Anomaly detection for data accountability of Mars telemetry data
作者：
Highlights：
• Δ-MADS is a new derivative-free optimization hybrid algorithm.
• It combines a model based global exploration with a direct search local intensification.
• It is tailored to optimize the configuration of a variational autoencoder for anomaly detection.
• Δ-MADS is fastest at finding the best solution among other blackbox optimization algorithms.
摘要
•Δ-MADS is a new derivative-free optimization hybrid algorithm.•It combines a model based global exploration with a direct search local intensification.•It is tailored to optimize the configuration of a variational autoencoder for anomaly detection.•Δ-MADS is fastest at finding the best solution among other blackbox optimization algorithms.
论文关键词：Anomaly detection,Variational autoencoder,Hyperparameter optimization,Architecture search,Derivative-free optimization
论文评审过程：Received 25 August 2020, Revised 25 July 2021, Accepted 8 October 2021, Available online 22 October 2021, Version of Record 28 October 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.116060
