Quantum Machine Learning for Large-Scale Classical Datasets with Applications in Earth Observation

Aug 3, 2026

Upload by Rowan

Big satellite datasets support monitoring the Earth’s surface and earth observation challenges such as land cover changes, enabling decision-ready use of artificial intelligence. Quantum machine learning motivated by quantum algorithms is explored as a way to process data-driven tasks faster than conventional methods. Using benchmark satellite datasets, the work formulates three key issues: selecting suitable ML tasks for quantum computers, embedding large-dimensional satellite data into quantum input states, and combining supercomputer and quantum resources efficiently. The dissertation identifies ML and satellite data-driven tasks deployable on quantum hardware, proposes a two-level encoding strategy, and develops quantum learning approaches for both a quantum annealer and a noisy intermediate-scale quantum computer, achieving competitive supervised learning performance and estimating quantum resources for potential advantage.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages160

Reading Time6 h 43 min

AccessFree to download

Abstract

Kurzfassung

Dissertation Information

Thesis Submission Details

Reviewers and Oral Examination Date

Oath Declaration (Eidesstattliche Versicherung)

Dissertation an der Fakult¨at f¨ur Mathematik, Informatik und Statistik der Ludwig-Maximilians-Universit¨at M¨unchen Quantum Machine Learning for Large-Scale Classical Datasets with Applications in Earth Observation eingereicht von Soronzonbold Otgonbaatar am 25 . October 2024 Dissertation an der Fakult¨at f¨ur Mathematik, Informatik und Statistik der Ludwig-Maximilians-Universit¨at M¨unchen Quantum Machine Learning for Large-Scale Classical Datasets with Applications in Earth Observation eingereicht von Soronzonbold Otgonbaatar am 25 . October 2024 Erster Gutachter: Prof. Dieter Kranzlm¨uller, Ludwig-Maximilians-Universit¨at M¨unchen, Germany Zweiter Gutachter: Prof. Mihai Datcu, POLITEHNICA Bucures, ti, Romania Dritter Gutachter: Prof. Martin Werner, Technische Universit¨at M¨unchen, Germany Tag der m¨undlichen Pr¨ufung: 13 . Februar 2025 Eidesstattliche Versicherung (Siehe Promotionsordnung vom 12.07.11, § 8, Abs. 2 Pkt. .5.) Hiermit erkl¨are ich an Eidesstatt, dass die Dissertation von mir selbstst¨andig, ohne unerlaubte Beihilfe angefertigt ist. Otgonbaatar, Soronzonbold Name, Vorname M¨unchen, 25. October 2024 Soronzonbold (Ort, Datum) (Unterschrift Doktorand/in) Abstract Big satellite datasets are used to monitor the Earth’s surface or for Earth observation challenges like land cover changes. They enable leveraging artificial intelligence, including machine and deep learning, for detecting changes on the ground. Compared to conventional computational algorithms, artificial intelligence helps find better optimal solutions to Earth observation challenges involving big datasets. Decision-makers and policymakers already extensively use these solutions to make fast, safety-critical, and human-centered decisions. Quantum machine learning inspired by quantum algorithms promises to process some data-driven tasks faster than its conventional counterparts. We aim to use various benchmark satellite datasets to develop and benchmark quantum machine learning approaches with traditional artificial intelligence models. There are three main issues for processing quantum machine learning on benchmark satellite datasets: 1 . which machine learning task for big satellite datasets and which satellite data-driven task can be efficiently and effectively processed on a quantum computer? 2 . how to embed large-dimensional satellite data points in input quantum states, and 3 . how to profit from both supercomputers and quantum computers. To find a scientific answer to these three questions, we examine and identify both machine learning and satellite data-driven tasks that can be deployed on a quantum computer, otherwise inherently intractable. We then propose the encoding strategy of classical problems involving big satellite datasets in a quantum computer, named two-level encoding. Further, we design and investigate quantum machine learning approaches for a quantum annealer anda noisy intermediate-scale quantum computer for supervised learning tasks. For supervised learning tasks, the performance of our quantum machine learning approaches is already competitive (and even better in some instances) compared to the ones of their classical counterparts. Additionally, we estimate the quantum resource required to gain an advantage over a supercomputer and profit from a supercomputer and a quantum computer. Doing so gives us insights into a future fault-tolerant quantum computer for tackling practical computational problems. Kurzfassung Man benutzt große Datens¨atze, um die Oberfl¨ache der Erde zu ¨uberwachen oder f¨ur Erdbeobachtungsaufgaben wie Ver¨anderungen der Bodenbedeckung. Diese Datens¨atze erlaubenes, k¨unstliche Intelligenz einschließlich maschinellem Lernen und Deep Learning einzusetzen, um Ver¨anderungen auf der Erdoberfl¨ache zu entdecken. Im Vergleich zu konventionellen Rechenverfahren hilft k¨unstliche Intelligenz, bessere optimale L¨osungen f¨ur Erdbeobachtungsaufgaben mit gr

FAQs

What problems does the dissertation target for quantum machine learning on satellite benchmark datasets?

It focuses on three issues: which ML and satellite data-driven tasks can be efficiently run on a quantum computer, how to embed large-dimensional satellite data into input quantum states, and how to profit from combining supercomputers and quantum computers.

What encoding strategy is proposed for mapping classical large-scale satellite data to quantum computers?

The dissertation proposes a two-level encoding strategy to represent classical problems involving big satellite datasets within a quantum computer.

How does the dissertation evaluate whether quantum approaches are competitive with classical models?

It designs and investigates quantum learning approaches for supervised learning using a quantum annealer and a noisy intermediate-scale quantum computer, showing competitive performance in comparison to classical counterparts and estimating required quantum resources for potential advantage.

Aug 3, 2026

Upload by Rowan