Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Transfer Learning for Large-Scale Image Classification Using Annealing-based Quantum Boltzmann Machines

Daniëlle Schuman, Leo Sünkel, Philipp Altmann, Jonas Stein, Christoph Roch, Thomas Gabor, Claudia Linnhoff-Popien

TL;DR

The paper addresses transfer learning for large-scale image classification using annealing-based Quantum Boltzmann Machines (QBMs) within a hybrid quantum-classical pipeline. It replaces gate-based variational circuits with an annealing-based QBM, simulated via Simulated Annealing, in a SEQUENT-like transfer-learning framework that uses a ResNet-18 feature extractor and a 64-dimensional compression layer before quantum classification. On the COVID-CT-MD dataset, the annealing-based QBM approach delivers higher test accuracy and AUC-ROC than a similarly sized classical baseline, and reaches these performance levels with fewer training epochs, albeit with longer wall-clock time in SA. The results suggest potential advantages of quantum-inspired learning in large-scale imaging, while also underscoring the need for hardware experiments, validation strategies, and broader datasets to assess practical quantum advantage.

Abstract

Quantum Transfer Learning (QTL) recently gained popularity as a hybrid quantum-classical approach for image classification tasks by efficiently combining the feature extraction capabilities of large Convolutional Neural Networks with the potential benefits of Quantum Machine Learning (QML). Existing approaches, however, only utilize gate-based Variational Quantum Circuits for the quantum part of these procedures. In this work we present an approach to employ Quantum Annealing (QA) in QTL-based image classification. Specifically, we propose using annealing-based Quantum Boltzmann Machines as part of a hybrid quantum-classical pipeline to learn the classification of real-world, large-scale data such as medical images through supervised training. We demonstrate our approach by applying it to the three-class COVID-CT-MD dataset, a collection of lung Computed Tomography (CT) scan slices. Using Simulated Annealing as a stand-in for actual QA, we compare our method to classical transfer learning, using a neural network of the same order of magnitude, to display its improved classification performance. We find that our approach consistently outperforms its classical baseline in terms of test accuracy and AUC-ROC-Score and needs less training epochs to do this.

Towards Transfer Learning for Large-Scale Image Classification Using Annealing-based Quantum Boltzmann Machines

TL;DR

The paper addresses transfer learning for large-scale image classification using annealing-based Quantum Boltzmann Machines (QBMs) within a hybrid quantum-classical pipeline. It replaces gate-based variational circuits with an annealing-based QBM, simulated via Simulated Annealing, in a SEQUENT-like transfer-learning framework that uses a ResNet-18 feature extractor and a 64-dimensional compression layer before quantum classification. On the COVID-CT-MD dataset, the annealing-based QBM approach delivers higher test accuracy and AUC-ROC than a similarly sized classical baseline, and reaches these performance levels with fewer training epochs, albeit with longer wall-clock time in SA. The results suggest potential advantages of quantum-inspired learning in large-scale imaging, while also underscoring the need for hardware experiments, validation strategies, and broader datasets to assess practical quantum advantage.

Abstract

Quantum Transfer Learning (QTL) recently gained popularity as a hybrid quantum-classical approach for image classification tasks by efficiently combining the feature extraction capabilities of large Convolutional Neural Networks with the potential benefits of Quantum Machine Learning (QML). Existing approaches, however, only utilize gate-based Variational Quantum Circuits for the quantum part of these procedures. In this work we present an approach to employ Quantum Annealing (QA) in QTL-based image classification. Specifically, we propose using annealing-based Quantum Boltzmann Machines as part of a hybrid quantum-classical pipeline to learn the classification of real-world, large-scale data such as medical images through supervised training. We demonstrate our approach by applying it to the three-class COVID-CT-MD dataset, a collection of lung Computed Tomography (CT) scan slices. Using Simulated Annealing as a stand-in for actual QA, we compare our method to classical transfer learning, using a neural network of the same order of magnitude, to display its improved classification performance. We find that our approach consistently outperforms its classical baseline in terms of test accuracy and AUC-ROC-Score and needs less training epochs to do this.
Paper Structure (9 sections, 3 equations, 3 figures, 1 table)

This paper contains 9 sections, 3 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. (Quantum) Boltzmann Machines
  4. (Quantum) Transfer Learning
  5. Methodology
  6. Experiments
  7. Dataset
  8. Experimental setup and results
  9. Discussion and Conclusion

Figures (3)

  • Figure 1: Our hybrid QTL pipeline. The QBM is the only part of this pipeline that can be run on a Quantum Annealer (even though we used SA in our experiments), the rest is executed on a classical computer. (CT image from COVID-CT-MD)
  • Figure 2: Accuracies and AUC-ROC-Scores of SA and classical models with the best identified hyperparameter settings over the course of the training. The solid line shows the average over 10 seeds, the transparent area around it the standard deviation.
  • Figure 3: Accuracies and AUC-ROC-Scores of SA and classical models with the best identified hyperparameter settings applied to the test set (averaged over 10 test seeds each).