Following our study of variational algorithms and parameterized quantum circuits (PQCs), this chapter focuses on Quantum Neural Networks (QNNs). These models adapt concepts from classical neural networks, using quantum circuits as their building blocks. We will examine various QNN architectures, such as Quantum Circuit Born Machines (QCBMs), Quantum Convolutional Neural Networks (QCNNs), and Quantum Graph Neural Networks (QGNNs). Additionally, the chapter covers the design of hybrid quantum-classical networks and addresses the practical difficulties in training QNNs, including gradient computation, optimization strategies, overfitting, and generalization.
5.1 Models of Quantum Neurons and Layers
5.2 Quantum Circuit Born Machines (QCBMs)
5.3 Quantum Convolutional Neural Networks (QCNNs)
5.4 Quantum Graph Neural Networks (QGNNs)
5.5 Hybrid Quantum-Classical Neural Network Architectures
5.6 Training QNNs Challenges and Strategies
5.7 Overfitting and Generalization in QNNs
5.8 Practice: Building and Training a Simple QNN