extreme bandit
Extreme bandits
In many areas of medicine, security, and life sciences, we want to allocate limited resources to different sources in order to detect extreme values. In this paper, we study an efficient way to allocate these resources sequentially under limited feedback. While sequential design of experiments is well studied in bandit theory, the most commonly optimized property is the regret with respect to the maximum mean reward. However, in other problems such as network intrusion detection, we are interested in detecting the most extreme value output by the sources. Therefore, in our work we study extreme regret which measures the efficiency of an algorithm compared to the oracle policy selecting the source with the heaviest tail. We propose the ExtremeHunter algorithm, provide its analysis, and evaluate it empirically on synthetic and real-world experiments.
Extreme bandits
Alexandra Carpentier, Michal Valko
In many areas of medicine, security, and life sciences, we want to allocate limited resources to different sources in order to detect extreme values. In this paper, we study an efficient way to allocate these resources sequentially under limited feedback. While sequential design of experiments is well studied in bandit theory, the most commonly optimized property is the regret with respect to the maximum mean reward. However, in other problems such as network intrusion detection, we are interested in detecting the most extreme value output by the sources. Therefore, in our work we study extreme regret which measures the efficiency of an algorithm compared to the oracle policy selecting the source with the heaviest tail.
- Europe > United Kingdom > England > Cambridgeshire > Cambridge (0.14)
- Europe > France (0.04)
- Information Technology > Artificial Intelligence > Machine Learning (1.00)
- Information Technology > Communications > Networks (0.69)
- Information Technology > Data Science > Data Mining > Anomaly Detection (0.50)
- Information Technology > Data Science > Data Mining > Big Data (0.47)
Extreme bandits SequeL team University of Cambridge, UK INRIA Lille - Nord Europe, France
In many areas of medicine, security, and life sciences, we want to allocate limited resources to different sources in order to detect extreme values. In this paper, we study an efficient way to allocate these resources sequentially under limited feedback. While sequential design of experiments is well studied in bandit theory, the most commonly optimized property is the regret with respect to the maximum mean reward. However, in other problems such as network intrusion detection, we are interested in detecting the most extreme value output by the sources. Therefore, in our work we study extreme regret which measures the efficiency of an algorithm compared to the oracle policy selecting the source with the heaviest tail.
- Europe > United Kingdom > England > Cambridgeshire > Cambridge (0.86)
- Europe > France (0.40)
- Information Technology > Artificial Intelligence > Machine Learning (1.00)
- Information Technology > Communications > Networks (0.69)
- Information Technology > Data Science > Data Mining > Anomaly Detection (0.50)
- Information Technology > Data Science > Data Mining > Big Data (0.47)
Extreme Bandits using Robust Statistics
Bhatt, Sujay, Li, Ping, Samorodnitsky, Gennady
We consider a multi-armed bandit problem motivated by situations where only the extreme values, as opposed to expected values in the classical bandit setting, are of interest. We propose distribution free algorithms using robust statistics and characterize the statistical properties. We show that the provided algorithms achieve vanishing extremal regret under weaker conditions than existing algorithms. Performance of the algorithms is demonstrated for the finite-sample setting using numerical experiments. The results show superior performance of the proposed algorithms compared to the well known algorithms.
- North America > United States > New York > New York County > New York City (0.14)
- Europe > United Kingdom > England > Cambridgeshire > Cambridge (0.04)
- North America > United States > Washington > King County > Bellevue (0.04)
- (16 more...)
Extreme bandits
Carpentier, Alexandra, Valko, Michal
In many areas of medicine, security, and life sciences, we want to allocate limited resources to different sources in order to detect extreme values. In this paper, we study an efficient way to allocate these resources sequentially under limited feedback. While sequential design of experiments is well studied in bandit theory, the most commonly optimized property is the regret with respect to the maximum mean reward. However, in other problems such as network intrusion detection, we are interested in detecting the most extreme value output by the sources. Therefore, in our work we study extreme regret which measures the efficiency of an algorithm compared to the oracle policy selecting the source with the heaviest tail. We propose the ExtremeHunter algorithm, provide its analysis, and evaluate it empirically on synthetic and real-world experiments.
- Information Technology > Artificial Intelligence > Machine Learning (0.94)
- Information Technology > Data Science > Data Mining > Big Data (0.48)