The authors present a hybrid quantum-classical architecture designed to address time reparameterization invariance in time series analysis. The system integrates quantum neural networks with path signatures, using feature layers that compute signature kernels between input paths via classical or quantum variational linear solvers (VQLS).
- Feature layers calculate signature kernels for classification tasks.
- A Quantum Convolutional Neural Network (QCNN) performs downstream learning.
- Experiments evaluate the architecture on binary classification of handwritten digits.
- The study analyzes computational limitations associated with the VQLS component.
The work demonstrates the potential advantages of implementing path signature kernel layers within quantum circuits for time series data.