Optimization problems with discrete and continuous variables are ubiquitous in numerous important areas, including operations and scheduling, drug discovery, wireless communications, finance, integrated circuit design, compressed sensing and machine learning. Despite rapid advances in both algorithm and digital computing technology, even modest sized optimization problems that arise in practice may be very difficult to solve on modern digital computers. One alternative of current interest is the adiabatic quantum computing (AQC) or quantum annealing (QA). Sophisticated AQC/QA devices are already under development, but providing dense connectivity between qubits remains a major challenge with serious implications for the efficiency of AQC/QA approaches. In this talk, we will introduce a novel computing system, coherent Ising machine, and describe its theoretical and experimental performance. We start with the physics of quantum-to-classical crossover as a computational mechanism and how to construct such physical devices as quantum neurons and synapses. We show the performance comparison against various classical neural network models implemented in CPU and supercomputers as algorithms. We end the talk by introducing the portal of the QNNCloud service system based on the coherent Ising machines.