Vacuum Spiker: A Spiking Neural Network-Based Model for Efficient Anomaly Detection in Time Series

TL;DR

The paper tackles energy-efficient anomaly detection in time series on resource-constrained devices by introducing Vacuum Spiker, a two-layer spiking neural network that processes data in real time using Interval Coding and a novel STDP rule. It trains exclusively on normal data to suppress routine patterns, and flags anomalies via a spike-count threshold in the processing layer, achieving single-step, online inference. Across 45 public datasets and a real-world solar inverter case, Vacuum Spiker shows competitive detection performance while dramatically reducing energy consumption compared to deep learning baselines. This work demonstrates the potential of Spiking Neural Networks for Green AI in edge-time-series anomaly detection and motivates future exploration of online learning and architectural variants.

Abstract

Anomaly detection is a key task across domains such as industry, healthcare, and cybersecurity. Many real-world anomaly detection problems involve analyzing multiple features over time, making time series analysis a natural approach for such problems. While deep learning models have achieved strong performance in this field, their trend to exhibit high energy consumption limits their deployment in resource-constrained environments such as IoT devices, edge computing platforms, and wearables. To address this challenge, this paper introduces the \textit{Vacuum Spiker algorithm}, a novel Spiking Neural Network-based method for anomaly detection in time series. It incorporates a new detection criterion that relies on global changes in neural activity rather than reconstruction or prediction error. It is trained using Spike Time-Dependent Plasticity in a novel way, intended to induce changes in neural activity when anomalies occur. A new efficient encoding scheme is also proposed, which discretizes the input space into non-overlapping intervals, assigning each to a single neuron. This strategy encodes information with a single spike per time step, improving energy efficiency compared to conventional encoding methods. Experimental results on publicly available datasets show that the proposed algorithm achieves competitive performance while significantly reducing energy consumption, compared to a wide set of deep learning and machine learning baselines. Furthermore, its practical utility is validated in a real-world case study, where the model successfully identifies power curtailment events in a solar inverter. These results highlight its potential for sustainable and efficient anomaly detection.

Figures (5)

• Figure 1: Vacuum Spiker algorithm process. First, an univariate time series is encoded using Interval Coding. The predefined interval in which a value falls determines which neuron is activated. Information then propagates through the $I \rightarrow R$ connection (black lines between $I$ and $R$) and, if present, through $R \rightarrow R$ connection (gray lines within layer $R$). If the number of firing neurons in layer $R$ exceeds a predefined threshold, an alert is generated.
• Figure 2: Spike counts generated by the Vacuum Spiker algorithm (blue). Green dots indicate the time series values, while orange lines mark intervals with labelled anomalies. The red lines denote the anomaly detection thresholds.
• Figure 3: Comparison of power output (kW) by the solar inverter with and without power curtailment.
• Figure 4: Power output (kW) of a solar inverter during periods affected by communication errors.
• Figure 5: Power production (in kW) and anomaly detection for the studied solar inverter under different operational conditions. In all figures, the power output is shown in blue and corresponds to the left axis, while the number of spikes generated over time is shown in orange and corresponds to the right axis. Spikes are considered alerts. The scenarios depicted include normal operation, communication errors, and varying degrees of power curtailment across different months.