Tracking the Best of Many Experts

György, András; Linder, Tamás; Lugosi, Gábor

doi:10.1007/11503415_14

Cited by 25 publications

(40 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the expert setting there is a long history of algorithms that can handle the case when the best expert is allowed to shift over time. This is achieved by first doing an exponential update and then mixing in a bit of either the uniform distribution over all experts or the average of all past distributions over the experts [18,3,10]. In all experimental evaluations that the authors are aware of, it was crucial to extend the online algorithms to the shifting expert case [16,15].…”

Section: Open Problemsmentioning

confidence: 99%

Combining initial segments of lists

Warmuth

Koolen

Helmbold

2014

Theoretical Computer Science

View full text Add to dashboard Cite

a r t i c l e i n f o a b s t r a c t Keywords:Online learning Ranking Worst-case analysis Relative loss boundsWe propose a new way to build a combined list from K base lists, each containing N items. A combined list consists of top segments of various sizes from each base list so that the total size of all top segments equals N. A sequence of item requests is processed and the goal is to minimize the total number of misses. That is, we seek to build a combined list that contains all the frequently requested items. We first consider the special case of disjoint base lists. There, we design an efficient algorithm that computes the best combined list for a given sequence of requests. In addition, we develop a randomized online algorithm whose expected number of misses is close to that of the best combined list chosen in hindsight. We prove lower bounds that show that the expected number of misses of our randomized algorithm is close to the optimum. In the presence of duplicate items, we show that computing the best combined list is NP-hard. We show that our algorithms still apply to a linearized notion of loss in this case. We expect that this new way of aggregating lists will find many ranking applications.

show abstract

Section: Open Problemsmentioning

confidence: 99%

Combining initial segments of lists

Warmuth

Koolen

Helmbold

2014

Theoretical Computer Science

View full text Add to dashboard Cite

show abstract

“…In the expert setting there is a long history of algorithms that can handle the case when the best expert is allowed to shift over time. This is achieved by first doing an exponential update and then mixing in a bit of either the uniform distribution over all experts or the average of all past distributions over the experts [HW98,BW02,GLL05]. In all experimental evaluations that the authors are aware of, it was crucial to extend the online algorithms to the shifting expert case [HLSS00,GWBA02].…”

Section: Open Problemsmentioning

confidence: 99%

Combining Initial Segments of Lists

Warmuth

Koolen

Helmbold

2011

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. We propose a new way to build a combined list from K base lists, each containing N items. A combined list consists of top segments of various sizes from each base list so that the total size of all top segments equals N . A sequence of item requests is processed and the goal is to minimize the total number of misses. That is, we seek to build a combined list that contains all the frequently requested items. We first consider the special case of disjoint base lists. There, we design an efficient algorithm that computes the best combined list for a given sequence of requests. In addition, we develop a randomized online algorithm whose expected number of misses is close to that of the best combined list chosen in hindsight. We prove lower bounds that show that the expected number of misses of our randomized algorithm is close to the optimum. In the presence of duplicate items, we show that computing the best combined list is NP-hard. We show that our algorithms still apply to a linearized notion of loss in this case. We expect that this new way of aggregating lists will find many ranking applications.

show abstract

“…While this complexity may be prohibitive for large classes of experts, in many situations, such as the shortest path problem in graphs, the special structure of the experts allows to implement the algorithms with significantly lower complexity, see, e.g., [12], [13], [14], [15], [16], [17], [18], [8], [19], [20].…”

Section: Sequential Decision Problemsmentioning

confidence: 99%

“…Next, following [19] and [20], we show how to implement the algorithm to compete with paths of a fixed length K. Then it is explained how this algorithm can be extended to compete with all paths, and what simplifications can be made for acyclic graphs.…”

Section: Theorem 42 (Littlestone and Warmuth [4])mentioning

confidence: 99%

“…, |V|}, and then choosing randomly I t from paths of length K as above. The computational complexity of this algorithm is O(n|V||E|), that is typically significantly less than the O(nN ) complexity of the original method of Algorithm 2 [20]. For acyclic graphs, there is no need for the second variable of G t , and a similar algorithm can be performed in O(n|E|) time, see [13], [17], [18].…”

Section: Theorem 42 (Littlestone and Warmuth [4])mentioning

confidence: 99%

See 1 more Smart Citation