Searching for Anomalies Over Composite Hypotheses

Hemo, Bar; Gafni, Tomer; Cohen, Kobi; Zhao, Qing

doi:10.1109/tsp.2020.2971438

Cited by 19 publications

(10 citation statements)

References 55 publications

(231 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Chernoff has established a randomized strategy, referred to as the Chernoff test which is asymptotically optimal in terms of minimizing the detection delay as the error probability diminishes. Variations and extensions of the problem and the Chernoff test were studied recently in [3], [4], [7]- [16]. When translating the sequential experimental design setting to the realm of the anomaly detection problem considered here, the experiments represent the set of K cells that can be probed at each time.…”

Section: A Connection With Sequential Design Of Experimentsmentioning

confidence: 99%

Searching for Anomalies with Multiple Plays under Delay and Switching Costs

Lambez

Cohen

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

The problem of searching for L anomalous processes among M processes is considered. At each time, the decision maker can observe a subset of K processes (i.e., multiple plays). The measurement drawn when observing a process follows one of two different distributions, depending whether the process is normal or abnormal. The goal is to design a policy that minimizes the Bayes risk which balances between the sample complexity, detection errors, and the switching cost associated with switching across processes. We develop a policy, dubbed consecutive controlled sensing (CCS), to achieve this goal. We prove theoretically that CCS is asymptotically optimal in terms of minimizing the Bayes risk as the sample complexity approaches infinity. Simulation results demonstrate strong performance of CCS in the finite regime as well.

show abstract

Section: A Connection With Sequential Design Of Experimentsmentioning

confidence: 99%

Searching for Anomalies with Multiple Plays under Delay and Switching Costs

Lambez

Cohen

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

show abstract

“…Anomaly detection problem is a well-studied research topic, and there are several active sensing schemes for anomaly detection designed to monitor the environments [3]- [6]. A popular approach for solving the anomaly detection problem is to use the active hypothesis testing framework [7], [8].…”

Section: A Related Literaturementioning

confidence: 99%

“…Here, the term P [s a = s ′ |s i = s], which depends on the statistical dependence between the processes, can be either assumed to be known or easily estimated from the training data 1 for every pair (i, j). Thus, (6) and (8) give the recursive update of σ(k).…”

Section: Estimation Algorithmmentioning

confidence: 99%

Scalable and Decentralized Algorithms for Anomaly Detection via Learning-Based Controlled Sensing

Joseph¹,

Chen²,

Gursoy³

et al. 2021

Preprint

View full text Add to dashboard Cite

We address the problem of sequentially selecting and observing processes from a given set to find the anomalies among them. The decision-maker observes a subset of the processes at any given time instant and obtains a noisy binary indicator of whether or not the corresponding process is anomalous. In this setting, we develop an anomaly detection algorithm that chooses the processes to be observed at a given time instant, decides when to stop taking observations, and declares the decision on anomalous processes. The objective of the detection algorithm is to identify the anomalies with an accuracy exceeding the desired value while minimizing the delay in decision making. We devise a centralized algorithm where the processes are jointly selected by a common agent as well as a decentralized algorithm where the decision of whether to select a process is made independently for each process. Our algorithms rely on a Markov decision process defined using the marginal probability of each process being normal or anomalous, conditioned on the observations. We implement the detection algorithms using the deep actor-critic reinforcement learning framework. Unlike prior work on this topic that has exponential complexity in the number of processes, our algorithms have computational and memory requirements that are both polynomial in the number of processes. We demonstrate the efficacy of these algorithms using numerical experiments by comparing them with state-ofthe-art methods.

show abstract

“…Intuitively, as more tests are performed, the learner becomes more certain about the true state of the population, which in turn leads to better choices of people to test. This approach was shown to achieve the asymptotic information theoretic bound in various hypothesis testing and dynamic search problems as the error probability approaches zero (see e.g., (15,25,27,28) and our previous work (4,5,12,17)). Here, we develop a controlled testing methodology to control the spread of the COVID-19 pandemic.…”

Section: Introductionmentioning

confidence: 95%

Suppressing the impact of the COVID-19 pandemic using controlled testing and isolation

Cohen

Leshem

2020

Preprint

Self Cite

View full text Add to dashboard Cite

The Corona virus disease 2019 has significantly affected lives of people around the world. Today, isolation policy is enforced by identifying infected individuals based on symptoms when these appear or by testing people and quarantining those who have been in close contact with infected people. In addition, many countries have imposed complete or partial lock-downs to control the spread of the disease. While this has resulted in some some success in slowing down the spread of the virus, lock-downs as well as widespread quarantine have devastating effects on the economy and social life. Thus, governments are urgently looking for efficient strategies to significantly relax lock-downs, while still controlling the spread of the virus. We argue that this can be done by using active feedback to control testing for infection by actively testing individuals with a high probability of being infected. We develop an active testing strategy to achieve this goal, and demonstrate that it would have tremendous success in controlling the spread of the virus on one million people, using 3, 000 tests per day. Our results show up to a 50% reduction in quarantine rate and morbidity rate in typical settings as compared to existing methods.

show abstract

Searching for Anomalies Over Composite Hypotheses

Cited by 19 publications

References 55 publications

Searching for Anomalies with Multiple Plays under Delay and Switching Costs

Searching for Anomalies with Multiple Plays under Delay and Switching Costs

Scalable and Decentralized Algorithms for Anomaly Detection via Learning-Based Controlled Sensing

Suppressing the impact of the COVID-19 pandemic using controlled testing and isolation

Contact Info

Product

Resources

About