Image Classification by Throwing Quantum Kitchen Sinks at Tensor Networks

TL;DR

This paper introduces a hybrid quantum circuit combining quantum kitchen sinks with tensor networks for image classification, demonstrating that feature optimization significantly enhances performance and achieves near-term feasible quantum advantage.

Contribution

The paper proposes a novel circuit ansatz combining quantum kitchen sinks with tensor networks and shows that feature optimization leads to state-of-the-art results in quantum image classification.

Findings

Combining quantum kitchen sinks with tensor networks alone does not improve performance.

Feature optimization greatly enhances classification accuracy.

The optimized circuits are shallow and use few qubits, suitable for near-term quantum devices.

Abstract

Several variational quantum circuit approaches to machine learning have been proposed in recent years, with one promising class of variational algorithms involving tensor networks operating on states resulting from local feature maps. In contrast, a random feature approach known as quantum kitchen sinks provides comparable performance, but leverages non-local feature maps. Here we combine these two approaches by proposing a new circuit ansatz where a tree tensor network coherently processes the non-local feature maps of quantum kitchen sinks, and we run numerical experiments to empirically evaluate the performance of the new ansatz on image classification. From the perspective of classification performance, we find that simply combining quantum kitchen sinks with tensor networks yields no qualitative improvements. However, the addition of feature optimization greatly boosts performance, leading to state-of-the-art quantum circuits for image classification, requiring only shallow circuits and a small number of qubits -- both well within reach of near-term quantum devices.