Quantum machine learning for universal quantum computation

We describe a systematic method, using machine learning, to ``program'' a large-scale quantum computer. Large-scale quantum computational tasks require that the quantum computer be prepared in states which are multiply entangled; our methods show a way to create GHZ states over hundreds of qubits, and also to tailor the particular entanglement desired for a particular computation. In addition, current algorithmic approaches use a ``building block'' strategy, in which a procedure is formulated as a sequence of steps from a universal set, e.g., a sequence of CNOT, Hadamard, and phase shift gates. Using quantum learning enables us to perform computations without breaking down an algorithm into its ``building blocks'', eliminating a difficult step and potentially increasing efficiency by simplifying and reducing unnecessary complexity. Moreover, we demonstrate robustness of quantum learning to noise and to decoherence.