Online metric algorithms with untrusted predictions

Antoniadis, Antonios; Coester, Christian; Eliáš, Marek; Polak, Adam G.; Simon, Bertrand

doi:10.48550/arxiv.2003.02144

Cited by 4 publications

(10 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…And interestingly, this value is a function of m, as the number of subsets, and θ = 2, as the algorithm proposed in [31] is a suboptimal algorithm. This is indeed our observation in Figure 4, where the degradation factor for m = 10 5 with θ = [2,3] is less than 1. This shows that the original algorithm proposed in [31] is not using the worst-case optimized parameter and one can find the worst-case optimized parameter θ by minimizing Eq.…”

Section: F Regret Analysis Of Data-driven Algorithm Design For One-wa...supporting

confidence: 85%

“…An important open question is how to achieve the "best-of-bothworlds", both near-optimal performance in typical settings, which requires data-driven adaptation, and a near-optimal competitive ratio. Toward this goal, there have been substantial efforts to improve the performance of competitive algorithms using predictions [1,2,3], ML advice [4,5,6,7], and advice from multiple experts [8]. In these approaches the goal is to allow online algorithms to use (potentially noisy) predictions (or advice) about future inputs.…”

Section: Introductionmentioning

confidence: 99%

“…Such predictions capture the fact that, often, something is known about the future that could be used to improve the performance in typical cases. These approaches have been successfully applied to design competitive algorithms that perform near-optimally in the typical case in settings such as the ski-rental problem [6,9,5], online optimization with switching costs [1], online caching [4,7], and metrical task systems [3], to name a few. However, in this literature, the power of ML models has been leveraged to first, predict the future input, and then modify the algorithms to use this additional input to further improve the performance.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Data-driven Competitive Algorithms for Online Knapsack and Set Cover

Zeynali¹,

Sun²,

Hajiesmaili³

et al. 2020

Preprint

View full text Add to dashboard Cite

The design of online algorithms has tended to focus on algorithms with worst-case guarantees, e.g., bounds on the competitive ratio. However, it is well-known that such algorithms are often overly pessimistic, performing sub-optimally on non-worst-case inputs. In this paper, we develop an approach for data-driven design of online algorithms that maintain near-optimal worst-case guarantees while also performing learning in order to perform well for typical inputs. Our approach is to identify policy classes that admit global worst-case guarantees, and then perform learning using historical data within the policy classes. We demonstrate the approach in the context of two classical problems, online knapsack and online set cover, proving competitive bounds for rich policy classes in each case. Additionally, we illustrate the practical implications via a case study on electric vehicle charging.

show abstract

Section: F Regret Analysis Of Data-driven Algorithm Design For One-wa...supporting

confidence: 85%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Data-driven Competitive Algorithms for Online Knapsack and Set Cover

Zeynali¹,

Sun²,

Hajiesmaili³

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Remark The independent, concurrent work of Antoniadas et al [2] has slight overlap with ours. In particular, they also showed that the PRP prediction model does not provide asymptotic benefits for randomized algorithms.…”

Section: Related Worksupporting

confidence: 66%

Online Algorithms for Weighted Paging with Predictions

Jiang

Panigrahi

Sun

2020

Preprint

View full text Add to dashboard Cite

In this paper, we initiate the study of the weighted paging problem with predictions. This continues the recent line of work in online algorithms with predictions, particularly that of Lykouris and Vassilvitski (ICML 2018) and Rohatgi (SODA 2020) on unweighted paging with predictions. We show that unlike unweighted paging, neither a fixed lookahead nor knowledge of the next request for every page is sufficient information for an algorithm to overcome existing lower bounds in weighted paging. However, a combination of the two, which we call the strong per request prediction (SPRP) model, suffices to give a 2-competitive algorithm. We also explore the question of gracefully degrading algorithms with increasing prediction error, and give both upper and lower bounds for a set of natural measures of prediction error.

show abstract

“…Closest to our work are the recent papers of Antoniadis et al (2020) and Rutten et al (2022). The former considers the problem of designing algorithms for metrical task systems (MTS) with blackbox advice.…”

Section: Online Algorithms With Black-box Advicementioning

confidence: 99%

Chasing Convex Bodies and Functions with Black-Box Advice

Christianson¹,

Handina²,

Wierman³

2022

Preprint

View full text Add to dashboard Cite

We consider the problem of convex function chasing with black-box advice, where an online decision-maker aims to minimize the total cost of making and switching between decisions in a normed vector space, aided by black-box advice such as the decisions of a machine-learned algorithm. The decision-maker seeks cost comparable to the advice when it performs well, known as consistency, while also ensuring worst-case robustness even when the advice is adversarial. We first consider the common paradigm of algorithms that switch between the decisions of the advice and a competitive algorithm, showing that no algorithm in this class can improve upon 3-consistency while staying robust. We then propose two novel algorithms that bypass this limitation by exploiting the problem's convexity. The first, INTERP, achieves ( √ 2 + )-consistency and O( C 2 )-robustness for any > 0, where C is the competitive ratio of an algorithm for convex function chasing or a subclass thereof. The second, BDINTERP, achieves (1 + )-consistency and O( CD )-robustness when the problem has bounded diameter D. Further, we show that BDINTERP achieves near-optimal consistency-robustness trade-off for the special case where cost functions are α-polyhedral.

show abstract

Online metric algorithms with untrusted predictions

Cited by 4 publications

References 23 publications

Data-driven Competitive Algorithms for Online Knapsack and Set Cover

Data-driven Competitive Algorithms for Online Knapsack and Set Cover

Online Algorithms for Weighted Paging with Predictions

Chasing Convex Bodies and Functions with Black-Box Advice

Contact Info

Product

Resources

About