What question did this study set out to answer?

Develop a mathematically guaranteed, unsupervised online anomaly detection algorithm tailored for complex sequential data streams.

August 27, 2020IEEE Transactions on Neural Networks and Learning Systems

Online Anomaly Detection With Bandwidth Optimized Hierarchical Kernel Density Estimators

View Full Paper

Ask AI

Bookmark

Develop a mathematically guaranteed, unsupervised online anomaly detection algorithm tailored for complex sequential data streams.
Constructed a hierarchical partitioning tree where each space partition trains an online kernel density estimator with dynamic, region-specific bandwidths.
Combined regional estimators sequentially to yield a final density estimate evaluated against a data-adaptive detection threshold.
Proved theoretically that the framework learns optimal partitions and time-varying bandwidths with computational complexity scaling linearly with tree depth and data length.
Achieved significant improvements in anomaly detection accuracy compared to existing state-of-the-art techniques.

Discussion

No takes yet. Share an insight, caveat, or question.

Implication

Algorithmic study demonstrates improved anomaly detection accuracy in complex sequential data, indicating scalable real-time monitoring with linear computational complexity.

Cite This Study

Kerpiççi et al. (2020) studied this question.

synapsesocial.com/papers/6a0e9b213903c9222ec84ada https://doi.org/10.1109/tnnls.2020.3017675

View Full Paper

Ask AI

Bookmark

Implication

Algorithmic study demonstrates improved anomaly detection accuracy in complex sequential data, indicating scalable real-time monitoring with linear computational complexity.

Online Anomaly Detection With Bandwidth Optimized Hierarchical Kernel Density Estimators

Key Points

Discussion

Implication

Cite This Study

Online Anomaly Detection With Bandwidth Optimized Hierarchical Kernel Density Estimators

Key Points

Discussion

Implication

Cite This Study

Also Consider

Also Consider