Quaternion Quantum Neural Network for Classification

Abstract

We propose the quaternionic quantum neural network (QQNN) for pattern recognition based on the formulation of quaternionic qubits and the construction of activation operators. In this model, the inputs and targets are represented by quaternionic qubits. The proposed neural network is evaluated through a series of experiments using different benchmark datasets, where the results show its superiority as a classifier in terms of accuracy when it is compared to conventional (real-valued) neural networks.