begin{bmatrix} 0 & 1 & 0 & 0 \\\\ 1 & 0 & 0 & 0 \\end{bmatrix} $$ *Test CVaR$_{0.04}$ utilitarian welfare:* 0.972 **Naïve approach:** *Allocation:* $$ \\begin{bmatrix} 0 & 0 & 0 & 1 \\\\ 0 & 0 & 1 & 0 \\end{bmatrix} $$ *Test CVaR$_{0.04}$ welfare:* 0.947 **Observations** The naïve approach selects items 3 and 4, as they have the highest mean values, but it does not account for the uncertainty in the preferences. The robust approach, being more conservative, chooses items 1 and 2 due to their lower uncertainty. The CVaR approach strikes a balance between these two methods, selecting items 1 and 3. Item 1 has a lower mean value with low uncertainty, while item 3 has a higher mean value but with slightly higher uncertainty than item 2. We then repeat the experiment, this time optimizing for egalitarian welfare across agents, and evaluate the results by measuring CVaR$_{0.04}$ of the egalitarian welfare. **CVaR approach:** *Allocation:* $$ \\begin{bmatrix} 0 & 0 & 1 & 0 \\\\ 1 & 0 & 0 & 0 \\end{bmatrix} $$ *Test CVaR$_{0.04}$ egalitarian welfare:* 0.47 **Robust approach:** *Allocation:* $$ \\begin{bmatrix} 1 & 0 & 0 & 0 \\\\ 0 & 1 & 0 & 0 \\end{bmatrix} $$ *Test CVaR$_{0.04}$ egalitarian welfare:* 0.38 **Naïve approach:** *Allocation:* $$ \\begin{bmatrix} 0 & 0 & 0 & 1 \\\\ 0 & 0 & 1 & 0 \\end{bmatrix} $$ *Test CVaR$_{0.04}$ egalitarian welfare:* 0.45 **Observations** We notice a significant decline in the performance of the robust approach. This decline occurs because, although items 1 and 2 have lower uncertainty compared to items 3 and 4, item 1 has significantly lower uncertainty than item 2, resulting in better worst-case utility for item 1. Since the CVaR approach selected items 3 and 1, it achieves a higher CVaR of egalitarian welfare at $\alpha = 0.04$ compared to the robust approach which selects items 1 and 2. We also see fairly similar results when setting $\alpha=0.2$ with the other problem parameters remaining the same. [1] https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.skewnorm.html

Reviewer 5Awd2024-08-12

Thank you for responding to my follow-up. The example is useful in helping me understand when uncertainty-aware solutions are better. I would recommend the authors to add an experiment that demonstrates it, to make a stronger case for their paper. Considering our discussion during this period, I will increase my score by one point.

Authorsrebuttal2024-08-13

Thank you!

Thank you so much for promptly checking our response and raising the score! We sincerely appreciate your feedback and constructive criticism on our paper. We will incorporate the suggested changes in the camera-ready version.

Reviewer ZdA86/10 · confidence 1/52024-07-13

Summary

The authors study a resource allocation problem where the objective is to optimize for efficiency and fairness under the presence of some uncertainty. The agents are partitioned into groups and the items need to be assigned so as to be fair to the groups. They study two maximization objectives : 1) a weighted sum of social welfare of groups and 2) the social welfare of the group with the least amount of welfare. They study two models of uncertainty over the agents' valuations of the items - the first is when there is some explicit uncertainty set and the objective is to maximize the worst-case outcome across this uncertainty set, and the second is when the distribution over the valuations is known and the objective is to maximize the Conditional Value at Risk (CVaR$_\alpha$). Their primary contribution is to present formulations of these problems as various mathematical programs. They make assumptions on the uncertainty sets so as to make these formulations suitable for optimization in certain cases. They present experiments supporting their models and formulations.

Strengths

1) Their experiments suggest that optimizing for CVaR has some benefit under a large uncertainty regime, as well as points to the benefit of considering the robust uncertainty model. 2) Their formulation and subsequent solution appears to be faster, and more successful, than applying the naive solver.

Weaknesses

The paper is a bit terse, and hard to follow at times. It would be beneficial to add more details where necessary. For example, in Line 180, please mention the randomized rounding procedure or at least the properties of the output rounded solution. Ideally, there would at least be some more details to get to the propositions.

Questions

1) Are there any tractable settings where the agents can be in multiple groups at once? There are many resource allocation, clustering, etc. papers where such settings are considered. 2) Can't the robust setting be viewed as a subcase of the CVaR setting where you maximize the worst (ie. the 0 percentile) outcome?

Rating

6

Confidence

1

Soundness

3

Presentation

2

Contribution

2

Limitations

N/A

Reviewer hqEo4/10 · confidence 2/52024-07-15

Summary

The authors of this paper investigate the fair multi-matching problem under uncertainty. Both stochastic and robust optimizations are considered to solve the proposed problem.

Strengths

S1. Fairness is an important and practical concern in resource allocation problems. S2. The theoretical results of this paper seem to be correct.

Weaknesses

W1. The authors fail to provide a clear motivation example of the proposed research problem in real applications. For the reviewer assignment application, I do not see strong reasons why we need to consider uncertainty (the reviewers have already revealed their preferences over papers clearly) and fairness (the utility of a reviewer is how the assigned papers match her/him?). W2. What are the major technical challenges of considering both group fairness and uncertainty in the resource allocation problem? Since there are works discussing each factor (fairness and uncertainty), can we just adapt existing techniques to solve the proposed problem? Are there any new technical challenges caused by the combination of fairness and uncertainty? W3. Does the process of obtaining the uncertainty set assume a true model (usually linear models in statistics) of the uncertain parameters? If so, I am afraid the way of considering uncertainty is still a toy research problem because in reality the data distribution is usually unknown and complicated. The uncertainty set itself may not be constructed in a reliable way. For example, under what conditions can the authors prove that their uncertainty sets described in Appendix D have the true parameters? To prove this, I guess the authors may need strong assumptions about the data distribution. W4. As the reviewer assignment datasets do not have groups of reviewers, the authors may want to try different ways of grouping reviewers in experiments to make the results more convincing w.r.t. group fairness. W5. It seems to me that the experiments are just numerical simulations where the objectives are simulated. This makes the research problem a toy one as no real datasets are used to verify the proposed method.

Questions

W1, W2, W3

Rating

4

Confidence

2

Soundness

3

Presentation

2

Contribution

2

Limitations

W1, W3, W4, W5

Reviewer UQiJ2024-08-07

The authors have addressed all my concerns and I have raised my score by one point. I found the submission very interesting, and think the paper will be a great contribution to the area!

Authorsrebuttal2024-08-07

Thank you!

Thank you so much for promptly reviewing our rebuttal and raising the score. We greatly appreciate it and are glad that you found our paper interesting!

Reviewer 5Awd2024-08-07

Thanks for answering my questions and concerns. The new experiment included in the global response helps show the value of the GESW function. This should be included in the main paper. I am still not convinced about the need to include CVaR optimization. I agree that it allows better analysis, and it may be widely used, but what is the contribution it makes in the context of this paper? I understand that the paper's contribution is not in designing a better solution to an existing problem, but to define a new optimization. Yet, the benefits of using this optimization could be made clearer. From Figure 1, yes, the uncertainty-unaware solutions do become worse as the level of uncertainty increases, but so do the uncertainty-aware solutions (especially looking at USW, the difference is negligible). Can the authors comment on why this is, and perhaps give an example where considering uncertainty helps over the uncertainty-unaware approaches (specifically for CVaR)?

Reviewer 69cR2024-08-11

I'd like to thank the authors for replying to my concerns. I still think that the running time restricts the application of algorithms. But I believe that the studied problem is very interesting. Considering that I am not able to judge the technical novelty of this paper, I will keep my score and confidence unchanged.

Authorsrebuttal2024-08-14

Thank you!

Thank you again for your time and effort in reviewing our paper, and for both the positive comments and critiques.

Program Chairsdecision2024-09-25

Decision

Accept (poster)

© 2026 NYSGPT2525 LLC