QTS2D: Quantum-based Image Encoding of Time Series

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21460%2F25%3A00385078" target="_blank" >RIV/68407700:21460/25:00385078 - isvavai.cz</a>

Výsledek na webu

<a href="https://doi.org/10.1016/j.softx.2025.102327" target="_blank" >https://doi.org/10.1016/j.softx.2025.102327</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1016/j.softx.2025.102327" target="_blank" >10.1016/j.softx.2025.102327</a>

Jazyk výsledku

angličtina

Název v původním jazyce

QTS2D: Quantum-based Image Encoding of Time Series

Popis výsledku v původním jazyce

Despite growing interest in quantum machine learning, the application of quantum principles to foundational tasks such as time series feature engineering remains underexplored. This paper introduces QTS2D, a Python library that addresses this gap by adapting established time series-to-image techniques, including Gramian Angular Fields, Markov Transition Fields, Recurrence Plots, and Spectrograms, using quantum-inspired formulations. Unlike many works focused on execution speed or hardware advantage, our primary aim is to investigate the representational power of quantum-derived transformations for feature extraction. QTS2D leverages concepts such as amplitude encoding, quantum fidelity, and the Quantum Fourier Transform to generate rich, structured representations simulated on classical hardware. Empirical results on physiological signals demonstrate that quantum-based representations not only improve classification accuracy over classical counterparts, even within fully classical machine learning pipelines, but also exhibit enhanced class separability in UMAP embeddings and consistently lower Davies–Bouldin Index scores. The library provides a practical and extensible foundation for exploring quantum-enhanced feature engineering, with promising applications in health monitoring, anomaly detection, and beyond.

Název v anglickém jazyce

QTS2D: Quantum-based Image Encoding of Time Series

Popis výsledku anglicky

Despite growing interest in quantum machine learning, the application of quantum principles to foundational tasks such as time series feature engineering remains underexplored. This paper introduces QTS2D, a Python library that addresses this gap by adapting established time series-to-image techniques, including Gramian Angular Fields, Markov Transition Fields, Recurrence Plots, and Spectrograms, using quantum-inspired formulations. Unlike many works focused on execution speed or hardware advantage, our primary aim is to investigate the representational power of quantum-derived transformations for feature extraction. QTS2D leverages concepts such as amplitude encoding, quantum fidelity, and the Quantum Fourier Transform to generate rich, structured representations simulated on classical hardware. Empirical results on physiological signals demonstrate that quantum-based representations not only improve classification accuracy over classical counterparts, even within fully classical machine learning pipelines, but also exhibit enhanced class separability in UMAP embeddings and consistently lower Davies–Bouldin Index scores. The library provides a practical and extensible foundation for exploring quantum-enhanced feature engineering, with promising applications in health monitoring, anomaly detection, and beyond.

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

<a href="/cs/project/TQ16000045" target="_blank" >TQ16000045: Chytrá platforma založená na monitorování fyziologických dat a biofeedbacku ve VR pro rehabilitaci pacientů</a><br>

Návaznosti

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

Rok uplatnění

2025

Kód důvěrnosti údajů

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

Název periodika

SoftwareX

ISSN

2352-7110

e-ISSN

2352-7110

Svazek periodika

31

Číslo periodika v rámci svazku

September

Stát vydavatele periodika

AT - Rakouská republika

Počet stran výsledku

1000

Strana od-do

—

Kód UT WoS článku

001566568600002

EID výsledku v databázi Scopus

2-s2.0-105014819739