Separable and non-separable data representation for pattern discrimination

TL;DR

This paper introduces a quantum-inspired data processing workflow for pattern recognition that is easy to implement, resource-efficient, and capable of exploring the effects of tensor product structures on classification performance.

Contribution

It presents a novel quantum information theory-based scheme for data representation in pattern recognition, emphasizing practical implementation and analysis of tensor product effects.

Findings

The scheme is easily implementable with modest resources.

It can differentiate pattern recognition outcomes based on tensor product structures.

Illustrative example demonstrates the scheme's application to 2D data classification.

Abstract

We provide a complete work-flow, based on the language of quantum information theory, suitable for processing data for the purpose of pattern recognition. The main advantage of the introduced scheme is that it can be easily implemented and applied to process real-world data using modest computation resources. At the same time it can be used to investigate the difference in the pattern recognition resulting from the utilization of the tensor product structure of the space of quantum states. We illustrate this difference by providing a simple example based on the classification of 2D data.