Realization of a quantum neural network using repeat-until-success circuits in a superconducting quantum processor

  1. M. S. Moreira,
  2. G. G. Guerreschi,
  3. W. Vlothuizen,
  4. J. F. Marques,
  5. J. van Straten,
  6. S. P. Premaratne,
  7. X. Zou,
  8. H. Ali,
  9. N. Muthusubramanian,
  10. C. Zachariadis,
  11. J. van Someren,
  12. M. Beekman,
  13. N. Haider,
  14. A. Bruno,
  15. C. G. Almudever,
  16. A. Y. Matsuura,
  17. and L. DiCarlo
Artificial neural networks are becoming an integral part of digital solutions to complex problems. However, employing neural networks on quantum processors faces challenges related
to the implementation of non-linear functions using quantum circuits. In this paper, we use repeat-until-success circuits enabled by real-time control-flow feedback to realize quantum neurons with non-linear activation functions. These neurons constitute elementary building blocks that can be arranged in a variety of layouts to carry out deep learning tasks quantum coherently. As an example, we construct a minimal feedforward quantum neural network capable of learning all 2-to-1-bit Boolean functions by optimization of network activation parameters within the supervised-learning paradigm. This model is shown to perform non-linear classification and effectively learns from multiple copies of a single training state consisting of the maximal superposition of all inputs.

eQASM: An Executable Quantum Instruction Set Architecture

  1. X. Fu,
  2. L. Riesebos,
  3. M. A. Rol,
  4. J. van Straten,
  5. J. van Someren,
  6. N. Khammassi,
  7. I. Ashraf,
  8. R.F.L. Vermeulen,
  9. V. Newsum,
  10. K. K. L. Loh,
  11. J. C. de Sterke,
  12. W. J. Vlothuizen,
  13. R. N. Schouten,
  14. C. G. Almudever,
  15. L. DiCarlo,
  16. and K. Bertels
Bridging the gap between quantum software and hardware, recent research proposed a quantum control microarchitecture QuMA which implements the quantum microinstruction set QuMIS. However,
QuMIS does not offer feedback control, and is tightly bound to the hardware implementation. Also, as the number of qubits grows, QuMA cannot fetch and execute instructions fast enough to apply all operations on qubits on time. Known as the quantum operation issue rate problem, this limitation is aggravated by the low information density of QuMIS instructions. In this paper, we propose an executable quantum instruction set architecture (QISA), called eQASM, that can be translated from the quantum assembly language (QASM), supports feedback, and is executed on a quantum control microarchitecture. eQASM alleviates the quantum operation issue rate problem by efficient timing specification, single-operation-multiple-qubit execution, and a very-long-instruction-word architecture. The definition of eQASM focuses on the assembly level to be expressive. Quantum operations are configured at compile time instead of being defined at QISA design time. We instantiate eQASM into a 32-bit instruction set targeting a seven-qubit superconducting quantum processor. We validate our design by performing several experiments on a two-qubit quantum processor.