Quantum-inspired complex convolutional neural networks

Quantum-inspired artificial neural network is an interesting research area, which combines quantum computing and deep learning. Several models of quantum-inspired neuron with real-valued weights have been proposed, they were mainly used to build the three-layer feedforward networks. In this work, we improve convolutional networks (CNNs) by utilizing way data representation operation. Specifically, first exploiting complex-valued weights, richer representational capacity better non-linearity. Moreover, extend method implementing neurons perform operations, naturally draw (QICNNs) capable processing high-dimensional data. Here five specific types QICNNs are different in layers fully connected layers. We establish detail mathematical framework implement QICNNs. The performances accuracy, convergence robustness against classical counterpart tested using MNIST CIFAR-10 datasets. results show that (1) QICNN can achieve higher classification accuracy (up 99.65%) than CNN when dataset; (2) has faster speed, means be trained easily a similar number parameters; (3) case employing weight initialization or rotating input It expected our outperform counterparts more practical learning tasks.