Learning diverse rankings with multi-armed bandits

By Filip Radlinski, Robert D. Kleinberg, Thorsten Joachims. 2008

569 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Networks and Communications, Management Science and Operations Research

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Networks and Communications

In Management Science and Operations Research

Citations in the corpus

Determinantal Point Processes for Machine Learning (2012, background): “This focus on diversity places DPPs alongside a number of recently developed techniques for working with diverse sets, particularly in the information retrieval community [23, 26, 121, 122, 140, 158, 159].”
Ranking with Fairness Constraints (2017, methodology): “Combining the learning of values along with the ranking of items has also been studied [RKJ08, SRG13]; in each round an algorithm chooses an ordered list of k documents as a function of the estimated values w ij and can receive a click on one of them.”
Diversifying search results (2009, background): “Radlinski, Kleinberg, and Joachims [13] propose a learning algorithm to compute an ordering of search results from a diverse set of orderings.”
Near-Optimal MAP Inference for Determinantal Point Processes (2012, background): “Informative subset selection problems arise in many applications where a small number of items must be chosen to represent or cover a much larger set; for instance, text summarization [1, 2], document and image search [3, 4, 5], sensor placement [6], viral marketing [7], and many others.”
Online learning to diversify from implicit feedback (2012, background): “While some online learning methods exist that can exploit click data, those methods either cannot generalize across queries [11] and/or have a hard-coded notion of diversity that cannot be adjusted through learning [15].”
Feeling lucky?: multi-armed bandits for ordering judgements in pooling-based evaluation (2016, background): “Klein-berg and colleagues [14] presented a multi-armed bandit al-gorithm that learns a diverse ranking of results based on clickthrough data.”
Submodular Function Maximization (2014, background): “…optimization, news recommendation systems, online ad selection, and combining multiple heuristics for computationally hard problems (Streeter and Golovin, 2008; Golovin et al., 2010b; Munagala et al., 2005; Babu et al., 2004; Streeter et al., 2009; Radlinski et al., 2008; Streeter et al., 2007a).”
Toward Controlling Discrimination in Online Ad Auctions (2019, unknown): “In addition, [6,13,44] study fair ranking algorithms; these could be used to generate a balanced list of results on job platforms and other search engines.”
Rank and relevance in novelty and diversity metrics for recommender systems (2011, methodology): “Our approach takes inspiration in related work on user click models in information retrieval systems [6,10,13,17,18], but any other alternative modeling approach could be plugged into our framework.”
Learning to Diversify from Implicit Feedback (2012, background): “While some online learning methods exist that can exploit click data, those methods either cannot generalize across queries [9] and/or have a hard-coded notion of diversity that cannot be adjusted through learning [13].”
Almost Optimal Exploration in Multi-Armed Bandits (2013, background): “, 2007), learning search-ranking from usage data (Radlinski et al., 2008) and more.”
Novelty and Diversity in Recommender Systems (2015, methodology): “Our approach takes inspiration in related work on user click models in information retrieval systems [6,10,13,17,18], but any other alternative modeling approach could be plugged into our framework.”
Cascading Bandits: Learning to Rank in the Cascade Model (2015, background): “The solutions in ranked bandits are (1− 1/e) approximate and the regret is Ω(K) (Radlinski et al., 2008), where K is the number of recommended items.”
A probabilistic method for inferring preferences from clicks (2011, methodology): “Methods for optimizing the click-through rates in ad placement [19] and web search [22] have also learned effectively from click data.”