Extending constrained-bodyness tensor network simulation to multi-class and higher-dimensional settings

Extend the constrained-bodyness tensor network method based on matrix product states—currently applicable to binary classification tasks on one-dimensional quantum systems—to support multi-class classification problems and higher-dimensional system topologies.

Background

The paper introduces a matrix product state (MPS) based constrained-bodyness tensor network approach for simulating QCNNs that operates efficiently by restricting to low-bodyness operator subspaces. This method is used to dequantize QCNNs for classical datasets, and complements Pauli propagation surrogate techniques for quantum datasets.

The authors note that the present implementation is limited to binary classification and performs best on one-dimensional systems, and explicitly state that extending the method to multi-class tasks and higher-dimensional architectures is left for future work.