The roles of Kerr nonlinearity in a bosonic quantum neural network

The emerging technology of quantum neural networks (QNNs) attracts great attention from both the fields machine learning and physics with capability to gain advantage an artificial network (ANN) system. Comparing classical counterparts, QNNs have been proven be able speed up information processing, enhance prediction or classification efficiency as well offer versatile experimentally friendly platforms. It is established that Kerr nonlinearity indispensable element in a ANN, while, QNN, roles are not yet fully understood. In this work, we consider bosonic QNN investigate (simulating XOR gate) (generating Schr\"odinger cat states) tasks demonstrate only enables non-trivial but also makes system more robust errors.