Potentials of Learning Entropy for Sub-Nyquist and Sub-Noise Anomaly Detection

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21220%2F24%3A00385305" target="_blank" >RIV/68407700:21220/24:00385305 - isvavai.cz</a>

Výsledek na webu

<a href="https://ieeexplore.ieee.org/document/10805128" target="_blank" >https://ieeexplore.ieee.org/document/10805128</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Potentials of Learning Entropy for Sub-Nyquist and Sub-Noise Anomaly Detection

Popis výsledku v původním jazyce

Recently, we have noted the intriguing ability of Learning Entropy (LE) with Higher-Order Neural Units (HONUs) to detect abrupt frequency changes in both sub-Nyquist signals and pure noise signals. The concept of the LE and its fundamental element, the Learning Information (LI), is revised and extended. The LI is presented as the logical counterpart of Shannon's probabilistic information and the LE as its aggregation, e.g. as an average. The approach emphasizes the computational efficiency and analyzability of HONUs with gradient learning, which implies minimal computational and energy requirements for real-time computations. For pure noise, we show that bandpass filtering and principal components improve LE performance to immediately detect frequency changes below the noise level. By demonstrating the ability of LE with linear units to detect anomalies in undersampled signals and also in pure noise, we show the potential of LE for industrial applications and perhaps also for research related to future digital communication systems or advanced analysis of data from distributed sensors in optical transmission infrastructures

Název v anglickém jazyce

Potentials of Learning Entropy for Sub-Nyquist and Sub-Noise Anomaly Detection

Popis výsledku anglicky

Recently, we have noted the intriguing ability of Learning Entropy (LE) with Higher-Order Neural Units (HONUs) to detect abrupt frequency changes in both sub-Nyquist signals and pure noise signals. The concept of the LE and its fundamental element, the Learning Information (LI), is revised and extended. The LI is presented as the logical counterpart of Shannon's probabilistic information and the LE as its aggregation, e.g. as an average. The approach emphasizes the computational efficiency and analyzability of HONUs with gradient learning, which implies minimal computational and energy requirements for real-time computations. For pure noise, we show that bandpass filtering and principal components improve LE performance to immediately detect frequency changes below the noise level. By demonstrating the ability of LE with linear units to detect anomalies in undersampled signals and also in pure noise, we show the potential of LE for industrial applications and perhaps also for research related to future digital communication systems or advanced analysis of data from distributed sensors in optical transmission infrastructures

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

21100 - Other engineering and technologies

Projekt

<a href="/cs/project/TN02000010" target="_blank" >TN02000010: Národní centrum kompetence Mechatroniky a chytrých technologií pro strojírenství</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2024

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

2024 SICE Festival with Annual Conference, SICE FES 2024

ISBN

979-8-3315-4446-1

ISSN

—

e-ISSN

—

Počet stran výsledku

7

Strana od-do

584-590

Název nakladatele

Institute of Electrical and Electronics Engineers Inc.

Místo vydání

—

Místo konání akce

Kochi

Datum konání akce

27. 8. 2024

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

001424958800050