Anomaly detection systems can be effectively constructed using machine learning methods. However, the baseline methods do not naturally support functionality for scenarios with constraints imposed by the environment and limited resources. Motivated by the lack of such approaches, we investigate scenarios where anomaly detection can be improved for these conditions. We develop and evaluate approaches for solving anomaly detection problems in case of online learning, limited reliability of input data and labels, and data acquisition constraints. Our approaches show improvements in detection and analysis of anomalies in data under these conditions.
«
Anomaly detection systems can be effectively constructed using machine learning methods. However, the baseline methods do not naturally support functionality for scenarios with constraints imposed by the environment and limited resources. Motivated by the lack of such approaches, we investigate scenarios where anomaly detection can be improved for these conditions. We develop and evaluate approaches for solving anomaly detection problems in case of online learning, limited reliability of input d...
»