Abstract:Support vector machines (SVMs) are widely used machine learning models (e.g., in remote sensing), with formulations for both classification and regression tasks. In the last years, with the advent of working quantum annealers, hybrid SVM models characterised by quantum training and classical execution have been introduced. These models have demonstrated comparable performance to their classical counterparts. However, they are limited in the training set size due to the restricted connectivity of the current quantum annealers. Hence, to take advantage of large datasets (like those related to Earth observation), a strategy is required. In the classical domain, local SVMs, namely, SVMs trained on the data samples selected by a k-nearest neighbors model, have already proven successful. Here, the local application of quantum-trained SVM models is proposed and empirically assessed. In particular, this approach allows overcoming the constraints on the training set size of the quantum-trained models while enhancing their performance. In practice, the FaLK-SVM method, designed for efficient local SVMs, has been combined with quantum-trained SVM models for binary and multiclass classification. In addition, for comparison, FaLK-SVM has been interfaced for the first time with a classical single-step multiclass SVM model (CS SVM). Concerning the empirical evaluation, D-Wave's quantum annealers and real-world datasets taken from the remote sensing domain have been employed. The results have shown the effectiveness and scalability of the proposed approach, but also its practical applicability in a real-world large-scale scenario.

What problem does this paper attempt to address?

### Main Problem Addressed by the Paper The paper primarily addresses the limitations faced by quantum-trained Support Vector Machines (SVM) when handling large-scale datasets. Specifically, due to the limited connectivity of current quantum annealers, the quantum-trained SVM model has constraints regarding the size of the training set. To solve this issue, the paper proposes a novel approach that combines the concept of local Support Vector Machines with quantum training. ### Overview of the Solution The paper proposes the following key points to address the aforementioned problem: 1. **Locally Applied Quantum-Trained SVM Model**: By leveraging the success of local SVMs in the classical domain (i.e., SVMs trained on data samples selected by the k-nearest neighbors algorithm), this concept is extended to quantum-trained SVM models. This approach can overcome the limitations of quantum-trained models concerning the size of the training set and improve model performance. 2. **Integration with FaLK-SVM Method**: The paper employs the Fast Local Kernel Support Vector Machine (FaLK-SVM) method, an efficient local SVM approach that reduces execution time using a cover tree data structure. This method is applied to quantum-trained binary classification and support vector machines (QBSVM and QMSVM) and is also compared for the first time with the classical single-step multi-class SVM (CS SVM). 3. **Evaluation and Validation**: To validate the effectiveness and scalability of the proposed method, the paper conducts experiments using D-Wave's quantum annealer and real-world datasets from the remote sensing domain. The results show that the proposed localized quantum-trained SVM method is not only effective but also highly applicable in large-scale real-world scenarios. ### Experiments and Results - **Binary Classification Tasks**: All considered binary classification methods were evaluated and compared. - **Multi-Class Classification Tasks**: Similarly, performance evaluations and comparisons were conducted for multi-class classification tasks. - **Performance Scalability Analysis**: The scalability of performance for local and global classical SVM methods (including both binary and multi-class) was analyzed. In summary, this paper aims to overcome the limitations of quantum-trained SVM models in handling large-scale datasets by introducing a novel approach and demonstrates the effectiveness and practicality of this method.

Local Binary and Multiclass SVMs Trained on a Quantum Annealer

A Single-Step Multiclass SVM based on Quantum Annealing for Remote Sensing Data Classification

A Quantum Extension of SVM-perf for Training Nonlinear SVMs in Almost Linear Time

Quantum-Inspired Support Vector Machine

Quantum-Assisted Support Vector Regression for Detecting Facial Landmarks

Variational Quantum Linear Solver enhanced Quantum Support Vector Machine

Detecting Clouds in Multispectral Satellite Images Using Quantum-Kernel Support Vector Machines

Machine Learning in the Quantum Age: Quantum vs. Classical Support Vector Machines

Towards Efficient Quantum Anomaly Detection: One-Class SVMs using Variable Subsampling and Randomized Measurements

An enhanced approach for predicting air pollution using quantum support vector machine

Enhancing Quantum Support Vector Machines through Variational Kernel Training

Application of quantum machine learning using the quantum kernel algorithm on high energy physics analysis at the LHC

Enhanced feature encoding and classification on distributed quantum hardware

Quantum Fuzzy Support Vector Machine for Binary Classification

A quantum-enhanced support vector machine for galaxy classification

Leveraging Data Locality in Quantum Convolutional Classifiers

Unlocking quantum SVM potential: optimal feature map generation and feature selection

Adaptive Learning for Quantum Linear Regression

Quantum support vector machine without iteration

Cloud Detection in Multispectral Satellite Images Using Support Vector Machines With Quantum Kernels

Quantum support vector machines for classification and regression on a trapped-ion quantum computer