Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multiple Embeddings for Quantum Machine Learning

Siyu Han, Lihan Jia, Lanzhe Guo

TL;DR

The paper targets the poor generalization of quantum classifiers arising from reliance on a single data embedding. It introduces MEDQ, a Multi-Encoding Data reuploading framework that stacks and reuploads data through multiple embedding schemes (Rot, QAOA, Angle) to create richer feature representations without increasing qubit count. Empirical results on linear separable benchmarks and MNIST with PCA demonstrate that MEDQ achieves superior generalization and often requires fewer layers than single-embedding baselines. This approach broadens the practical potential of quantum classifiers by leveraging diverse encodings to better capture dataset structure, with implications for hardware-accelerated quantum learning.

Abstract

This work focuses on the limitations about the insufficient fitting capability of current quantum machine learning methods, which results from the over-reliance on a single data embedding strategy. We propose a novel quantum machine learning framework that integrates multiple quantum data embedding strategies, allowing the model to fully exploit the diversity of quantum computing when processing various datasets. Experimental results validate the effectiveness of the proposed framework, demonstrating significant improvements over existing state-of-the-art methods and achieving superior performance in practical applications.

Multiple Embeddings for Quantum Machine Learning

TL;DR

The paper targets the poor generalization of quantum classifiers arising from reliance on a single data embedding. It introduces MEDQ, a Multi-Encoding Data reuploading framework that stacks and reuploads data through multiple embedding schemes (Rot, QAOA, Angle) to create richer feature representations without increasing qubit count. Empirical results on linear separable benchmarks and MNIST with PCA demonstrate that MEDQ achieves superior generalization and often requires fewer layers than single-embedding baselines. This approach broadens the practical potential of quantum classifiers by leveraging diverse encodings to better capture dataset structure, with implications for hardware-accelerated quantum learning.

Abstract

This work focuses on the limitations about the insufficient fitting capability of current quantum machine learning methods, which results from the over-reliance on a single data embedding strategy. We propose a novel quantum machine learning framework that integrates multiple quantum data embedding strategies, allowing the model to fully exploit the diversity of quantum computing when processing various datasets. Experimental results validate the effectiveness of the proposed framework, demonstrating significant improvements over existing state-of-the-art methods and achieving superior performance in practical applications.

Paper Structure

This paper contains 15 sections, 4 equations, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Preliminary
  4. Data Reuploading Classifier
  5. Circuit Architecture
  6. Data Embeddings
  7. Reuploading Process
  8. Measurement Process
  9. Training Process
  10. Experiments
  11. Linear Separable
  12. MNIST
  13. Conclusion & Discussion
  14. Discussion
  15. Conclusion

Figures (5)

  • Figure 1: Visual illustration of Data Reuploading Model
  • Figure 2: An illustration of classic machine learning: the input data is initially encoded linearly, and non-linear methods are applied later to enhance the model’s generalization capability. The yellow arrow means nonlinear methods such as kernel methods or activation functions.
  • Figure 3: An illustration of MEDQ: Left shows the original quantum machine learning model where the input data is encoded non-linearly, while the right figure shows after integrating multiple embeddings, the MEDQ earned a better generalization capability.
  • Figure 4: Visual illustration of Data Reuploading Model
  • Figure 5: Representation of the Bloch sphere, each point representing a class vector and single-qubit classifier will be trained to distributed the data points in one of these vertices