arXiv subjects

Ireneusz Jabłoński

Publications and source records attributed to Ireneusz Jabłoński.

4 recordsLinked to original sources

arXiv2026

Real-time and adaptive anomaly detection algorithm for cyclostationary models

This article introduces PeriodicCALM, an effective real-time anomaly detection framework designed for cyclostationary data streams. While classical cyclostationary processes feature periodically time-varying statistical properties, real-world signals often contain recurring impulsive components that conceal abnormal behavior. Existing real-time methods for struggle with these dynamics, frequently misinterpreting phase-dependent variability as non-cyclic anomalies and causing excessive false alarms. To address this, PeriodicCALM incorporates cycle-dependent variability to systematically ignore regular cyclic impulses while accurately isolating genuine anomalies. Operating in real time with continuous retraining capabilities, the method adapts dynamically to evolving signal characteristics. Comparative evaluations against the baseline CALM framework using simulated data demonstrate significant improvements in detection accuracy and training efficiency, alongside a reduction in prediction latency. Furthermore, the practical utility of PeriodicCALM is validated on real-world vibration signals collected from a compressor monitoring system.

stat.ME↗

arXiv2026

Real-time anomaly detection in base station testbeds via scalable kernel density estimation framework

Large-scale testing infrastructures are critical for validating telecommunication systems, yet their growing complexity makes efficient resource utilization and anomaly detection increasingly challenging. In reservation-based testbed environments, errors in resource allocation or preparation often manifest as abrupt spikes or regime changes in time-based metrics. This paper proposes a scalable, unsupervised framework for real-time anomaly detection in such environments. We introduce CALM (Continuous Anomaly Localization for univariate and Multivariate data), a nonparametric method based on kernel density estimation and bootstrap-based thresholding, designed for anomaly detection at the individual testbed level. To address system-wide visibility, we further propose AggCALM, an aggregation framework that consolidates local anomaly signals across multiple testbeds to detect statistically significant global anomalies while mitigating alarm fatigue. The methodology is evaluated using simulated multivariate data and real-world data from a large-scale base station testing platform. Results demonstrate that the proposed framework enables timely, flexible, and accurate anomaly detection without requiring labeled data, supporting reliable operation of complex test environments. Although the proposed methodology is presented within the context of a telecommunication testing labs, it can be effectively used to other applications, such as condition monitoring, where anomaly detection serves as a pivotal pre-processing step for diagnostic signals.

stat.AP↗

arXiv2025

Robust variance estimators in application to segmentation of measurement data distorted by impulsive and non-Gaussian noise

The paper algorithmizes the problem of regime change point identification for data measured in a system exhibiting impulsive behaviors. This is a fundamental challenge for annotation of measurement data relevant, e.g., for designing data-driven autonomous systems. The contribution consists in the formulation of an offline robust methodology based on the classical approach for structural break detection. The problem of data segmentation is considered in the context of scale change, which physically can be translated into the occurrence of a critical event that reorganizes the system structure. The main advantage of our approach is that it does not require the existence of a variance of the data distribution. The efficiency has been evaluated for simulated data from two distributions and for real-world datasets measured in financial, mechanical, and medical systems. Simulation studies show that in the most challenging case, the error in estimating regime change is 20 times smaller for robust approach compared to the classical one.

stat.ME↗

arXiv2024

Modeling configuration-performance relation in a mobile network: a data-driven approach

Mobile network performance modeling typically assumes either a fixed cell's configuration or only considers a limited number of parameters. This prohibits the exploration of multidimensional, diverse configuration space for, e.g., optimization purposes. This paper presents a method for performance predictions based on a network cell's configuration and network conditions, which utilizes neural network architecture. We evaluate the idea by extensive experiments, with data from more than 50,000 5G cells. The assessment included a comparison of the proposed method against models developed for fixed configuration. Results show that combined configuration-performance modeling outperforms single-configuration models and allows for performance prediction of unknown configurations, i.e., it is not used for model training. A substantially lower mean absolute error was achieved (0.25 vs. 0.45 for fixed-configuration MLP-based models).

cs.NI↗