<div class="csl-bib-body">
<div class="csl-entry">Drmota, M., Jacquet, P., Wu, C., & Szpankowski, W. (2024). Minimax Regret with Unbounded Weights. In <i>2024 IEEE International Symposium on Information Theory (ISIT)</i> (pp. 2305–2310). IEEE. https://doi.org/10.1109/ISIT57864.2024.10619590</div>
</div>
-
dc.identifier.uri
http://hdl.handle.net/20.500.12708/210053
-
dc.description.abstract
In online learning, a learner receives data in rounds 1 t T and at each round predicts a label which is then compared to the true label resulting in a loss. The total loss over T rounds, when compared to a loss over the best expert from a class of experts, is called the regret. This paper focuses on logarithmic loss over a class of experts represented by a probability distribution p and parameterized by addimensional weight vector w. Unlike previous work that studied bounded weights, we assume that the norm of the weights can be unbounded. This unboundedness poses a challenging problem that leads to unexpected results. For such a class of weighted experts we analyze the (fixed design) minimax regret for the best predictor and worst label sequence. Such a minimax regret turns out to be a universal lower bound for most regrets analyzed in the literature. For bounded weights it is known that the minimax regret can grow like where R is an upper bound on the weight norm. In contrast, we show in this paper that for unbounded norm with R the minimax regret is asymptotically (d-1) for a logistic-like expert class which we also extend to R We prove our findings by introducing the so called splittable label sequences that partition the weight space into regions with maximum sequence probability equal to 1. Finally, for a general class of monotone experts we present an upper bound 2d log T for the regret.
en
dc.language.iso
en
-
dc.subject
Minimax regret
en
dc.subject
Unbounded Weights
en
dc.title
Minimax Regret with Unbounded Weights
en
dc.type
Inproceedings
en
dc.type
Konferenzbeitrag
de
dc.contributor.affiliation
Institut national de recherche en informatique et en automatique, France
-
dc.contributor.affiliation
Purdue University West Lafayette, United States of America (the)
-
dc.contributor.affiliation
Purdue University West Lafayette, United States of America (the)
-
dc.relation.isbn
979-8-3503-8284-6
-
dc.relation.doi
10.1109/ISIT57864.2024
-
dc.description.startpage
2305
-
dc.description.endpage
2310
-
dc.type.category
Full-Paper Contribution
-
tuw.booktitle
2024 IEEE International Symposium on Information Theory (ISIT)
-
tuw.peerreviewed
true
-
tuw.relation.publisher
IEEE
-
tuw.researchTopic.id
A3
-
tuw.researchTopic.name
Fundamental Mathematics Research
-
tuw.researchTopic.value
100
-
tuw.publication.orgunit
E104-05 - Forschungsbereich Kombinatorik und Algorithmen
-
tuw.publisher.doi
10.1109/ISIT57864.2024.10619590
-
dc.description.numberOfPages
6
-
tuw.author.orcid
0000-0001-7919-1206
-
tuw.author.orcid
0000-0001-5255-1275
-
tuw.author.orcid
0000-0001-9062-0067
-
tuw.event.name
2024 IEEE International Symposium on Information Theory (ISIT)
en
tuw.event.startdate
07-07-2024
-
tuw.event.enddate
12-07-2024
-
tuw.event.online
On Site
-
tuw.event.type
Event for scientific audience
-
tuw.event.place
Athen
-
tuw.event.country
GR
-
tuw.event.presenter
Wu, Changlong
-
wb.sciencebranch
Informatik
-
wb.sciencebranch
Mathematik
-
wb.sciencebranch.oefos
1020
-
wb.sciencebranch.oefos
1010
-
wb.sciencebranch.value
5
-
wb.sciencebranch.value
95
-
item.languageiso639-1
en
-
item.openairetype
conference paper
-
item.grantfulltext
restricted
-
item.fulltext
no Fulltext
-
item.cerifentitytype
Publications
-
item.openairecristype
http://purl.org/coar/resource_type/c_5794
-
crisitem.author.dept
E104 - Institut für Diskrete Mathematik und Geometrie
-
crisitem.author.dept
Institut national de recherche en informatique et en automatique