Adaptive Routing Using Expert Advice

György, András; Ottucsák, György

doi:10.1093/comjnl/bxh168

Cited by 12 publications

(11 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…if I t = i and S t = 1; 0 otherwise, based on György and Ottucsák [11]. Therefore, the estimated loss is an unbiased estimate of the true loss with respect to its natural filtration, that is,…”

Section: The Algorithmmentioning

confidence: 99%

“…In many applications, including regression problems (Györfi and Lugosi [9]) or routing in communication networks (cf. György and Ottucsák [11]) the loss is unbounded. The main aim of this paper is to show Hannan consistency of on-line algorithms for unbounded losses under partial monitoring.…”

Section: Introductionmentioning

confidence: 99%

“…al [5]). Finally the combination of the label efficient and the multi-armed bandit problem, where after choosing a decision, the algorithm learns its own loss if and only if it asks for it (see György and Ottucsák [11]). …”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Hannan Consistency in On-Line Learning in Case of Unbounded Losses Under Partial Monitoring

Allenberg

Auer

Györfi

et al. 2006

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Abstract. In this paper the sequential prediction problem with expert advice is considered when the loss is unbounded under partial monitoring scenarios. We deal with a wide class of the partial monitoring problems: the combination of the label efficient and multi-armed bandit problem, that is, where the algorithm is only informed about the performance of the chosen expert with probability ε ≤ 1. For bounded losses an algorithm is given whose expected regret scales with the square root of the loss of the best expert. For unbounded losses we prove that Hannan consistency can be achieved, depending on the growth rate of the average squared losses of the experts.

show abstract

Section: The Algorithmmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Hannan Consistency in On-Line Learning in Case of Unbounded Losses Under Partial Monitoring

Allenberg

Auer

Györfi

et al. 2006

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

show abstract

“…This general framework of learning from expert advice was introduced by Littlestone and Warmuth [1] and Vovk [2]. Besides online recommendation systems, the framework has been applied to various other applications, for instance, finding the shortest path problem [3], [4], [5] , the metrical task system [6], and online paging [7].…”

Section: Introductionmentioning

confidence: 99%

Optimal adversarial strategies in learning with expert advice

Truong

Kiyavash

2013

52nd IEEE Conference on Decision and Control

View full text Add to dashboard Cite

Abstract-We propose an adversarial setting for the framework of learning with expert advice in which one of the experts has the intention to compromise the recommendation system by providing wrong recommendations. The problem is formulated as a Markov Decision Process (MDP) and solved by dynamic programming. Somewhat surprisingly, we prove that, in the case of logarithmic loss, the optimal strategy for the malicious expert is the greedy policy of lying at every step. Furthermore, a sufficient condition on the loss function is provided that guarantees the optimality of the greedy policy. Our experimental results, however, show that the condition is not necessary since the greedy policy is also optimal when the square loss is used, even though the square loss does not satisfy the condition. Moreover, the experimental results suggest that, for absolute loss, the optimal policy is a threshold one.

show abstract

“…The application of learning automata in dynamic routing has been widely studied in different applications such as telephone routing [11], [12], wavelength routing in WDM networks [13] and MPLS routing [14]. An application of multi-armed bandit algorithms in adaptive routing was presented in [15].…”

Section: Introductionmentioning

confidence: 99%

Reinforcement Learning-Based Load Shared Sequential Routing

Heidari

Mannor

Mason

2007

NETWORKING 2007. Ad Hoc and Sensor Networks, Wireless Networks, Next Generation Internet

View full text Add to dashboard Cite

Abstract. We consider event dependent routing algorithms for on-line explicit source routing in MPLS networks. The proposed methods are based on load shared sequential routing in which load sharing factors are updated using learning algorithms. The learning algorithms we employ are either based on learning automata or on online learning algorithms that were originally devised for solving the adversarial multi-armed bandit problem. While simple to implement, the performance of the proposed learning algorithms in terms of blocking probability compares favorably with the performance of other event dependent routing methods proposed for MPLS routing such as the Success-to-the-top algorithm. We demonstrate the convergence of one of the learning algorithms to the user equilibrium within a set of discrete event simulations.

show abstract

Adaptive Routing Using Expert Advice

Cited by 12 publications

References 32 publications

Hannan Consistency in On-Line Learning in Case of Unbounded Losses Under Partial Monitoring

Hannan Consistency in On-Line Learning in Case of Unbounded Losses Under Partial Monitoring

Optimal adversarial strategies in learning with expert advice

Reinforcement Learning-Based Load Shared Sequential Routing

Contact Info

Product

Resources

About