Deterministic Tensor Network Classifiers

TL;DR

This paper introduces deterministic tensor network classifiers that efficiently extract features from images, refine predictions with quantum stacking, and are suitable for implementation on NISQ devices, achieving good accuracy without variational training.

Contribution

The paper proposes a novel deterministic tensor network approach for feature extraction and classification, including a quantum stacking method compatible with NISQ devices, and demonstrates competitive performance on standard datasets.

Findings

Achieves good accuracy on MNIST and fashion MNIST without variational training.

Uses a logarithmic number of qubits for feature encoding.

Introduces a quantum stacking method for performance refinement.

Abstract

We present tensor networks for feature extraction and refinement of classifier performance. These networks can be initialised deterministically and have the potential for implementation on near-term intermediate-scale quantum (NISQ) devices. Feature extraction proceeds through a direct combination and compression of images amplitude-encoded over just $\log N_{\text{pixels}}$ qubits. Performance is refined using `Quantum Stacking', a deterministic method that can be applied to the predictions of any classifier regardless of structure, and implemented on NISQ devices using data re-uploading. These procedures are applied to a tensor network encoding of data, and benchmarked against the 10 class MNIST and fashion MNIST datasets. Good training and test accuracy are achieved without any variational training.