Article

A Training-Free Channel-Intervention Variance Score for One-Class Multivariate Time-Series Anomaly Detection

2026-05-19

Abstract excerpt

<title>Abstract</title> <p>One-class scoring on a deep encoder is the dominant approach to multivariate time-series anomaly detection (MTS-AD). Two engineering constants of recent methods are (i) a single scalar score s(x) produced by a learned head and(ii) a training loss that pushes normal windows toward zero. We study a training-free inference-time enhancement: apply K random channel interventions to each test...

Topics

Open a Topic to create a Post that cites this publication.

Identifiers and source

Literature Corpus work: 99cdfe7c-2d34-5b27-8297-36b109ef1665
DOI: 10.21203/rs.3.rs-9578810/v1

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

A Training-Free Channel-Intervention Variance Score for One-Class Multivariate Time-Series Anomaly DetectionDOI 10.21203/rs.3.rs-9578810/v1Read

A Generalized Iterative Polar Transform for Readable Time-Series ImagingSemantic proximity Read

Efficient Uncertainty Aware Human Activity Recognition On Microcontrollers Using Hyperdimensional Computing And Conformal PredictionSemantic proximity Read

<p style="-qt-block-indent: 0; text-indent: 0px; margin: 0px;">ASBR-CL: Stage-Aware Balanced Replay for Memory-Limited Continual Fault DiagnosisSemantic proximity Read

A foundation model for multivariate time series forecastingSemantic proximity Read

Multiple Instance Learning Framework can Facilitate Explainability in Murmur DetectionSemantic proximity Read

A Temporal Difference and Cross-variate Fusion Network for Multivariate Time Series ClassificationSemantic proximity Read

STAR-RL: Stealth-Aware Targeted Adversarial Attack on Multimodal Sensors in Human Activity Recognition via Reinforcement LearningSemantic proximity Read

Agentic AI and LLM-Driven Framework for Contextual Anomaly DetectionSemantic proximity Read

Improved Human Activity Recognition Through Controllable GAN-Generated Synthetic Data and Large Language Models for ClassificationSemantic proximity Read

ATAT: Astronomical Transformer for time series And Tabular dataSemantic proximity Read

Improved Human Activity Recognition Through Controllable GAN-Generated Synthetic Data and Large Language Models for ClassificationSemantic proximity Read

HIFN-Mamba: Hybrid Information Flow Networks with Selective State-Space Sequence Modeling for Intrusion Detection in the Internet of Medical ThingsSemantic proximity Read