QTS2D: Quantum-Based Image Encoding of Time Series

QTS2D is a Python library for encoding time series data into image representations using quantum computing principles. These image-based representations can then be used as input for various machine learning models, particularly Convolutional Neural Networks (CNNs), for tasks like time series classification or forecasting.

Description

The library provides several quantum-inspired transformers that convert 1D time series into 2D matrices (images). These methods are analogous to classical time series imaging techniques but leverage quantum computations or quantum-inspired mathematical formulations.

Currently implemented transformers:

• QGAF: Quantum Gramian Angular Field
• QRP: Quantum Recurrence Plot
• QSG: Quantum Spectrogram
• QMTF: Quantum Markov Transition Field

Installation

You can install library using pip:

pip install qts2d

or using pip+git for the latest version of the code:

pip install git+

The library requires Python >= 3.10 and the following main dependencies:

• numpy (>=2.2.5)
• scikit-learn (>=1.6.1)
• qiskit (==1.4.2)
• qiskit-aer (==0.17.0)
• qiskit-machine-learning (==0.8.2)
• pyts (>=0.13.0)

Quick Example

Each transformer follows the scikit-learn Estimator and TransformerMixin interface.

import numpy as np
from qts2d.encoding import QGAF

# Example time series data (n_samples, n_timestamps)
X = np.random.rand(1, 1000)

# Using Quantum Gramian Angular Field (QGAF)
qgaf = QGAF(image_size=32, method='summation', computation_method='hadamard', scaling='minmax')
X_qgaf = qgaf.fit_transform(X)
print(f"QGAF output shape: {X_qgaf.shape}")

For more detailed examples, please refer to the examples/ directory for each specific transformer. To run the examples Matplotlib is required.

Contributing

We welcome everyone to contribute to qts2d! The library is still evolving and far from perfect, so your help is especially valuable. Please feel free to submit a pull request or open an issue. Detailed contribution guidelines will be shared soon along with the documentation. Stay tuned!

Citation

Coming soon!

License

This project is licensed under the BSD 3-Clause License - see the LICENSE.txt file for details.

References

• [1] Xu, Z., Wang, Y., Feng, X., Wang, Y., Li, Y., & Lin, H. Quantum-enhanced forecasting: Leveraging quantum gramian angular field and CNNs for stock return predictions. Finance Research Letters (2024)

• [2] Z. Wang and T. Oates, "Encoding Time Series as Images for Visual Inspection and Classification Using Tiled Convolutional Neural Networks." AAAI Workshop (2015).

• [3] J.-P Eckmann, S. Oliffson Kamphorst and D Ruelle, "Recurrence Plots of Dynamical Systems". Europhysics Letters (1987).