A Rules-Based Approach for Configuring Chains of Classifiers in Real-Time Stream Mining Systems

Foo, Brian; Schaar, Mihaela van der

doi:10.1155/2009/975640

Cited by 5 publications

(2 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The proposed algorithms are compared against several benchmarks: random, safe experimentation, UCB1 [29], and optimal. Safe experimentation (SE) is a method used in [40] when there is no uncertainty about the accuracy of the classifiers. In each period t, each classifier selects its baseline action with probability 1 − t or selects a new random action with probability t .…”

Section: Accelerating Learning Through Reward Informativenessmentioning

confidence: 99%

Multiagent Systems: Learning, Strategic Behavior, Cooperation, and Network Formation

Tekin

Zhang

et al. 2018

Cooperative and Graph Signal Processing

Self Cite

View full text Add to dashboard Cite

Proliferation of web-based multimedia content sources and servers led to a tremendous growth in the volume and diversity of multimedia content that is consumed by a diverse set of users. This diversity results in users exhibiting a vast range of preferences over the content, which often depends on the context in which they consume the content. Such demand led to the emergence of multimedia content aggregators (MCAs) [1,2] that gather and fuse content from numerous multimedia sources to provide a ubiquitous content delivery experience for their users. It is thus essential for these systems to learn the context-specific content preferences of their users using past feedback of their users on the content that they provide. The context of a user includes information that is related to its content preferences, including but not limited to the location information, search query, gender, age, and the type of the device that the user is using (e.g., mobile phone, tablet, PC) to access the content [3]. Thus, the goal of the MCA is to match its users with the most appropriate content by learning how users with different contexts react to different contents. Such learning is necessary for continuous satisfaction of a user's request for content, which dynamically evolves over time depending on how the user's context evolves. This problem can be formulated as an online learning problem where the MCA learns the best content for its users through exploring how its users react to different content. In order to maximize the user satisfaction, an MCA needs to connect with other MCAs that have access to other multimedia sources to find the right content for its users. This requires cooperation between the MCAs: In addition to serving its own users, an MCA should also serve content to the users of other MCAs when a request is made. Thus, each MCA has two types of users: direct users, who visit the website of the MCA to search for content, and indirect users, who are users of another MCA Cooperative and Graph Signal Processing.

show abstract

Section: Accelerating Learning Through Reward Informativenessmentioning

confidence: 99%

Multiagent Systems: Learning, Strategic Behavior, Cooperation, and Network Formation

Tekin

Zhang

et al. 2018

Cooperative and Graph Signal Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…(2) Safe Experimentation (SE): This is a method used in [6] when there is no uncertainty about the accuracy of the classifiers. In each period t, each classifier selects its baseline action with probability 1 − ϵ t or selects a new random action with probability ϵ t .…”

Section: Performance Comparisonmentioning

confidence: 99%

Distributed Multi-Agent Online Learning Based on Global Feedback

Tekin

Zhang

et al. 2015

IEEE Trans. Signal Process.

Self Cite

View full text Add to dashboard Cite

In many types of multi-agent systems, distributed agents cooperate with each other to take actions with the goal of maximizing an overall system reward. However, in many of these systems, agents only receive a (perhaps noisy) global feedback about the realized overall reward rather than individualized feedback about the relative merit of their own actions with respect to the overall reward. If the contribution of an agent's actions to the overall reward is unknown a priori, it is crucial for the agents to utilize a distributed algorithm that still allows them to learn their best actions. In this paper, we rigorously formalize this problem and develop online learning algorithms which enable the agents to cooperatively learn how to maximize the overall reward in these global feedback scenarios without exchanging any information among themselves. We prove that, if the agents observe the global feedback without errors, the distributed nature of the considered multi-agent system results in no performance loss compared with the case where agents can exchange information. When the agents' individual observations are erroneous, existing centralized algorithms, including popular ones like UCB1, break down. To address this challenge, we propose a novel class of distributed algorithms that are robust to individual observation errors and whose performance can be analytically bounded. We prove that our algorithms' learning regrets -the losses incurred by the algorithms due to uncertainty -are logarithmically increasing in time and thus the time average reward converges to the optimal average reward. Moreover, we also illustrate how the regret depends on the size of the action space, and we show that this relationship is influenced by the informativeness of the reward structure with regard to each agent's individual action.We prove that when the overall reward is fully informative, regret is linear in the total number of actions , IEEE Transactions on Signal Processing 2 of all the agents. When the reward function is not informative, regret is linear in the number of joint actions. Our analytic and numerical results show that the proposed learning algorithms significantly outperform existing online learning solutions in terms of regret and learning speed. We illustrate how our theoretical framework can be used in practice by applying it to online Big Data mining using distributed classifiers. However, our framework can be applied to many other applications including online distributed decision making in cooperative multi-agent systems (e.g. packet routing or network coding in multi-hop networks), cross-layer optimization (e.g. parameter selection in different layers), multi-core processors etc. Index TermsMulti-agent learning, online learning, multi-armed bandits, Big Data mining, distributed cooperative learning, reward informativeness.

show abstract

A New Data Stream Mining Algorithm for Interestingness-Rich Association Rules

Kuthadi¹

2013

Journal of Computer Information Systems

View full text Add to dashboard Cite

A Rules-Based Approach for Configuring Chains of Classifiers in Real-Time Stream Mining Systems

Cited by 5 publications

References 29 publications

Multiagent Systems: Learning, Strategic Behavior, Cooperation, and Network Formation

Multiagent Systems: Learning, Strategic Behavior, Cooperation, and Network Formation

Distributed Multi-Agent Online Learning Based on Global Feedback

A New Data Stream Mining Algorithm for Interestingness-Rich Association Rules

Contact Info

Product

Resources

About