ABMecik/Brain-Inspired-Cortical-Network: Brain-Inspired Cortical Network for Unsupervised Time-Series Anomaly Detection

Overview This release contains the stable, fully autonomous version of the Brain-Inspired Cortical Network anomaly detection framework. It is designed for unsupervised anomaly detection in chaotic, non-stationary time-series data. This implementation includes the full experimental pipeline used in the study on anomaly detection in chaotic systems. 🛠 Included Components Cortical Structure: spectron, tree (Brain-inspired clustering) Signal Encoding: wavelet (Multi-resolution wavelet packet transform) Pipeline: anomaly, anomaly_detector (Fully unsupervised learning) Data Tools: data_generator (Chaotic signal generator based on the Lorenz system) 📊 Datasets & Reproducibility Supports Synthetic Lorenz attractor data generation. Anomaly Injection: Spike, drift, drop, noise burst, and chaotic transition. Includes all scripts required to reproduce the experiments from the reference study. ✨ Key Features Fully Unsupervised: No labeled data required for training. Bio-inspired: Adaptive cortical structure with energy-based evolution. Robust: Designed specifically for chaotic, noisy, and non-stationary signals. Proven: Tested on Lorenz attractor signals and real-world KPI/AIOps datasets. 📖 Research Context Unsupervised Anomaly Detection in Chaotic Time Series Using Brain-Inspired Cortical Coding This framework is intended for research in predictive maintenance, AIOps, KPI monitoring, and brain-inspired computing.