RAVEN-UCB proposes a variance-adaptive UCB algorithm for non-stationary bandits, but the proof of its main regret bound is mathematically invalid.
Applying Multi-armed Bandit Algorithms to Computational Advertising
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
Over the last two decades, we have seen extensive industrial research in the area of computational advertising. In this paper, our goal is to study the performance of various online learning algorithms to identify and display the best ads/offers with the highest conversion rates to web users. We formulate our ad-selection problem as a Multi-Armed Bandit problem which is a classical paradigm in Machine Learning. We have been applying machine learning, data mining, probability, and statistics to analyze big data in the ad-tech space and devise efficient ad selection strategies. This article highlights some of our findings in the area of computational advertising from 2011 to 2015.
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
REJECT 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
From Theory to Practice with RAVEN-UCB: Addressing Non-Stationarity in Multi-Armed Bandits through Variance Adaptation
RAVEN-UCB proposes a variance-adaptive UCB algorithm for non-stationary bandits, but the proof of its main regret bound is mathematically invalid.