Constructing and programming quantum hardware for quantum annealing processes

Abstract

Quantum adiabatic computations may become non-adiabatic due to excitations caused, e.g., by thermal fluctuations. Instead of remaining at the instantaneous ground state, the evolving quantum state initially started at the ground state of the initial Hamiltonian can reach an excited state of the evolving Hamiltonian. Quantum hardware may be constructed and programmed to suppress such excitations from an instantaneous ground state to a higher energy state during an early stage of the computation. Quantum hardware may further be constructed and programmed to assist relaxation from higher energy states to lower energy states or the ground state during a later stage of the computation.

Description

CONSTRUCTING AND PROGRAMMING QUANTUM HARDWARE FOR QUANTUM ANNEALING PROCESSES This application is a divisional application of Canadian patent application No. 2,936,114. BACKGROUND This specification relates to constructing and programming quantum hardware for quantum annealing processes. SUMMARY Artificial intelligence tasks can be translated into machine learning optimization problems. To perform an artificial intelligence task, quantum hardware, e.g., a quantum processor, is constructed and programmed to encode the solution to a corresponding machine optimization problem into an energy spectrum of a many-body quantum Hamiltonian characterizing the quantum hardware. For example, the solution is encoded in the ground state of the Hamiltonian. The quantum hardware performs adiabatic quantum computation starting with a known ground state of a known initial Hamiltonian. Over time, as the known initial Hamiltonian evolves into the Hamiltonian for solving the problem, the known ground state evolves and remains at the instantaneous ground state of the evolving Hamiltonian. The energy spectrum or the ground state of the Hamiltonian for solving the problem is obtained at the end of the evolution without diagonalizing the Hamiltonian. Sometimes the quantum adiabatic computation becomes non-adiabatic due to excitations caused, e.g., by thermal fluctuations. Instead of remaining at the instantaneous ground state, the evolving quantum state initially started at the ground state of the initial Hamiltonian can reach an excited state of the evolving Hamiltonian. The quantum hardware is constructed and programmed to suppress such excitations from an instantaneous ground state to a higher energy state during an early stage of the computation. In addition, the quantum hardware is also constructed and programmed to assist relaxation from higher energy states to lower energy states or the ground state during a later stage of the computation. The robustness of finding the ground state of the Hamiltonian for solving the problem is improved. The details of one or more embodiments of the subject matter of this specification are set forth in the accompanying drawings and the description below. Other features, aspects, and advantages of the subject matter will be apparent from the description, the drawings, and the claims. DESCRIPTION OF DRAWINGS FIG. 1 is a schematic perspective view of interacting qubits in a quantum processor. FIG. 2 is a schematic diagram showing the structures and interactions of two qubits in a quantum processor. FIG. 3 is a schematic diagram showing the effect of a quantum governor on a quantum annealing process. FIG. 4 is a schematic diagram showing the interplay of an initial Hamiltonian, a problem Hamiltonian, and a Hamiltonian of a quantum governor chosen for the problem Hamiltonian during a quantum annealing process. FIG. 5 is a flow diagram of an example process for determining a quantum governor distribution. FIG. 6 is a flow diagram of an example process for performing an artificial intelligence task. DETAILED DESCRIPTION Overview Solutions to hard combinatorial problems, e.g., NP-hard problems and machine learning problems, can be encoded in the ground state of a many-body quantum Hamiltonian system, which is also called a quantum annealer or "QA." A QA of a system can sometimes be obtained through adiabatic quantum computation, in which the QA is initialized to a ground state of an initial Hamiltonian <semantics>Hi<annotation encoding="application / x-tex">H_i< / annotation>< / semantics> that is known and easy to prepare. Over time, the QA is adiabatically guided to a problem Hamiltonian <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics> that encodes the problem. In theory, during the adiabatic quantum computation, the QA can remain in the instantaneous ground state of a Hamiltonian <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> evolving from <semantics>Hi<annotation encoding="application / x-tex">H_i< / annotation>< / semantics> to <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>, where <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> can be expressed as: [Image disponible dans le document PDF, Image available in the PDF document] where s is a time dependent control parameter: [Image disponible dans le document PDF, Image available in the PDF document] and <semantics>tT<annotation encoding="application / x-tex">t_{T}< / annotation>< / semantics> is the total time of the adiabatic quantum computation. The QA will reach the ground state of the problem Hamiltonian <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics> with certainty, if the evolution of system is sufficiently slow with respect to the intrinsic energy scale of the system. In reality, the quantum computation may not be completely adiabatic and the QA may reach an excited state of <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> during the computation, which can lead to inaccurate result at the end of the quantum computation. For example, in many hard combinatorial optimization problems, e.g., in decision problems, when the problem Hamiltonian demonstrates a phase transition in its computational complexity, the size of a gap between an excited state and the ground state of <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> can be small, e.g., exponentially small, with respect to the intrinsic energy scale of the system. In such situations, the QA may undergo a quantum phase transition and can reach a large number, e.g., an exponentially large number, of excited states. In addition, the QA may also deviate from the ground state of <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> due to other factors such as quantum fluctuations induced by environmental interactions with the system and system imperfection errors, including control errors and fabrication imperfections. In this specification, the process of driving the QA from the ground state of <semantics>Hi<annotation encoding="application / x-tex">H_i< / annotation>< / semantics> to the ground state of <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics> is called a QA schedule or a quantum annealing process. Quantum hardware, such as quantum processors, of this specification includes a quantum chip that defines a quantum governor ("QG") in addition to <semantics>Hi<annotation encoding="application / x-tex">H_i< / annotation>< / semantics> and <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>, such that the evolving Hamiltonian <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> becomes <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics>: [Image disponible dans le document PDF, Image available in the PDF document] where <semantics>I(t)<annotation encoding="application / x-tex">I(t)< / annotation>< / semantics> and <semantics>P(t)<annotation encoding="application / x-tex">P(t)< / annotation>< / semantics> represent the time-dependency of the initial and problem Hamiltonians, <semantics>Hi<annotation encoding="application / x-tex">H_i< / annotation>< / semantics> and <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>, respectively; <semantics>G(t)<annotation encoding="application / x-tex">G(t)< / annotation>< / semantics> represents the time-dependency of the QG related Hamiltonian, <semantics>HG<annotation encoding="application / x-tex">H_G< / annotation>< / semantics>; and <semantics>HAG−B<annotation encoding="application / x-tex">H_{AG-B}< / annotation>< / semantics> is the interaction of the combined QA-QG system with its surrounding environment, commonly referred to as a bath. Generally it can be assumed that <semantics>HAG−B<annotation encoding="application / x-tex">H_{AG-B}< / annotation>< / semantics> is non-zero but constant and well-characterized. In some implementations: [Image disponible dans le document PDF, Image available in the PDF document] Generally, the QG can be considered as a class of non-information-bearing degrees of freedom that can be engineered to steer the dissipative dynamics of an information-bearing degree of freedom. In the example of <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics>, the information- bearing degree of freedom is the QA. In particular, a QG can be defined as an auxiliary subsystem with an independent tensor product structure. The QG can be characterized by <semantics>HG<annotation encoding="application / x-tex">H_G< / annotation>< / semantics>, which characterizes both the interaction between information- bearing degree of freedom and the QG and the free Hamiltonian of QG. The QG can interact with the QA coherently and in a non-perturbative manner. The interaction can also demonstrate strong non-Markovian effects. The quantum hardware is constructed and programmed to allow the QG to navigate the quantum evolution of a disordered quantum annealing hardware at finite temperature in a robust manner and improve the adiabatic quantum computation process. For example, the QG can facilitate driving the QA towards a quantum phase transition, while decoupling the QA from excited states of <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> by making the excited states effectively inaccessible by the QA. After the quantum phase transition, the QA enters another phase in which the QA is likely to be frozen in excited states due to quantum localization. The QG can adjust the energy level of the QA to be in tune with vibrational energies of environment to facilitate the QA to relax into a lower energy state or the ground state. Such an adjustment can increase the ground state fidelity, i.e., the fidelity of the QA being in the ground state at the end of the computation, and allow the QA to avoid pre-mature freeze in suboptimal solutions due to quantum localization. Generally, the QA experiences four phases in a quantum annealing process of the specification, including initialization, excitation, relaxation, and freezing, which are explained in more detailed below. The QG can assist the QA in the first two phases by creating a mismatch between average phonon energy of the bath and an average energy level spacing of the <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics> to suppress QA excitations. In the third and fourth stages, the QG can enhance thermal fluctuations by creating an overlap between the spectral densities of the system and the bath. The enhanced thermal fluctuations can allow the QA to have high relaxation rates from higher energy states to lower energy states or to the ground state of <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics>. In particular, due to quantum localization, the QA may be frozen at non-ground states. The QG can allow the QA to defreeze from the non-ground states. The QG can be used to achieve universal adiabatic quantum computing when quantum interactions are limited due to either natural or engineered constraints of the quantum hardware. For example, a quantum chip can have engineering constraints such that the Hamiltonian representing the interactions of qubits on the quantum chip becomes a k-local stochastic Hamiltonian. In some implementations, the quantum hardware is constructed and programmed to manipulate the structural and dynamical effects of environmental interactions and disorders, even without any control over the degrees of freedom of the environment. Generally, the QG is problem-dependent. The quantum hardware of the specification can be programmed to provide different QGs for different problem Hamiltonians. In some implementations, a QG can be determined for a given <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics> using a quantum control strategy developed based on mean-field and microscopic approaches. In addition or alternatively, the quantum control strategy can also implement random matrix theory and machine learning techniques in determining the QG. The combined QA and QG can be tuned and trained to generate desired statistical distributions of energy spectra for <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>, such as Poisson, Levy, or Boltzmann distributions. Example Quantum Hardware As shown in FIG. 1, in a quantum processor, a programmable quantum chip 100 includes 4 by 4 unit cells 102 of eight qubits 104, connected by programmable inductive couplers as shown by lines connecting different qubits. Each line may represent one or multiple couplers between a pair of qubits. The chip 100 can also include a larger number of unit cells 102, e.g., 8 by 8 or more. Within each unit cell 102, such as the cell circled with dashed lines and including qubit numbers <semantics>q1<annotation encoding="application / x-tex">q_1< / annotation>< / semantics> to <semantics>q8<annotation encoding="application / x-tex">q_8< / annotation>< / semantics>, one or more of the eight qubits 104, e.g., four qubits 104 labeled <semantics>q1<annotation encoding="application / x-tex">q_1< / annotation>< / semantics>, <semantics>q2<annotation encoding="application / x-tex">q_2< / annotation>< / semantics>, <semantics>q5<annotation encoding="application / x-tex">q_5< / annotation>< / semantics>, <semantics>q6<annotation encoding="application / x-tex">q_6< / annotation>< / semantics>, are logical qubits for use in computations carried out by the quantum processor. The other qubits 104, e.g., the other four qubits labeled <semantics>q3<annotation encoding="application / x-tex">q_3< / annotation>< / semantics>, <semantics>q4<annotation encoding="application / x-tex">q_4< / annotation>< / semantics>, <semantics>q7<annotation encoding="application / x-tex">q_7< / annotation>< / semantics>, <semantics>q𝒮<annotation encoding="application / x-tex">q_{\mathcal{S}}< / annotation>< / semantics>, are control qubits to be programmed to perform the function of the QG of this specification. The control qubits do not participate in the computations for which the logical qubits are configured. In some implementations, in each unit cell 102, half of the qubits are used as control qubits and the other half of the qubits are used as logical qubits. The size of the chip 100 implementing the QG is scalable by increasing the number of unit cells. In some implementations, the logical qubits and the control qubits have the same construction. In other implementations, the control qubits have simpler or less precise structures than the logical qubits. FIG. 2 shows an example pair of coupled qubits 200, 202 in the same unit cell of a chip. For example, the qubits 200, 202 can correspond to two logical qubits, e.g., qubits <semantics>q1<annotation encoding="application / x-tex">q_1< / annotation>< / semantics>, <semantics>q5<annotation encoding="application / x-tex">q_5< / annotation>< / semantics>, two control qubits, e.g., qubits <semantics>q3<annotation encoding="application / x-tex">q_3< / annotation>< / semantics>, <semantics>q7<annotation encoding="application / x-tex">q_7< / annotation>< / semantics>, or a logical qubit, e.g., qubit <semantics>q1<annotation encoding="application / x-tex">q_1< / annotation>< / semantics>, and a control qubit, e.g., qubit <semantics>q7<annotation encoding="application / x-tex">q_7< / annotation>< / semantics>. In this example, each qubit is a superconducting qubit and includes two parallelly connected Josephson boxes 204a, 204b or 208a, 208b. Each Josephson box can include a Josephson junction 206 parallelly connected to a capacitance 207. The qubits 200, 202 are subject to an external magnetic field B applied along a e3 direction perpendicular to the surface of the paper on which the figure is shown; the B field is labeled by the symbol <semantics>⊗<annotation encoding="application / x-tex">\otimes< / annotation>< / semantics> . A set of inductive couplers 210 laced between the qubits 200, 202 such that the qubits are coupled along the e3- e3 directions. In some implementations, a logical qubit and a control qubit are both constructed using two Josephson boxes connected in parallel. However, the logical qubit can be constructed with a higher precision than the control qubit. The less precisely constructed control qubit can therefore perform the function of the QG at a reduced cost. In other implementations, while a logical qubit is constructed using two Josephson boxes connected in parallel as shown in FIG. 2, a control qubit can be constructed using structures other than a Josephson box, such as a quantum Harmonic oscillator. In the chip 100 of FIG. 1, the control qubits in each unit cell interact with each other and with the logical qubits. These interactions act as a QG to assist the QA characterized by the logical qubits to reach the ground state of <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics> at the end of a quantum annealing process. As one example, The Hamiltonian of the QA can be written as: [Image disponible dans le document PDF, Image available in the PDF document] where <semantics>σix<annotation encoding="application / x-tex">\sigma_i^x< / annotation>< / semantics> and <semantics>σiz<annotation encoding="application / x-tex">\sigma_i^z< / annotation>< / semantics> are Pauli operator and each represents the spin of the ith qubit along the x direction or the z direction, respectively. <semantics>hi<annotation encoding="application / x-tex">h_i< / annotation>< / semantics> and <semantics>Jij<annotation encoding="application / x-tex">J_{ij}< / annotation>< / semantics> are parameters that can be programmed for different problems to be solved by adjusting the coupling provided by the inductive coupler set 210. <semantics>hi<annotation encoding="application / x-tex">h_i< / annotation>< / semantics> and <semantics>Jij<annotation encoding="application / x-tex">J_{ij}< / annotation>< / semantics> have real values. N is the total number of logical qubits for computation. The sparsity of the parameter <semantics>Jij<annotation encoding="application / x-tex">J_{ij}< / annotation>< / semantics> is constrained by the hardware connectivity, i.e., the connectivity of the qubits shown in FIG. 1. For unconnected qubits, the corresponding <semantics>Jij<annotation encoding="application / x-tex">J_{ij}< / annotation>< / semantics> is 0. Again, <semantics>I(t)<annotation encoding="application / x-tex">I(t)< / annotation>< / semantics> and <semantics>P(t)<annotation encoding="application / x-tex">P(t)< / annotation>< / semantics> represent the time- dependency of initial and problem Hamiltonians, respectively. In a simplified example, <semantics>I(t)<annotation encoding="application / x-tex">I(t)< / annotation>< / semantics> equals (1-s), and <semantics>P(t)<annotation encoding="application / x-tex">P(t)< / annotation>< / semantics> equals s, where s equals <semantics>t / tT<annotation encoding="application / x-tex">t / t_T< / annotation>< / semantics>. The additional M control qubits, where M can be the same as N or can be different, introduce additional quantum control mechanisms to the chip 100. The control mechanisms define an auxiliary subsystem that interacts with the QA of <semantics>HSG<annotation encoding="application / x-tex">H_{SG}< / annotation>< / semantics> or <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics>. The auxiliary subsystem includes tensor product structures and can be characterized by an intrinsic Hamiltonian of QG that commutes with QG-QA interaction Hamiltonian <semantics>HSG<annotation encoding="application / x-tex">H_{SG}< / annotation>< / semantics>. The auxiliary subsystem can be controlled by a global time-varying control knob <semantics>G(t)<annotation encoding="application / x-tex">G(t)< / annotation>< / semantics> and possibly a set of macroscopic programmable control parameters of the environment. In some implementations, the control knob is <semantics>s(1−s)<annotation encoding="application / x-tex">s(1-s)< / annotation>< / semantics>, where <semantics>s=t / tT<annotation encoding="application / x-tex">s = t / t_T< / annotation>< / semantics>; and a programmable control parameter of the environment is the temperature T. Accordingly, the Hamiltonian <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics> for the combined QA-QG system in the chip 100 is: [Image disponible dans le document PDF, Image available in the PDF document] where <semantics>I(t)<annotation encoding="application / x-tex">I(t)< / annotation>< / semantics>, <semantics>G(t)<annotation encoding="application / x-tex">G(t)< / annotation>< / semantics>, and <semantics>P(t)<annotation encoding="application / x-tex">P(t)< / annotation>< / semantics> describe the general time-dependency of global control parameters. In this Hamiltonian, the initial Hamiltonian is: [Image disponible dans le document PDF, Image available in the PDF document] The problem Hamiltonian <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics> is: [Image disponible dans le document PDF, Image available in the PDF document] and the QG Hamiltonian <semantics>HG<annotation encoding="application / x-tex">H_G< / annotation>< / semantics> is: [Image disponible dans le document PDF, Image available in the PDF document] [Image disponible dans le document PDF, Image available in the PDF document] [Image disponible dans le document PDF, Image available in the PDF document] where HGF is the free Hamiltonian of QG, and HGA the interaction Hamiltonian of QG and QA. Note that NG and NA are the total number of QG and QA qubits respectively. As an example, the total Hamiltonian can have the following explicit global time- dependency: [Image disponible dans le document PDF, Image available in the PDF document] Programming the Quantum Hardware For a given problem and its corresponding problem Hamiltonian <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>, a QG is determined to improve the ground state fidelity of the QA without diagonalizing <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>. Various QG realizations can be repeated to improve knowledge about the computational outcomes. In some implementations, a QG is determined such that before a system characterized by <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> experiences a quantum phase transition, the QG Hamiltonian <semantics>HQG<annotation encoding="application / x-tex">H_{QG}< / annotation>< / semantics> acts to suppress excitations of the QA. In particular, the QG is out of resonance with the average phonon energy of the bath, which creates a mismatch between the average phonon energy and average energy level spacing of the combined QA and QG, or <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics> to reduce unwanted excitations. After the system undergoes the quantum phase transition, the QG Hamiltonian <semantics>HQG<annotation encoding="application / x-tex">H_{QG}< / annotation>< / semantics> acts to enhance relaxation of the QA from any excited state to the ground state of <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics>. In particular, the average energy level spacing of <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics> is in resonance with the average phonon energy. The QG enhances thermal fluctuations by creating an overlap between the spectral densities of the system and its bath. The thermal fluctuations can facilitate the QA to reach the ground state of <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics> at a high relaxation rate and prevent the QA from being prematurely frozen at an excited state due to quantum localization. An example of desirable QG functions is shown in FIG. 3. The energy levels <semantics>E0,E1,E2,…Ei<annotation encoding="application / x-tex">E_0, E_1, E_2, \dots E_i< / annotation>< / semantics> (not shown) of <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> are plotted as a function of time <semantics>t<annotation encoding="application / x-tex">t< / annotation>< / semantics>. At <semantics>t=0<annotation encoding="application / x-tex">t = 0< / annotation>< / semantics>, <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> is <semantics>Hi<annotation encoding="application / x-tex">H_i< / annotation>< / semantics>, and at <semantics>t=tT<annotation encoding="application / x-tex">t = t_T< / annotation>< / semantics>, <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> is <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>. During a quantum annealing process from <semantics>t=0<annotation encoding="application / x-tex">t = 0< / annotation>< / semantics> to <semantics>t=0<annotation encoding="application / x-tex">t = 0< / annotation>< / semantics> <semantics>tT<annotation encoding="application / x-tex">t_{T}< / annotation>< / semantics>, the QA approximately experiences an initialization phase from <semantics>t=0<annotation encoding="application / x-tex">t=0< / annotation>< / semantics> to <semantics>t=tI<annotation encoding="application / x-tex">t=t_{I}< / annotation>< / semantics>, an excitation phase from <semantics>t=t1<annotation encoding="application / x-tex">t = t_1< / annotation>< / semantics> to <semantics>t=t2<annotation encoding="application / x-tex">t = t_2< / annotation>< / semantics>, a relaxation phase from <semantics>t=t2<annotation encoding="application / x-tex">t = t_2< / annotation>< / semantics> to <semantics>t=t3<annotation encoding="application / x-tex">t = t_3< / annotation>< / semantics>, and a freezing phase from <semantics>t=t3<annotation encoding="application / x-tex">t = t_3< / annotation>< / semantics> to <semantics>t=tT<annotation encoding="application / x-tex">t = t_T< / annotation>< / semantics>. The time <semantics>t2<annotation encoding="application / x-tex">t_2< / annotation>< / semantics> can correspond to a time at which a quantum phase transition occurs in a system characterized by <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics>. During the excitation phase, the QG increases, as indicated by arrows 300, 302, the energy spacing between adjacent energy levels <semantics>Δεi<annotation encoding="application / x-tex">\Delta \varepsilon_i< / annotation>< / semantics>, such as <semantics>Δε1=E2−E1<annotation encoding="application / x-tex">\Delta \varepsilon_1 = E_2 - E_1< / annotation>< / semantics> and <semantics>Δε0=E1−E0<annotation encoding="application / x-tex">\Delta \varepsilon_0 = E_1 - E_0< / annotation>< / semantics>, such that the increased energy spacing is much larger than the average phonon energy. During the relaxation phase, the QG adjusts the energy spacing <semantics>Δε0,Δε1,…<annotation encoding="application / x-tex">\Delta \varepsilon_0, \Delta \varepsilon_1, \dots< / annotation>< / semantics> to be comparable to the average phonon energy to facilitate relaxation of the QA from excited states to lower energy states or the ground state, as indicated by arrows 304, 306, 308, 310. The interplay of the three Hamiltonians, <semantics>Hi<annotation encoding="application / x-tex">H_i< / annotation>< / semantics>, <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>, and <semantics>HQG<annotation encoding="application / x-tex">H_{QG}< / annotation>< / semantics> over time in different phases of the quantum annealing process is schematically shown in FIG. 4. The control parameters <semantics>I(t)<annotation encoding="application / x-tex">I(t)< / annotation>< / semantics>, <semantics>P(t)<annotation encoding="application / x-tex">P(t)< / annotation>< / semantics>, and <semantics>G(t)<annotation encoding="application / x-tex">G(t)< / annotation>< / semantics> control the shapes of the curves for the corresponding Hamiltonians. In this example, <semantics>I(t)<annotation encoding="application / x-tex">I(t)< / annotation>< / semantics> and <semantics>P(t)<annotation encoding="application / x-tex">P(t)< / annotation>< / semantics> are linear and <semantics>G(t)<annotation encoding="application / x-tex">G(t)< / annotation>< / semantics> is parabolic. In addition, the QG can be chosen to allow the QA of <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics> to steadily evolve over the QA schedule and reach a final state that has a maximum overlap with the ground state of <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>. Ideally, the ground state fidelity of the QA at time <semantics>tT<annotation encoding="application / x-tex">t_T< / annotation>< / semantics> is 1. However, unity fidelity is hard to achieve within a finite period of time. Other than at time 0 and at time <semantics>tT<annotation encoding="application / x-tex">t_T< / annotation>< / semantics>, the QA of <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics> is in a mixed state of the combined <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>, <semantics>Hi<annotation encoding="application / x-tex">H_i< / annotation>< / semantics>, and <semantics>HQG<annotation encoding="application / x-tex">H_{QG}< / annotation>< / semantics>. The evolution of the QA can be expressed as: [Image disponible dans le document PDF, Image available in the PDF document] where <semantics>|ϵ0i⟩<annotation encoding="application / x-tex">|\epsilon_0^i\rangle< / annotation>< / semantics> is the state of the QA at time 0, <semantics>|ϵ0p⟩<annotation encoding="application / x-tex">|\epsilon_0^p\rangle< / annotation>< / semantics> is the state of the QA at time <semantics>tT<annotation encoding="application / x-tex">t_T< / annotation>< / semantics>, and <semantics>ρA(t)<annotation encoding="application / x-tex">\rho_A(t)< / annotation>< / semantics> is the density function of the QA at other times. By assigning a probability, e.g., using a probability mass function, to each state <semantics>|ϵ0p⟩<annotation encoding="application / x-tex">|\epsilon_0^p\rangle< / annotation>< / semantics>, the evolution of the QA can be further expressed as: [Image disponible dans le document PDF, Image available in the PDF document] where <semantics>fG(k)<annotation encoding="application / x-tex">f^{G}(k)< / annotation>< / semantics> is the probability mass function, <semantics>k=0,1,...,<annotation encoding="application / x-tex">k=0, 1, ...,< / annotation>< / semantics> and corresponds to quantum state levels, and <semantics>∑kfG(k)=1<annotation encoding="application / x-tex">\sum_{k} f^{G}(k) = 1< / annotation>< / semantics>. If the ground state fidelity is 1, then <semantics>fG(θ)=1<annotation encoding="application / x-tex">f^{G}(\theta) = 1< / annotation>< / semantics>, and <semantics>fG(k≠0)=0<annotation encoding="application / x-tex">f^{G}(k\neq 0) = 0< / annotation>< / semantics>. As described above, such a unity fidelity is hard to realize. Instead, a desirable QG can be selected to provide an exponential distribution function as: [Image disponible dans le document PDF, Image available in the PDF document] where <semantics>λG<annotation encoding="application / x-tex">\lambda_G< / annotation>< / semantics> defines the distribution of a QG family suitable for use with <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>. The probability mass function can be any probability distribution function. Examples include Poisson distribution functions, Levy distribution functions, and Boltzmann distribution functions. To determine a QG with desirable functions for a problem, including those functions described above with reference to FIGS. 3 and 4, one or more techniques can be used, including, for example, open quantum system models, random matrix theory, quantum Monte Carlo, and machine learning. An example process 500 for determining a QG is shown in FIG. 5, which can be performed by a classical processor, such as a classical computer, or a quantum processor, or a combination of them. In the process 500, information about energy states of <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> (502) are generally unknown. Thus, finding the global optimal QG is as at least as hard as solving the problem Hamiltonian itself. However, applications of QG in this context are highly robust to variations in its microscopic degrees of freedom. In some implementations the approach is to introduce a mean-field description of QG, such as a random matrix theory model that relies on a few effective control parameters. Then, a hybrid of unsupervised and supervised classical machine learning algorithms may be used to pre-train and train the macroscopic degrees of freedoms of QG. In order to generate sufficient training data for learning steps, state-of-art classical (metaheuristic) solvers may be used to sample from energy eigenfunction properties of the problem Hamiltonian that may include classical solvers such as Path-integral Monte Carlo, Diffusion Monte Carlo, Spin-vector Monte Carlo, Markov-Chain Monte Carlo, Simulated Annealing, Tabu search, large neighborhood search, and cluster updates. Training data can also be generated by previous generations of quantum processors, if available, including quantum annealers and gate-model quantum computers. The ansatz time at which the phase transition takes place can be chosen as the approximation <semantics>t=tT / 2<annotation encoding="application / x-tex">t = t_T / 2< / annotation>< / semantics>, and can be further fine-tuned during the learning procedures to enhance the fidelity of the spectral properties of QA with the Boltzmann distribution of the problem Hamiltonian that have been already sampled by a chosen classical solver. For a given problem Hamiltonian in training data, some useful information regarding the energy spectrum of the total Hamiltonian can be estimated, including approximate energy levels <semantics>Ei<annotation encoding="application / x-tex">E_i< / annotation>< / semantics>, the energy level spacings <semantics>Δεi<annotation encoding="application / x-tex">\Delta \varepsilon_i< / annotation>< / semantics>, the distribution of the energy level spacings, and the average energy level spacing <semantics>Δε¯<annotation encoding="application / x-tex">\overline{\Delta \varepsilon}< / annotation>< / semantics>. In some implementations, the average energy level spacing at time t is estimated as: [Image disponible dans le document PDF, Image available in the PDF document] where <semantics>εi(t)<annotation encoding="application / x-tex">\varepsilon_i(t)< / annotation>< / semantics> is the energy of the ith instantaneous eigenstate energy of <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics>, and N is the total number of eigenstates. Also in the process 500, the average phonon energy of the bath in which the system characterized by <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> is located can be calculated or experimentally measured (504). In approximation, the average phonon energy can be taken as <semantics>kT<annotation encoding="application / x-tex">kT< / annotation>< / semantics>, where <semantics>k<annotation encoding="application / x-tex">k< / annotation>< / semantics> is the Boltzmann constant, and T is the temperature. The average phonon energy can also be calculated in a more precise manner. For example, an open quantum system model of dynamics, such as the Lindblad formalism, can be selected for the calculation. The selection can be based on calibration data or quantum tomography of the quantum processor. Under the open quantum system model, the average phonon energy of a bath, in which a system represented by <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> is located, at any given temperature T can be defined as: [Image disponible dans le document PDF, Image available in the PDF document] where <semantics>J(ω)<annotation encoding="application / x-tex">J(\omega)< / annotation>< / semantics> can be the Omhic spectral density, i.e., <semantics>J(ω)=λωe−ωγ<annotation encoding="application / x-tex">J(\omega) = \lambda \omega e^{-\frac{\omega}{\gamma}}< / annotation>< / semantics>, the super-Omhic spectral density, i.e., <semantics>J(ω)=λω3e−ωγ<annotation encoding="application / x-tex">J(\omega) = \lambda \omega^3 e^{-\frac{\omega}{\gamma}}< / annotation>< / semantics>, the Drude-Lorentz spectral density, i.e., <semantics>J(ω)=121γ1γ<annotation encoding="application / x-tex">J(\omega) = \frac{1}{2} \frac{1}{\gamma} \frac{1}{\gamma}< / annotation>< / semantics> <semantics>2λωγω2+γ2<annotation encoding="application / x-tex">\frac{2\lambda\omega\gamma}{\omega^2+\gamma^2}< / annotation>< / semantics>, or a flat spectral distribution, i.e., <semantics>J(ω)=1<annotation encoding="application / x-tex">J(\omega)=1< / annotation>< / semantics>. In these equations, <semantics>λ<annotation encoding="application / x-tex">\lambda< / annotation>< / semantics> is the reorganization energy and <semantics>γ<annotation encoding="application / x-tex">\gamma< / annotation>< / semantics> is the bath frequency cut-off. A probability mass function for the ground state fidelity of the QA is selected (506). Based on the obtained information, training data, the calculated average phonon energy, and the selected probability mass function, the process 500 then determines (508) a QG distribution for <semantics>Hp<annotation encoding="application / x-tex">H_p< / annotation>< / semantics>. For example, the QG distribution can be represented by an exponential family, such as a Gaussian unitary ensemble, of random matrices selected using a random matrix theory model. The average energy level spacing <semantics>Δg<annotation encoding="application / x-tex">\Delta g< / annotation>< / semantics> and the maximum and minimum energy eigenvalues of the QG or <semantics>HQG<annotation encoding="application / x-tex">H_{QG}< / annotation>< / semantics> are determined to allow the QG to function as desired. In particular, in the second phase of the QA schedule, e.g., during time <semantics>t1<annotation encoding="application / x-tex">t_1< / annotation>< / semantics> to <semantics>t2<annotation encoding="application / x-tex">t_2< / annotation>< / semantics> shown in FIG. 3, the average energy level spacing of the QG is chosen such that: [Image disponible dans le document PDF, Image available in the PDF document] Where <semantics>Δεe<annotation encoding="application / x-tex">\Delta \varepsilon_{e}< / annotation>< / semantics> is the modified average energy level of QA and QG. This choice increases the energy level spacing of <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> such that the combined energy level spacing of <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics> is much larger than the average phonon energy. Accordingly, possible excitations of the QA to a higher energy state by thermal fluctuation are suppressed. In addition, the QG is also selected such that in the third phase of the QA schedule, e.g., during time t2 to t3 shown in FIG. 3, the average energy level spacing of the QG satisfies: [Image disponible dans le document PDF, Image available in the PDF document] This choice allows the energy level spacing of <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> to be similar to the thermal fluctuation. The QA can relax to a lower energy state or the ground state at a high rate. The selected exponential family can be parameterized with respect to the controllable parameters, such as the coupling between qubits, of the quantum hardware. In some implementations, a deep neural network is used to represent the QG- QA system or the system characterized by <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics>, and stochastic gradient descent is used to train the QG distribution. As an example, the training is done by selecting a statistically meaningful number, e.g., 1000, of random matrices from a parameterized exponential family that can in average generate path-integral Monte-Carlo outputs, within the desired probability mass function for a given <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> of interest. In some implementations, the training can start with an initial QG distribution selected based on the desired average combined energy level spacing <semantics>Δεs¯<annotation encoding="application / x-tex">\overline{\Delta \varepsilon_s}< / annotation>< / semantics> discussed above. FIG. 6 shows an example process 600 in which a control system programs QA hardware, such as a quantum processor, for the QA hardware to perform an artificial intelligence task. The control system includes one or more classical, i.e., non- quantum, computers, and may also include a quantum computer. The task is translated into a machine learning optimization problem, which is represented in a machine-readable form. The control system receives (602) the machine-readable machine learning optimization problem. The control system encodes (606) the optimization problem into the energy spectrum of an engineered <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics>. The encoding is based on structure of the QA hardware, such as the couplings between qubits. An example of <semantics>Htotal<annotation encoding="application / x-tex">H_{total}< / annotation>< / semantics> is the Ising Hamiltonian <semantics>HSG<annotation encoding="application / x-tex">H_{SG}< / annotation>< / semantics>, and the encoding determines the values for the parameters <semantics>hi<annotation encoding="application / x-tex">h_i< / annotation>< / semantics> and <semantics>Jij<annotation encoding="application / x-tex">J_{ij}< / annotation>< / semantics>. The encoded information, such as <semantics>hi<annotation encoding="application / x-tex">h_i< / annotation>< / semantics> and <semantics>Jij<annotation encoding="application / x-tex">J_{ij}< / annotation>< / semantics>, is provided to the QA hardware, which receives (620) the information as initialization parameters for the hardware. To stabilize the QA during a quantum annealing process to be performed by the QA hardware, the control system further devises (608) a QG, e.g., by selecting one QG from a QG distribution determined using the process 500 of FIG. 5. The devised QG is characterized by control parameters including <semantics>hiG<annotation encoding="application / x-tex">h_i^G< / annotation>< / semantics>, <semantics>JijG<annotation encoding="application / x-tex">J_{ij}^G< / annotation>< / semantics>, <semantics>JijGA<annotation encoding="application / x-tex">J_{ij}^{GA}< / annotation>< / semantics> which are sent to the QA hardware to program the QA hardware. The QA hardware receives (620) the initialization parameters, such as <semantics>hi<annotation encoding="application / x-tex">h_i< / annotation>< / semantics> and <semantics>Jij<annotation encoding="application / x-tex">J_{ij}< / annotation>< / semantics>, and also receives (622) the control parameters for the QG, such as <semantics>hiG<annotation encoding="application / x-tex">h_i^G< / annotation>< / semantics>, <semantics>JijG<annotation encoding="application / x-tex">J_{ij}^G< / annotation>< / semantics>, <semantics>JijGA<annotation encoding="application / x-tex">J_{ij}^{GA}< / annotation>< / semantics>, and is programmed and initialized by the control system according to the received initialization parameters and the ansatz for QG parameters. The QA hardware implements (624) the quantum annealing schedule to obtain eigenstates of the combined QA-QG system characterized by <semantics>Htot<annotation encoding="application / x-tex">H_{tot}< / annotation>< / semantics>. The solution to the machine learning optimization problem is encoded in these eigenstates. After a predetermined amount of time, the QA schedule ends and the QA hardware provides (626) an output represented by the eigenstates and their corresponding energy spectra. The output can be read by the control system or by another classical computer or quantum computer. The predetermined amount of time can be in the order of <semantics>1 / (Δℰg¯)2<annotation encoding="application / x-tex">1 / (\overline{\Delta \mathcal{E}_g})^2< / annotation>< / semantics>. However, shorter or longer periods of time can be used. A shorter time period may provide efficiency, and a longer time period may provide a high ground state fidelity. As described above, in the output provided by the QA hardware, the ground state fidelity of the QA is generally smaller than 1. When the fidelity is smaller than 1, the one-time output provided by the QA hardware may not accurately encode the solution to the problem. In some implementations, the QA hardware implements the QA schedule multiple times, using the same QG or different QGs provided by the control system that have different sets of control parameters, such as <semantics>hiG<annotation encoding="application / x-tex">h_i^G< / annotation>< / semantics>, <semantics>JijG<annotation encoding="application / x-tex">J_{ij}^G< / annotation>< / semantics>, <semantics>JijGA<annotation encoding="application / x-tex">J_{ij}^{GA}< / annotation>< / semantics>, selected from the same QG distribution determined for the problem, to provide multiple outputs. The multiple outputs can be statistically analyzed and the problem or the artificial intelligent task can be resolved or performed based on the statistical results. In particular, in the process 600, after the control system receives and stores (610) the output provided by the QA hardware, the control system determines (612) whether the QA hardware has completed the predetermined number of iterations of QA schedules. If not, then the control system returns to the step 608 by devising another QG, which can be the same as the previously used QG or a different QG selected from the previously determined QG distribution. The QA hardware receives (622) another set of control parameters for the QG and is re-programmed by the control system based on this set of control parameters and the previously determined initialization parameters that encode the problem. The QA schedule is implemented again (624) and another output is provided (626). If the QA hardware has completed the predetermined number of iterations of QA schedule, then the control system or another data processing system statistically processes (614) all outputs to provide solutions to the problem. The predetermined number of iterations can be 100 iterations or more, or 1000 iterations or more. In some implementations, the number of iterations can be chosen in connection with the length of the QA schedule, so that the process 600 can be performed with high efficiency and provide solutions to the problems with high accuracy. For example, when the length of each QA schedule is relatively short, e.g., shorter than <semantics>1 / (Δℰg)2<annotation encoding="application / x-tex">1 / (\Delta \mathcal{E}_{g})^{2}< / annotation>< / semantics>, the predetermined number of iterations can be chosen to be relatively large, e.g., 1000 iterations or more. In other situations when the length of each QA schedule is relatively long, e.g., longer than <semantics>1 / (Δℰg)2<annotation encoding="application / x-tex">1 / (\Delta \mathcal{E}_g)^2< / annotation>< / semantics>, the predetermined number of iterations can be chosen to be relatively small, e.g., less than 1000 iterations. Embodiments of the digital, i.e., non quantum, subject matter and the digital functional operations described in this specification can be implemented in digital electronic circuitry, in tangibly-embodied computer software or firmware, in computer hardware, including the structures disclosed in this specification and their structural equivalents, or in combinations of one or more of them. Embodiments of the digital subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible non transitory storage medium for execution by, or to control the operation of, data processing apparatus. The computer storage medium can be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them. Alternatively or in addition, the program instructions can be encoded on an artificially generated propagated signal, e.g., a machine-generated electrical, optical, or electromagnetic signal, that is generated to encode information for transmission to suitable receiver apparatus for execution by a data processing apparatus. The term "data processing apparatus" refers to digital data processing hardware and encompasses all kinds of apparatus, devices, and machines for processing data, including by way of example a programmable digital processor, a digital computer, or multiple digital processors or computers. The apparatus can also be, or further include, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit). The apparatus can optionally include, in addition to hardware, code that creates an execution environment for computer programs, e.g., code that constitutes processor firmware, a protocol stack, a database management system, an operating system, or a combination of one or more of them. A computer program, which may also be referred to or described as a program, software, a software application, a module, a software module, a script, or code, can be written in any form of programming language, including compiled or interpreted languages, or declarative or procedural languages, and it can be deployed in any form, including as a stand alone program or as a module, component, subroutine, or other unit suitable for use in a digital computing environment. A computer program may, but need not, correspond to a file in a file system. A program can be stored in a portion of a file that holds other programs or data, e.g., one or more scripts stored in a markup language document, in a single file dedicated to the program in question, or in multiple coordinated files, e.g., files that store one or more modules, sub programs, or portions of code. A computer program can be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a data communication network. The processes and logic flows described in this specification can be performed by one or more programmable digital computers, operating with one or more quantum processors, as appropriate, executing one or more computer programs to perform functions by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus can also be implemented as, special purpose logic circuitry, e.g., an FPGA or an ASIC, or by a combination of special purpose logic circuitry and one or more programmed computers. For a system of one or more digital computers to be "configured to" perform particular operations or actions means that the system has installed on it software, firmware, hardware, or a combination of them that in operation cause the system to perform the operations or actions. For one or more computer programs to be configured to perform particular operations or actions means that the one or more programs include instructions that, when executed by digital data processing apparatus, cause the apparatus to perform the operations or actions. Digital computers suitable for the execution of a computer program can be based on general or special purpose microprocessors or both, or any other kind of central processing unit. Generally, a central processing unit will receive instructions and data from a read only memory or a random access memory or both. The essential elements of a computer are a central processing unit for performing or executing instructions and one or more memory devices for storing instructions and data. The central processing unit and the memory can be supplemented by, or incorporated in, special purpose logic circuitry. Generally, a digital computer will also include, or be operatively coupled to receive data from or transfer data to, or both, one or more mass storage devices for storing data, e.g., magnetic, magneto optical disks, or optical disks. However, a computer need not have such devices. Computer readable media suitable for storing computer program instructions and data include all forms of non volatile memory, media and memory devices, including by way of example semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magneto optical disks; and CD ROM and DVD-ROM disks. Control of the various systems described in this specification, or portions of them, can be implemented in a computer program product that includes instructions that are stored on one or more non-transitory machine-readable storage media, and that are executable on one or more digital processing devices. The systems described in this specification, or portions of them, can each be implemented as an apparatus, method, or electronic system that may include one or more digital processing devices and memory to store executable instructions to perform the operations described in this specification. While this specification contains many specific implementation details, these should not be construed as limitations on the scope of what may be claimed, but rather as descriptions of features that may be specific to particular embodiments. Certain features that are described in this specification in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable subcombination. Moreover, although features may be described above as acting in certain combinations and even initially claimed as such, one or more features from a claimed combination can in some cases be excised from the combination, and the claimed combination may be directed to a subcombination or variation of a subcombination. Similarly, while operations are depicted in the drawings in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve desirable results. In certain circumstances, multitasking and parallel processing may be advantageous. Moreover, the separation of various system modules and components in the embodiments described above should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products. Particular embodiments of the subject matter have been described. Other embodiments are within the scope of the following claims. For example, the actions recited in the claims can be performed in a different order and still achieve desirable results. As one example, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some cases, multitasking and parallel processing may be advantageous.

Claims

<pat:Claims com:id="claims"> <pat:Claim com:id="CLM-00001"> <pat:ClaimNumber>1< / pat:ClaimNumber> <pat:ClaimText>1. A computer implemented method comprising: obtaining information representing energy states of a total Hamiltonian, wherein the total Hamiltonian characterizes dynamics of a respective quantum system, the quantum system comprising a quantum annealing subsystem and quantum governor subsystem; calculating an average phonon energy of a bath in which the quantum system is located; selecting a probability mass function for a ground state fidelity of the quantum annealing subsystem; and determining a quantum-governor distribution based on i) the information representing the energy states of the total Hamiltonian, ii) the average phonon energy, and iii) the probability mass function; determining a quantum-governor Hamiltonian using the quantum-governor distribution; and programming quantum annealing hardware using the quantum-governor Hamiltonian. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00002"> <pat:ClaimNumber>2< / pat:ClaimNumber> <pat:ClaimText>2. The method of claim 1, wherein the information representing the energy states of the total Hamiltonian comprises one or more of energy levels, spacings between energy levels, a distribution of the spacings between energy levels, or an average spacing between energy levels. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00003"> <pat:ClaimNumber>3< / pat:ClaimNumber> <pat:ClaimText>3. The method of claim 2, wherein the average energy level spacing is given by <semantics>Δϵ(t)=<annotation encoding="application / x-tex">\Delta \epsilon(t) =< / annotation>< / semantics> <semantics>2∑i=0N−1|ϵi(t)−ϵj(t)|N(N−1)<annotation encoding="application / x-tex">\frac{2\sum_{i=0}^{N-1}|\epsilon_i(t)-\epsilon_j(t)|}{N(N-1)}< / annotation>< / semantics> where <semantics>ϵi(t)<annotation encoding="application / x-tex">\epsilon_i(t)< / annotation>< / semantics> represents an energy of an i-th instantaneous eigenstate energy of the total Hamiltonian and N represents a total number of eigenstates of the total Hamiltonian. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00004"> <pat:ClaimNumber>4< / pat:ClaimNumber> <pat:ClaimText>4. The method of claim 1, wherein the information representing the energy states of the total Hamiltonian comprises information representing the energy states of the total Hamiltonian at a time of a quantum phase transition. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00005"> <pat:ClaimNumber>5< / pat:ClaimNumber> <pat:ClaimText>5. The method of claim 1, wherein calculating the average phonon energy of the bath comprises experimentally measuring the average phonon energy of the bath. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00006"> <pat:ClaimNumber>6< / pat:ClaimNumber> <pat:ClaimText>6. The method of claim 1, wherein calculating the average phonon energy of the bath comprises computing <semantics>kT<annotation encoding="application / x-tex">kT< / annotation>< / semantics>, where <semantics>k<annotation encoding="application / x-tex">k< / annotation>< / semantics> represents the Boltzmann constant and <semantics>T<annotation encoding="application / x-tex">T< / annotation>< / semantics> represents temperature. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00007"> <pat:ClaimNumber>7< / pat:ClaimNumber> <pat:ClaimText>7. The method of claim 1, wherein the average phonon energy satisfies the following equation: [Image disponible dans le document PDF, Image available in the PDF document] where <semantics>J(ω)<annotation encoding="application / x-tex">J(\omega)< / annotation>< / semantics> represents one of an Omhic spectral density: <semantics>J(ω)=λωe−ωγ<annotation encoding="application / x-tex">J(\omega) = \lambda \omega e^{-\frac{\omega}{\gamma}}< / annotation>< / semantics>, a super-Omhic spectral density: <semantics>J(ω)=λω3e−ωγ<annotation encoding="application / x-tex">J(\omega) = \lambda \omega^3 e^{-\frac{\omega}{\gamma}}< / annotation>< / semantics>, a Drude-Lorentz spectral density: <semantics>J(ω)=2λωγω2+γ2<annotation encoding="application / x-tex">J(\omega) = \frac{2\lambda\omega\gamma}{\omega^2 + \gamma^2}< / annotation>< / semantics>, or a flat spectral distribution: <semantics>J(ω)=1<annotation encoding="application / x-tex">J(\omega) = 1< / annotation>< / semantics>, <semantics>λ<annotation encoding="application / x-tex">\lambda< / annotation>< / semantics> represents a reorganization energy, and <semantics>γ<annotation encoding="application / x-tex">\gamma< / annotation>< / semantics> represents a bath frequency cut-off. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00008"> <pat:ClaimNumber>8< / pat:ClaimNumber> <pat:ClaimText>8. The method of claim 1, wherein calculating the average phonon energy of the bath comprises calculating the average phonon energy using an open quantum system model. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00009"> <pat:ClaimNumber>9< / pat:ClaimNumber> <pat:ClaimText>9. The method of claim 1, wherein the total Hamiltonian is given by [Image disponible dans le document PDF, Image available in the PDF document] wherein i represents an <semantics>ith<annotation encoding="application / x-tex">i^{th}< / annotation>< / semantics> superconducting unit, <semantics>hi<annotation encoding="application / x-tex">h_i< / annotation>< / semantics> and <semantics>Jij<annotation encoding="application / x-tex">J_{ij}< / annotation>< / semantics> represent parameters included in a problem Hamiltonian and take real values that are associated with a coupling strength between superconducting units, <semantics>hiG<annotation encoding="application / x-tex">h_i^G< / annotation>< / semantics>, <semantics>JijG<annotation encoding="application / x-tex">J_{ij}^G< / annotation>< / semantics> and <semantics>JijGA<annotation encoding="application / x-tex">J_{ij}^{GA}< / annotation>< / semantics> represent parameters included in a quantum governor Hamiltonian and take real values that are associated with a coupling strength between superconducting units, <semantics>I(t)<annotation encoding="application / x-tex">I(t)< / annotation>< / semantics>, <semantics>G(t)<annotation encoding="application / x-tex">G(t)< / annotation>< / semantics>, and <semantics>P(t)<annotation encoding="application / x-tex">P(t)< / annotation>< / semantics> represent time-dependent control parameters, and <semantics>σix<annotation encoding="application / x-tex">\sigma_i^x< / annotation>< / semantics>, <semantics>σiz<annotation encoding="application / x-tex">\sigma_i^z< / annotation>< / semantics> represent Pauli-x and Pauli-z operators for superconducting unit i. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00010"> <pat:ClaimNumber>10< / pat:ClaimNumber> <pat:ClaimText>10. The method of claim 1, wherein the quantum-governor distribution is represented by an exponential family of random matrixes selected using a random matrix theory model. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00011"> <pat:ClaimNumber>11< / pat:ClaimNumber> <pat:ClaimText>11. The method of claim 10, wherein the exponential family is parameterized with respect to controllable quantum annealing hardware parameters, the parameters comprising coupling between qubits. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00012"> <pat:ClaimNumber>12< / pat:ClaimNumber> <pat:ClaimText>12. The method of claim 11, further comprising implementing, by the programmed quantum annealing hardware, a quantum annealing schedule. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00013"> <pat:ClaimNumber>13< / pat:ClaimNumber> <pat:ClaimText>13. The method of claim 12, wherein in a second phase of the quantum annealing schedule, an average energy level spacing of the quantum governor subsystem satisfies <semantics>Δϵa¯≫ω¯<annotation encoding="application / x-tex">\overline{\Delta\epsilon_a} \gg \overline{\omega}< / annotation>< / semantics> where <semantics>Δϵa¯<annotation encoding="application / x-tex">\overline{\Delta\epsilon_a}< / annotation>< / semantics> represents a modified average energy level of the quantum annealing subsystem and quantum governor subsystem and <semantics>ω¯<annotation encoding="application / x-tex">\overline{\omega}< / annotation>< / semantics> represents the average phonon energy. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00014"> <pat:ClaimNumber>14< / pat:ClaimNumber> <pat:ClaimText>14. The method of claim 12, wherein in a third phase of the quantum annealing schedule, an average energy level spacing of the quantum governor subsystem satisfies <semantics>Δϵa¯≈ω¯<annotation encoding="application / x-tex">\overline{\Delta\epsilon_a} \approx \overline{\omega}< / annotation>< / semantics> where <semantics>Δϵa¯<annotation encoding="application / x-tex">\overline{\Delta\epsilon_a}< / annotation>< / semantics> represents a modified average energy level of the quantum annealing subsystem and quantum governor subsystem and <semantics>ω¯<annotation encoding="application / x-tex">\overline{\omega}< / annotation>< / semantics> represents the average phonon energy. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00015"> <pat:ClaimNumber>15< / pat:ClaimNumber> <pat:ClaimText>15. A control system for programming quantum annealing hardware, the control system comprising: one or more computers; and one or more computer-readable media coupled to the one or more computers having instructions stored thereon which, when executed by the one or more computers, cause the one or more computers to perform operations comprising: obtaining information representing energy states of a total Hamiltonian, wherein the total Hamiltonian characterizes dynamics of a respective quantum system, the quantum system comprising a quantum annealing subsystem and quantum governor subsystem; calculating an average phonon energy of a bath in which the quantum system is located; selecting a probability mass function for a ground state fidelity of a quantum annealing sub-Hamiltonian in the total Hamiltonian; and determining a quantum-governor distribution based on i) the information representing the energy states of the total Hamiltonian, ii) the average phonon energy, and iii) the probability mass function; determining a quantum-governor Hamiltonian using the quantum-governor distribution; and programming quantum annealing hardware using the quantum-governor Hamiltonian. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00016"> <pat:ClaimNumber>16< / pat:ClaimNumber> <pat:ClaimText>16. The control system of claim 15, wherein the operations further comprise implementing, by the programmed quantum annealing hardware, a quantum annealing schedule. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00017"> <pat:ClaimNumber>17< / pat:ClaimNumber> <pat:ClaimText>17. The control system of claim 16, wherein in a second phase of the quantum annealing schedule, an average energy level spacing of the quantum governor subsystem satisfies <semantics>Δϵq¯≫ω¯<annotation encoding="application / x-tex">\overline{\Delta\epsilon_q} \gg \overline{\omega}< / annotation>< / semantics> where <semantics>Δϵq¯<annotation encoding="application / x-tex">\overline{\Delta\epsilon_q}< / annotation>< / semantics> represents a modified average energy level of the quantum annealing subsystem and quantum governor subsystem and <semantics>ω¯<annotation encoding="application / x-tex">\overline{\omega}< / annotation>< / semantics> represents the average phonon energy. < / pat:ClaimText> < / pat:Claim> <pat:Claim com:id="CLM-00018"> <pat:ClaimNumber>18< / pat:ClaimNumber> <pat:ClaimText>18. The control system of claim 16, wherein in a third phase of the quantum annealing schedule, an average energy level spacing of the quantum governor subsystem satisfies <semantics>Δϵq¯≈ω¯<annotation encoding="application / x-tex">\overline{\Delta\epsilon_q} \approx \overline{\omega}< / annotation>< / semantics> where <semantics>Δϵq¯<annotation encoding="application / x-tex">\overline{\Delta\epsilon_q}< / annotation>< / semantics> represents a modified average energy level of the quantum annealing subsystem and quantum governor subsystem and <semantics>ω¯<annotation encoding="application / x-tex">\overline{\omega}< / annotation>< / semantics> represents the average phonon energy. < / pat:ClaimText> < / pat:Claim> < / pat:Claims>