Machine learning-based quantum noise decoder

Abstract

A quantum noise decoder including a quantum error detection model trained by machine learning is trained using training data including empirical operation data captured at least in part based on the operation of a particular quantum processor. A noise model for the particular quantum processor is generated based on the quantum error detection model trained by machine learning. The noise model is provided. Providing the noise model includes at least one of (a) causing a graphical representation of the noise model to be provided via a display of a computing entity such that components or parameters of the particular quantum processor are modified or changed based on the graphical representation of the noise model, or (b) providing the noise model as an input associated with executable instructions for execution by a controller of the particular quantum processor or a computing entity communicating with the controller such that components or parameters of the particular quantum processor are modified or changed based on the noise model.

Description

Technical Field

[0001] Cross - Reference to Related Applications This application claims the benefit of U.S. Patent Application No. 18 / 339,693, filed Jun. 22, 2023, which claims the benefit of U.S. Patent Application No. 63 / 367,770, filed Jul. 6, 2022, the contents of which are hereby incorporated by reference in their entirety.

[0002] Various embodiments relate to the use of models trained by machine learning to characterize noise in a quantum processor. Various embodiments relate to leveraging the characterization of noise in a quantum processor to reduce the noise in the quantum processor. For example, an exemplary embodiment relates to a quantum noise detector that includes a model trained by machine learning that is trained using operational data of a particular quantum processor and configured to be used in determining a noise model for the particular quantum processor.

Background Art

[0003] Large - scale quantum computers are expected to solve problems that are currently intractable with today's technology in fields such as chemistry, materials science, and biology. Solving such problems involves computations employing quantum algorithms implemented using deep quantum circuits. Achieving the required level of accuracy in these deep circuits requires a high level of reliability in quantum operations. To achieve such reliability, quantum error correction (QEC) is employed during computation to suppress noise to the required level. Through expended efforts, ingenuity, and innovation, many of the drawbacks of conventional QEC processes and quantum computer controllers configured to perform QEC have been solved by developing solutions constructed in accordance with embodiments of the present invention, many examples of which are detailed herein.

Summary of the Invention

Means for Solving the Problems

[0004] Exemplary embodiments provide methods, systems, apparatus, computer program products, etc., for characterizing the noise of a quantum processor, such that at least one component and / or parameter of the quantum processor and / or controller of a quantum computer may be modified and / or changed so as to reduce the overall noise of the quantum processor. In various embodiments, the noise of the quantum processor is characterized by a noise model. The noise model is generated based at least in part on a quantum error judgment model trained using machine learning techniques. The quantum error judgment model is trained using training data that includes empirical behavioral data of the quantum processor. In particular, the training data includes empirical behavioral data that characterizes the behavior of a particular quantum processor (e.g., of a particular instance of hardware and hardware configuration) in which at least one component and / or parameter will be modified, tuned and / or changed.

[0005] A first aspect of this disclosure provides a method for reducing noise present in computations performed by a particular quantum processor. In an exemplary embodiment, the method includes the steps of: training a quantum noise decoder, including a machine learning-trained quantum error judgment model, using training data, including operational data, incorporated at least in part on the operation of a particular quantum processor, by one or more processors; generating a noise model for a particular quantum processor based on the machine learning-trained quantum error judgment model, by one or more processors; and having one or more processors provide the noise model. Providing the noise model includes at least one of the following: (a) having a graphical representation of at least a portion of the noise model provided via a display of a computing entity, such that at least one component or parameter of a particular quantum processor is modified or changed based on a graphical representation of at least a portion of the noise model; or (b) providing at least a portion of the noise model as input related to executable instructions to be executed by a controller of a particular quantum processor or a computing entity communicating with the controller of a particular quantum processor, such that at least one component or parameter of a particular quantum processor is modified or changed based on at least a portion of the noise model.

[0006] In an exemplary embodiment, the operational data includes calibration data generated through the operation of a particular quantum processor.

[0007] In an exemplary embodiment, calibration data is periodically acquired during the operation of a particular quantum processor.

[0008] In an exemplary embodiment, the operation data includes spectator object data captured by direct or indirect observation of one or more spectator objects controlled by a specific quantum processor, the one or more spectator objects being controlled independently of a quantum algorithm being executed by the specific quantum processor.

[0009] In an exemplary embodiment, the quantum noise decoder includes a generative adversarial network (GAN) comprising a generator and a discriminator, the generator configured to generate simulated behavioral data.

[0010] In an exemplary embodiment, the discriminator includes or communicates with a quantum error judgment model trained by machine learning.

[0011] In an exemplary embodiment, the quantum noise decoder includes a noise model generation module configured to generate a noise model for a particular quantum processor based at least in part on the output of a quantum error judgment model trained by machine learning.

[0012] In an exemplary embodiment, at least one component or parameter is part of or used by a real-time quantum error decoder for correcting quantum errors during the operation of a particular quantum processor.

[0013] In an exemplary embodiment, at least one component or parameter is a hardware component or physical parameter of a particular quantum processor.

[0014] In exemplary embodiments, at least one component or parameter corresponds to the recalibration of a hardware component of a particular quantum processor or a software process of a controller of a particular quantum processor.

[0015] In an exemplary embodiment, the quantum noise decoder is a real-time quantum error decoder configured to cause quantum error correction during the operation of a particular quantum processor.

[0016] In an exemplary embodiment, the noise model characterizes the noise present in the operational data of a particular quantum processor.

[0017] In another embodiment, an apparatus is provided. In an exemplary embodiment, the apparatus includes at least one non-temporary memory for storing computer-executable instructions and a processing device. The computer-executable instructions are configured, when executed by the processing device, to cause the apparatus to at least train a quantum noise decoder including a machine learning-trained quantum error judgment model using training data including operation data incorporated at least in part based on the operation of a particular quantum processor; to generate a noise model for a particular quantum processor based on the machine learning-trained quantum error judgment model; and to provide the noise model. Providing the noise model includes at least one of (a) providing a graphical representation of at least a portion of the noise model via a display of a computing entity so that at least one component or parameter of a particular quantum processor is modified or changed based on a graphical representation of at least a portion of the noise model, or (b) providing at least a portion of the noise model as input related to an executable instruction to be executed by a controller of a particular quantum processor or a computing entity communicating with the controller of a particular quantum processor so that at least one component or parameter of a particular quantum processor is modified or changed based on at least a portion of the noise model.

[0018] In an exemplary embodiment, the operational data includes calibration data generated through the operation of a particular quantum processor.

[0019] In an exemplary embodiment, calibration data is periodically acquired during the operation of a particular quantum processor.

[0020] In an exemplary embodiment, the operation data includes spectrator object data captured by direct or indirect observation of one or more spectrator objects controlled by a specific quantum processor, the one or more spectrator objects being controlled independently of a quantum algorithm being executed by the specific quantum processor.

[0021] In an exemplary embodiment, the quantum noise decoder includes a generative adversarial network (GAN) comprising a generator and a discriminator, the generator configured to generate simulated behavioral data.

[0022] In an exemplary embodiment, the discriminator includes or communicates with a quantum error judgment model trained by machine learning.

[0023] In an exemplary embodiment, the quantum noise decoder includes a noise model generation module configured to generate a noise model for a particular quantum processor based at least in part on the output of a quantum error judgment model trained by machine learning.

[0024] In an exemplary embodiment, at least one component or parameter is part of or used by a real-time quantum error decoder for correcting quantum errors during the operation of a particular quantum processor.

[0025] In an exemplary embodiment, at least one component or parameter is a hardware component or physical parameter of a particular quantum processor.

[0026] In an exemplary embodiment, at least one component or parameter corresponds to recalibration of a hardware component of a particular quantum processor or a software process of a controller of a particular quantum processor.

[0027] In an exemplary embodiment, the quantum noise decoder is a real-time quantum error decoder configured to cause correction of quantum errors during operation of a particular quantum processor.

[0028] In an exemplary embodiment, the noise model characterizes noise present in the operation data of a particular quantum processor.

[0029] According to another aspect, a computer program product is provided. In an exemplary embodiment, the computer program product includes a non-transitory computer-readable medium storing computer-executable instructions. When the computer-executable instructions are executed by a processing device of the apparatus, the apparatus is caused to train a quantum noise decoder including a quantum error determination model trained by machine learning using training data including operation data captured at least in part based on the operation of a particular quantum processor, generate a noise model for the particular quantum processor based on the quantum error determination model trained by machine learning, and provide the noise model. Providing the noise model includes at least one of (a) causing at least a graphical representation of at least a part of the noise model to be provided via a display of a computing entity such that at least one component or parameter of the particular quantum processor is modified or changed based on the at least a graphical representation of at least a part of the noise model, or (b) providing at least a part of the noise model as an input related to executable instructions to be executed by a controller of the particular quantum processor or a computing entity communicating with the controller of the particular quantum processor such that at least one component or parameter of the particular quantum processor is modified or changed based on at least a part of the noise model.

[0030] In an exemplary embodiment, the operation data includes calibration data generated through the operation of a particular quantum processor.

[0031] In an exemplary embodiment, the calibration data is periodically captured during the operation of a particular quantum processor.

[0032] In an exemplary embodiment, the operation data includes spectrator object data captured by direct or indirect observation of one or more spectrator objects controlled by a specific quantum processor, the one or more spectrator objects being controlled independently of a quantum algorithm being executed by the specific quantum processor.

[0033] In an exemplary embodiment, the quantum noise decoder includes a generative adversarial network (GAN) comprising a generator and a discriminator, the generator configured to generate simulated behavioral data.

[0034] In an exemplary embodiment, the discriminator includes or communicates with a quantum error judgment model trained by machine learning.

[0035] In an exemplary embodiment, the quantum noise decoder includes a noise model generation module configured to generate a noise model for a particular quantum processor based at least in part on the output of a quantum error judgment model trained by machine learning.

[0036] In an exemplary embodiment, at least one component or parameter is part of or used by a real-time quantum error decoder for correcting quantum errors during the operation of a particular quantum processor.

[0037] In an exemplary embodiment, at least one component or parameter is a hardware component or physical parameter of a particular quantum processor.

[0038] In exemplary embodiments, at least one component or parameter corresponds to the recalibration of a hardware component of a particular quantum processor or a software process of a controller of a particular quantum processor.

[0039] In an exemplary embodiment, the quantum noise decoder is a real-time quantum error decoder configured to cause quantum error correction during the operation of a particular quantum processor.

[0040] In an exemplary embodiment, the noise model characterizes the noise present in the operational data of a particular quantum processor.

[0041] Having given a general overview of the present invention, the attached drawings, which are not necessarily drawn to the correct scale, will be referenced from here on. [Brief explanation of the drawing]

[0042] [Figure 1] This is a schematic diagram illustrating an exemplary quantum computing system including a quantum system controller according to an exemplary embodiment. [Figure 2] To provide a noise model for characterizing the noise of a particular quantum processor using a quantum noise decoder, according to various embodiments, the flowchart shows, for example, the processes, procedures, and / or operations performed by the controller in Figure 8 or the computing entity in Figure 9. [Figure 3] This flowchart illustrates various processes, operations, and / or procedures performed by, for example, the controller in Figure 8 or the computing entity in Figure 9, in order to obtain empirical operational data corresponding to the operation of a particular quantum processor in various embodiments. [Figure 4A] This flowchart illustrates various processes, operations, and / or procedures for providing a noise model that characterizes the noise of a particular quantum processor by operating a quantum noise decoder using, for example, the controller in Figure 8 or the computing entity in Figure 9, according to various embodiments. [Figure 4B] This block diagram schematically shows at least some of the architectures of quantum noise decoders according to various embodiments. [Figure 5A]This flowchart illustrates various processes, operations, and / or procedures for providing a noise model that characterizes the noise of a particular quantum processor by operating a quantum noise decoder, including a generative adversarial network (GAN), using various embodiments, for example, the controller in Figure 8 or the computing entity in Figure 9. [Figure 5B] This block diagram schematically shows at least some of the architectures of quantum noise decoders according to various embodiments. [Figure 6] This flowchart illustrates various processes, operations, and / or procedures performed by, for example, the controller in Figure 8 or the computing entity in Figure 9, in order to provide a noise model such that at least one component and / or parameter of a quantum processor in various embodiments is modified, changed, tuned, etc., based on the noise model. [Figure 7] This flowchart illustrates various processes, operations, and / or procedures performed by, for example, the controller in Figure 8 or the computing entity in Figure 9, in order to provide a noise model such that at least one component and / or parameter of a quantum processor in various embodiments is modified, changed, tuned, etc., based on the noise model. [Figure 8] This is a schematic diagram of exemplary controllers for quantum computers in various embodiments. [Figure 9] This is a schematic diagram of an exemplary computing entity of a quantum computer system that may be used in exemplary embodiments. [Modes for carrying out the invention]

[0043] The present invention will be more fully described below with reference to the accompanying drawings, which illustrate embodiments that are part of but not all of the present invention. Indeed, the present invention may be carried out in many different forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided to satisfy any legal requirements to which this disclosure may apply. The term “or” (also written as “ / ”) is used herein in both disjunctive and conjunctive senses unless otherwise indicated. The terms “explanatory” and “exemplary” are used to mean examples and do not indicate a level of quality. The terms “generally,” “substantially,” and “approximately” mean, unless otherwise indicated, within engineering and / or manufacturing tolerances, and / or within the user’s measuring ability. Throughout, similar numbers refer to similar elements.

[0044] Exemplary embodiments provide methods, systems, apparatus, computer program products, etc., for characterizing the noise of a quantum processor such that at least one component and / or parameter of the quantum processor and / or controller of a quantum computer may be modified and / or changed so as to reduce the overall noise of the quantum processor. In various embodiments, the noise of the quantum processor is characterized by a noise model. The noise model is generated based at least in part on a quantum error judgment model trained using machine learning techniques. The quantum error judgment model is trained using training data that includes empirical behavioral data of the quantum processor. In particular, the training data includes empirical behavioral data that characterizes the behavior of a particular quantum processor (e.g., a particular instance of hardware and hardware configuration) in which at least one component and / or parameter will be modified, tuned and / or changed.

[0045] Large-scale quantum computers are expected to solve problems in fields such as chemistry, materials science, and biology that are currently beyond the capabilities of modern technology. Solving such problems involves computations employing quantum algorithms implemented using deep quantum circuits. Achieving the required level of precision in these deep circuits requires a high level of reliability in quantum operations. To achieve such reliability, quantum error correction (QEC) is employed during computation to suppress noise to the required level. However, understanding the noise present in a particular quantum processor is helpful in suppressing the noise present within that particular quantum processor to the required level.

[0046] When used herein, a particular quantum processor corresponds to a particular instance of hardware and its configuration for providing a particular quantum processor. For example, in the field of quantum charge-coupled device (QCCD) based quantum computing, a particular quantum processor corresponds to a particular ion trap, a magnetic field generating component, a manipulation source (e.g., a laser), and an optical path defined to provide a manipulation signal (e.g., a laser beam) to each of the particular ion traps. For example, if a particular mirror or lens in the optical path is slightly misaligned, or if the optical fiber defining part of the optical path is nearly burnt out, the manipulation signal provided along the optical path may carry less optical power than expected and / or result in an uncompensated shift in the optical phase of the manipulation signal. Thus, the function performed using the optical path contributes to the noise of the particular quantum processor. However, a second quantum processor of a similar design does not suffer its particular contribution to the noise of the second quantum processor.

[0047] Current techniques for managing noise in quantum processors include the use of real-time quantum error decoders. However, due to the time constraints of performing real-time quantum error correction while executing quantum circuits, current real-time quantum error decoders tend to be simple programs that rely on algorithms such as the blossom algorithm or Dijkstra's algorithm. These real-time quantum error decoders generally fail to provide a broader characterization of the noise of a particular quantum processor and may rely on general noise models that cannot characterize the noise contributions that are different and / or specific to the particular quantum processor to which the real-time quantum error decoder is associated. Thus, there are technical problems in the field of characterizing the noise of quantum processors and performing quantum error correction (including real-time quantum error correction) of quantum processors using noise models that characterize the noise of quantum processors.

[0048] Various embodiments provide technical solutions to these technical problems. In particular, various embodiments provide a quantum noise decoder that includes a machine learning-based quantum error judgment model, which is trained using machine learning techniques to characterize the noise of a particular quantum processor. The quantum noise decoder (e.g., the machine learning-based quantum error judgment model) is trained using operation data (e.g., empirical operation data) corresponding to and / or generated during the operation of a particular quantum processor. The quantum noise decoder is then used to generate a noise model that characterizes the noise of a particular quantum processor.

[0049] Based on noise related to a particular quantum processor, at least one component and / or parameter of the quantum processor may be modified, tuned, or altered to reduce noise in computations performed by that particular quantum processor. For example, a graphical representation of at least a portion of the noise model for a particular quantum processor may be displayed by a graphical user interface (GUI) provided via a display of the computing entity, so that a human engineer may modify, tune, or alter at least one component and / or parameter. For example, a human engineer may modify or tune the physical components and / or parameters of a particular quantum processor. For example, a human engineer may replace a burnt-out optical fiber, adjust the alignment of a mirror or lens, etc., based on the noise contribution of a particular quantum processor identified and / or indicated by the noise model. For example, a quantum computer controller may modify, tune, or alter at least one component and / or parameter of a quantum processor based on the noise model (for example, it may tune hardware components and / or parameters, software components and / or parameters, and / or calibration components and / or parameters). In an exemplary embodiment, a noise model is provided to a real-time quantum error decoder for use when performing real-time quantum error correction for a particular quantum processor. In various embodiments, real-time quantum error correction includes tracking one or more quantum errors, phase shifts, etc., in software, and physically applying quantum error correction to the appropriate qubit at the appropriate time during execution of the quantum circuit.

[0050] Accordingly, various embodiments provide methods, apparatus, systems, computer program products, etc., for determining a noise model that characterizes the noise of a particular quantum processor. Various embodiments also provide methods, apparatus, systems, computer program products, etc., for reducing the noise of a particular quantum processor based on the determined noise model that characterizes the noise of a particular quantum processor. Accordingly, various embodiments provide practical applications that offer technical solutions and technical advantages to quantum computing, including areas such as quantum error correction, real-time quantum error correction, and quantum processor noise reduction.

[0051] In this specification, various embodiments are described in detail with respect to QCCD-based quantum processors. However, those skilled in the art will understand, based on the disclosures provided herein, that various embodiments may be used to characterize and / or reduce noise in various types of quantum processors (including, but not limited to, superconducting quantum processors (using Josephson junctions as qubits), neutral atom quantum processors in optical lattice quantum processors, spin-based or space-based quantum dot quantum processors, nuclear magnetic resonance quantum processors, etc.).

[0052] Exemplary quantum computing system including an atomic object confinement device Figure 1 provides a schematic diagram of an exemplary quantum computing system 100. The quantum computing system 100 includes one or more computing entities 10 and a quantum computer 110. The quantum computer includes a controller 30 and a quantum processor 115. In various embodiments, the controller 30 is programmed and / or configured to control the operation of various components, assemblies, elements, etc. of the quantum processor 115. The computing entities 10 communicate with the controller 30 of the quantum computer 110 via wired and / or wireless means.

[0053] In the illustrated embodiment, the quantum processor 110 is a QCCD-based quantum computer, and the quantum processor 115 includes an atomic object confinement device 120 (e.g., an ion trap) that confines a plurality of atomic objects (e.g., atoms, ions, etc.). In an exemplary embodiment, the quantum processor 115 includes a plurality of qubits (e.g., data qubits, which may be organized into logical qubits, ancilla qubits, etc.). For example, at least a portion of the atomic objects (e.g., atoms, ions, etc.) confined by the atomic object confinement device 120 (e.g., an ion trap, etc.) are used as qubits of the quantum processor 115.

[0054] In various embodiments, the quantum processor 115 includes means for controlling the evolution of quantum states of qubits. For example, in an exemplary embodiment, the quantum processor 115 includes a cryostat and / or vacuum chamber 40 surrounding a confinement device 120 (e.g., an ion trap), one or more operating sources 60, one or more voltage sources 50, and / or one or more optical collection systems 70. For example, the cryostat and / or vacuum chamber 40 may be a chamber with controlled temperature and / or pressure. In an exemplary embodiment, one or more operating sources 60 may include one or more lasers (e.g., optical lasers, microwave sources, etc.). In various embodiments, one or more operating sources 60 are configured to manipulate and / or induce controlled quantum state evolution of one or more atomic objects in the confinement device. In various embodiments, the atomic objects in the confinement device (e.g., ions trapped in an ion trap) act as data qubits and / or auxiliary qubits of the quantum processor 115 of the quantum computer 110. For example, in an exemplary embodiment in which one or more operating sources 60 include one or more lasers, the lasers may provide one or more laser beams to an atomic object trapped in a cryostat and / or confinement device 120 within a vacuum chamber 40. For example, the operating source 60 may generate and / or provide laser beams configured to ionize the atomic object, initialize the atomic object in a defined two-state qubit space of a quantum processor, execute a gate on one or more qubits of the quantum processor, read the quantum state of one or more qubits of the quantum processor, and so on.

[0055] In various embodiments, the quantum processor 115 includes an optical collection system 70 configured to collect and / or detect photons generated by the qubits (for example, during a read procedure). The optical collection system 70 may include one or more optical elements (e.g., lenses, mirrors, waveguides, optical fiber cables, etc.) and one or more photodetectors. In various embodiments, the photodetectors may be photodiodes, photomultiplier tubes, charge-coupled device (CCD) sensors, complementary metal-oxide-semiconductor (CMOS) sensors, micro-electromechanical system (MEMS) sensors, and / or other photodetectors that are highly sensitive to light of the expected fluorescence wavelength of the qubits of the quantum processor 115. In various embodiments, the detectors may communicate electronically with the controller 30 via one or more A / D converters 825 (see Figure 8), etc.

[0056] In various embodiments, the quantum processor 115 includes one or more voltage sources 50. For example, the voltage source 50 may include a plurality of voltage drivers and / or voltage sources, and / or at least one RF driver and / or voltage source. In exemplary embodiments, the voltage source 50 may be electrically coupled to a corresponding potential generating element (e.g., an electrode) of the confinement device 120.

[0057] In various embodiments, the computing entity 10 is configured to allow a user to provide input to the quantum computer 110 (for example, through the user interface of the computing entity 10) and receive, view, etc., outputs from the quantum computer 110. In various embodiments, the computing entity 10 is configured to train a quantum noise decoder and / or communicate with the quantum noise decoder. For example, in various embodiments, the computing entity 10 is configured to provide empirical operational data captured by one or more sensors coupled to a particular quantum processor 115 to the quantum noise decoder and to receive the output of the quantum noise decoder, which includes a noise model for the particular quantum processor 115. In an exemplary embodiment, the computing entity 10 is configured to provide a noise model such that at least one component and / or parameter of a particular quantum processor 115 is modified, tuned, changed, etc., based on the noise model.

[0058] Computing entities 10 may communicate with the controller 30 of the quantum computer 110 and / or other computing entities 10 via one or more wired or wireless networks 20, and / or via direct wired and / or wireless communication. In exemplary embodiments, computing entities 10 may convert, configure, format, etc., information / data, quantum computing algorithms and / or circuits, etc., into a computing language, executable instructions, command sets, etc., that the controller 30 can understand and / or implement.

[0059] In various embodiments, the controller 30 is configured to control a voltage source 50, a cryostat system and / or vacuum system that controls the temperature and pressure within the cryostat and / or vacuum chamber 40, an operating source 60, and / or other systems that control various environmental conditions within the cryostat and / or vacuum chamber 40 (e.g., temperature, pressure, magnetic field, etc.), and / or to manipulate and / or induce a controlled evolution of the quantum state of one or more atomic objects within the confinement device. For example, the controller 30 may induce a controlled evolution of the quantum state of one or more atomic objects within the confinement device 120 in order to execute a quantum circuit and / or algorithm. For example, the controller 30 may cause a reading procedure, possibly including coherent shelving, to be executed, possibly as part of the execution of a quantum circuit and / or algorithm.

[0060] Furthermore, the controller 30 is configured to transmit and / or receive input data from the optical acquisition system 70 corresponding to reading the quantum states of the qubits of the quantum processor 115. In various embodiments, the controller 30 is configured to control the calibration of one or more components and / or parameters of the quantum processor 115. In various embodiments, the controller 30 is configured to modify, adjust, or change one or more hardware, software, calibration, and / or operating components and / or parameters of the quantum processor 115, at least in part, based on processing and / or analyzing a noise model of the quantum processor 115.

[0061] Exemplary operation and use of quantum noise decoders In various embodiments, a quantum noise decoder including a machine learning-based quantum error judgment model is provided and / or used to modify, adjust, change, etc., at least one component and / or parameter of a particular quantum processor 115 based on a noise model for that particular quantum processor provided, generated and / or determined by the quantum noise decoder. In various embodiments, the quantum noise decoder is trained and / or operated by a computing entity 10 (e.g., through the execution of computer-executable instructions by processing device 908) and / or by a controller 30 (e.g., through the execution of computer-executable instructions by processing device 805).

[0062] In various embodiments, the quantum noise decoder is trained using operational data corresponding to the operation of a particular quantum processor 115. The particular quantum processor 115 is a quantum processor controlled by a controller 30. In various embodiments, training the quantum noise decoder includes training a quantum error judgment model using machine learning techniques. In various embodiments, the quantum error judgment model is trained using training data. The training data includes empirical operational data corresponding to the operation of a particular quantum processor 115.

[0063] In various embodiments, empirical operation data includes circuit execution data generated during the execution of a quantum circuit by a particular quantum processor 115. For example, while a particular quantum processor 115 is executing a quantum circuit, one or more sensors coupled to the quantum processor 115 (e.g., communicating with a controller 30) capture and provide the circuit execution data to the controller 30. In various embodiments, the circuit execution data includes the results of performing read operations on one or more qubits of the quantum processor, optical power indications showing the optical power of various operational signals applied to one or more qubits during the execution of the quantum circuit, and characterization of the circuit's performance.

[0064] In various embodiments, empirical operation data includes calibration data generated by performing calibration on a particular quantum processor 115. For example, the calibration process may be triggered at one or more set points before execution of a quantum circuit, after execution of a quantum circuit, at one or more set points during execution of a quantum circuit, and / or periodically during execution of a quantum circuit. During an exemplary calibration process, one or more set operations are performed and sensors coupled to the particular quantum processor 115 capture calibration data. For example, the power of a specific operational signal at a specific location along the optical path may be measured, the electric field generated by applying a specific voltage signal to one or more potential generating elements (e.g., electrodes) of the confinement device 120 may be measured and / or determined, and the alignment of various components of the particular quantum processor 115 (e.g., defining the optical path) may be checked. Such a calibration process results in the generation of calibration data that, in various embodiments, is included in the empirical operation data corresponding to the operation of the particular quantum processor 115. Some non-limiting examples of calibration processes include one-qubit gate fidelity tests, two-qubit gate fidelity tests, measurements of magnetic field fluctuations at one or more locations of the confinement device 120 over a period of time, and phase relaxation noise of atomic objects (e.g., qubits).

[0065] In various embodiments, calibration data is captured by one or more (classical) sensors and includes data characterizing the environment at one or more locations in the confinement device (e.g., magnetic field, temperature, pressure, ambient light, electric field, voltage changes across a portion of the surface of the confinement device 120). For example, one or more magnetometers, voltage sensors, piezoelectric thermal and / or pressure sensors, etc., coupled to and / or communicating with the environment surrounding the confinement device 120 (e.g., within the cryostat and / or vacuum chamber 40) are used to capture at least a portion of the calibration data.

[0066] In various embodiments, calibration data includes spectator object data. In various embodiments, one or more spectator objects are confined by the confinement device 120. As used herein, a spectator object is an atomic object (e.g., an atom, ion, etc.) that is not used as a qubit in the quantum processor 115 and is not used as a sympathetic cooling atomic object in the quantum processor 115 (for example, configured to be used when laser cooling the corresponding qubit by sympathetic cooling). In exemplary embodiments, one or more spectator objects are different chemical species from the atomic objects used as qubits and / or sympathetic cooling atomic objects in the quantum processor 115. In exemplary embodiments, the spectator objects include atomic objects of one or more chemical species that may be highly sensitive to various environmental properties (e.g., magnetic field strength, magnetic field fluctuations / noise, potential fluctuations / noise, temperature, temperature fluctuations / noise, etc.).

[0067] In various embodiments, spectator objects are used to explore various aspects of the operation of a particular quantum processor 115. For example, a calibration process may include performing one or more functions on one or more spectator objects confined by a confinement device 120 and measuring the response of one or more spectator objects to the performance of one or more functions. For example, one or more operational signals may be incident on a spectator object or a group of two or more spectator objects, and any fluorescence (e.g., light emitted by a spectator object in response to the incident of one or more operational signals on a spectator object) may be captured and / or measured. In another example, the movement of one or more spectator objects within the confinement device 120 as a result of a voltage signal applied to a potential generating element (e.g., an electrode) of the confinement device 120 may be determined and / or measured. In various embodiments, data captured regarding the response of one or more spectator objects to the performance of various functions on one or more spectator objects is referred to herein as spectator object data. In various embodiments, the calibration data includes spectator object data. In various embodiments, the spectator object data is used to supplement and / or as part of the calibration data.

[0068] A quantum noise decoder is configured to receive empirical operation data (e.g., captured during operation) corresponding to the operation of a particular quantum processor 115, and to generate and / or determine and provide a noise model based at least in part on that empirical operation data. The noise model characterizes the noise of the particular quantum processor 115 present in the operation data. In various embodiments, the noise model represents a probability distribution of noise characteristics on a multivariate time series of input data, which may be used, for example, to determine when a particular quantum processor 115 is operating within or outside set limits. For example, the noise model may be used to identify anomalies in the operation of a particular quantum processor 115. For example, in an exemplary embodiment, the noise model is an anomaly detection model configured to determine when a particular quantum processor 115 is operating within statistically normal limits or outside predetermined normal limits. For example, in various embodiments, the noise model is a time and / or spatially parameterized distribution of how noise affects and / or is added to computations performed by a particular quantum processor 115. For example, the noise model may include the frequency profile of noise present in the electrical signals applied to the potential-generating elements of the confinement device 120, the wavelength / frequency fluctuations, phase shifts, and / or fluctuations of optical power of various operating signals, the magnitude, direction, and / or frequency profile of magnetic field fluctuations at one or more locations within the confinement device 120, and fluctuations in the quantum state indications of a cohort of physical qubits used as logical qubits and / or a cohort of spectator objects. In various embodiments, the noise model is parameterized at least spatially and / or temporally. For example, different and / or independent noise profiles may be associated with different zones of the confinement device 120. For example, the evolution of noise at one or more locations within the confinement device 120 over time may be determined and / or tracked.

[0069] In various embodiments, the noise model may show trends in various noise types and / or contributors over time. For example, empirical operation data may correspond to the operation of a particular quantum processor 115 over a first period of time, and the noise model may show how the noise of that particular quantum processor 115 evolved over the first period of time and / or how the noise of that particular quantum processor 115 is expected to evolve in a second period of time (preceding or following the first period of time).

[0070] In various embodiments, the noise model includes a description of the noise present in the operation of various subsystems and / or assemblies of a particular quantum processor 115. For example, the noise model may include noise profiles of wavelength / frequency fluctuations, phase shifts, and / or optical power fluctuations of the operational signal used to perform a two-qubit gate, and noise profiles of wavelength / frequency fluctuations, phase shifts, and / or optical power fluctuations of the operational signal used to perform a qubit read operation. In various embodiments, the noise model may include indications of the source or inducement of the profile features. For example, a quantum error judgment model trained by machine learning is trained, in an exemplary embodiment, to identify instances where the noise profile of a particular subsystem of quantum processor 115 exceeds a baseline noise amplitude and contains identifiable features (e.g., the phase shift profile includes a peak of above-average amplitude at a point temporally corresponding to above-average optical power fluctuations). For example, if the probability of adding various noises increases with the execution time of a particular quantum processor 115 (e.g., due to a correlation between noise amplitude and execution time), the noise may be increasing due to heating. In another example, one or more measurements of one or more spectator objects may indicate that one or more quantum operations have an increased or decreased probability of adding a particular type of noise. Based on subsystem noise profiles and / or correlations between subsystem noise profiles, machine learning-trained quantum error judgment models and / or noise model generation modules are configured to identify likely noise sources in a particular quantum processor subsystem.

[0071] In various embodiments, a noise model is provided for a particular quantum processor 115 such that at least one component and / or parameter of the quantum processor may be modified, tuned, or changed in order to reduce the noise of the computation performed by the particular quantum processor.

[0072] For example, a graphical representation of at least a portion of the noise model for a particular quantum processor may be displayed via a graphical user interface (GUI) provided through the display of the computing entity, so that a human engineer may modify, adjust, or change at least one component and / or parameter. For example, a human engineer may replace a burnt-out optical fiber, adjust the alignment of a mirror or lens, etc., based on the noise model's contribution to the noise of a particular quantum processor.

[0073] For example, a quantum computer controller may modify, adjust, or change at least one component and / or parameter of a quantum processor based on a noise model (e.g., hardware components and / or parameters, software components and / or parameters, and / or calibration components and / or parameters). In an exemplary embodiment, the noise model is provided to a real-time quantum error decoder for use when performing real-time quantum error correction for a particular quantum processor. For example, one or more parameters, weights, etc., of a quantum error decoder configured to perform quantum error correction for a particular quantum processor 115 may be updated, modified, or changed based on the noise model. In an exemplary embodiment, one or more calibration processes may be executed more regularly / less regularly (e.g., according to shorter / longer periodicity), executed more / less often each time a process is triggered, and one or more new calibration processes may be defined at least partially based on the noise model. In exemplary embodiments, techniques for performing functions of a quantum computer (e.g., performing one or two qubit gates, performing transfer operations, performing read operations, etc.) may be modified, updated, or altered based on a noise model to reduce noise present in the computations performed by a particular quantum processor 115.

[0074] Determine a noise model for a specific quantum processor and use the noise model to improve the functionality of that specific quantum processor. Figure 2 provides a flowchart illustrating various processes, procedures, and operations for using a quantum noise decoder to determine a noise model for a particular quantum processor 115 and for using the noise model to improve the functionality of the particular quantum processor 115 (for example, by reducing the noise present in the computations performed by the particular quantum processor 115). In various embodiments, the processes, procedures, and operations shown in Figure 2 are performed by the computing entity 10 (for example, through the execution of computer-executable instructions by processing device 908) and / or by the controller 30 (for example, through the execution of computer-executable instructions by processing device 805).

[0075] Starting from step / operation 202, operational data for a specific quantum processor is acquired. For example, a processing device 805 of the controller 30 (see Figure 8) or a processing device 908 of the computing entity 10 (see Figure 9) acquires operational data for a specific quantum processor 115. In various embodiments, operational data is acquired by accessing operational data from memories 810, 922, 924. In various embodiments, operational data is acquired by receiving operational data via a communication interface 820, one or more A / D converters 825, a network interface 920, a receiver 906, etc. In an exemplary embodiment, the controller 30 and / or computing entity 10 may cause a specific quantum processor 115 to perform one or more calibration processes and / or execute at least a portion of a quantum circuit, and receive operational data generated as a result of and / or during the execution of one or more calibration processes, and / or as a result of and / or during the execution of at least a portion of a quantum circuit. In various embodiments, the acquired operational data is empirical operational data corresponding to the operation of a particular quantum processor and therefore includes noise signatures and / or noise profiles specific to a particular quantum processor 115.

[0076] In step / operation 204, operation data is provided to the quantum noise decoder. For example, in an exemplary embodiment, the operation data is kept available for use by the quantum noise decoder so that the quantum noise decoder can read the operation data. In an exemplary embodiment, the operation data is provided to the quantum noise decoder via an application programming interface (API) call. One or more modules of the quantum noise decoder then use the operation data to train a machine learning-based quantum error judgment model and generate a noise model (e.g., a specific noise signature and / or noise profile for a particular quantum processor) that characterizes the noise of a particular quantum processor 115. For example, processing devices 805, 908 may execute computer-executable instructions that cause the quantum noise decoder to provide the operation data so that the quantum noise decoder uses the operation data to generate and / or determine a noise model (e.g., a specific noise signature and / or noise profile for a particular quantum processor) that characterizes the noise of a particular quantum processor 115.

[0077] In step / operation 206, the output from the quantum noise decoder is received. The output from the quantum noise decoder includes a noise model for a specific quantum processor 115. For example, the output including the noise model for a specific quantum processor 115 may be stored in memories 810, 922, and 924 and accessed by processing devices 805 and 908. For example, the output including the noise model for a specific quantum processor 115 may be provided to processing devices 805 and 908 via an API call or response.

[0078] In step / operation 208, at least a portion of a noise model is provided so that at least one component and / or parameter related to the operation of a particular quantum processor may be modified, adjusted, changed, etc., based at least partially on the noise model. For example, the noise model may be provided in various embodiments via a communication interface 820, a network interface 920, a transmitter 904, and / or a display 916. For example, a graphical representation of at least a portion of the noise model for a particular quantum processor may be displayed by a graphical user interface (GUI) provided via the display 916 of the computing entity 10, so that a human technician may modify, adjust, change, etc., at least one component and / or parameter related to the operation of a particular quantum processor. For example, a human technician may replace a burnt-out optical fiber, adjust the alignment of a mirror or lens, etc., based on the contribution of a particular quantum processor to noise identified and / or indicated by the noise model.

[0079] In exemplary embodiments, processing devices 805, 908 may provide at least a portion of the noise model as input to programs, modules, applications, etc., running on the processing devices 805, 908. For example, a calibration manager may receive the noise model as input and, based on the results of processing and / or analyzing the noise model, modify, adjust, or change components and / or parameters of the calibration process, or generate a new calibration process.

[0080] For example, the controller 30 of the quantum computer may modify, adjust, or change at least one component and / or parameter of the quantum processor based on the noise model (e.g., hardware components and / or parameters, software components and / or parameters, and / or calibration components and / or parameters). In an exemplary embodiment, the noise model is provided to a real-time quantum error decoder (e.g., operating on the controller 30) for use when performing real-time quantum error correction for a particular quantum processor 115. In an exemplary embodiment, techniques for performing functions of the quantum computer (e.g., performing one or two qubit gates, performing transfer operations, performing read operations, etc.) may be modified, updated, or changed based on the noise model and / or the results of processing and / or analyzing the noise model to reduce noise present in the computations performed by the particular quantum processor 115. In an exemplary embodiment, a new calibration process may be used based on the noise model and / or the results of processing and / or analyzing the noise model to ensure proper functioning of a particular component, element, assembly, etc. of the particular quantum processor 115.

[0081] In an exemplary embodiment, the controller 30 and / or computing entity 10 processes a noise model to determine whether there are any components and / or parameters that may be automatically modified, adjusted, or changed in an attempt to reduce the noise afflicted by a particular quantum processor 115 (for example, by reducing the amplitude of the noise in one or more noise profiles provided by the noise model, or by reducing the presence of certain features present in one or more noise profiles provided by the noise model). In an exemplary embodiment, when it is determined that automated modifications, adjustments, or changes may be made to one or more components and / or parameters, the controller 30 and / or computing entity 10 may have at least one of the one or more components and / or parameters modified, adjusted, or changed accordingly, and may provide human-perceptible notifications and / or requests for permission for automated execution such as modifications, adjustments, or changes (for example, via the display 916), and / or update a log with information about the automated modifications, adjustments, or changes that have been performed.

[0082] In various embodiments, when no automated modifications, adjustments, or changes are identified for at least one component and / or parameter, and when possible manual modifications, adjustments, or changes are identified for at least one component and / or parameter, a graphical representation of at least a portion of the noise model (which may, for example, show the identified possible manual modifications, adjustments, or changes) is displayed for human user review (e.g., via display 916). In an exemplary embodiment, regardless of any identified possible automated and / or manual modifications, adjustments, or changes for one or more components and / or parameters of a particular quantum processor 115, a graphical representation of at least a portion of the noise model is displayed for human user review (e.g., via display 916).

[0083] In step / operation 210, the noise model may be stored in memories 810, 922, and 924. For example, the controller 30 and / or computing entity 10 may store the noise model for future use. For example, the noise model may be accessed from memory at a later point in time, for example, to be referenced by the (real-time) quantum error decoder of the quantum computer 110, for comparison with a newly determined noise model.

[0084] Obtaining a sample of operational data Figure 3 provides a flowchart illustrating various processes, procedures, and operations performed by the controller 30 and / or computing entity 10 as part of acquiring empirical operational data for a particular quantum processor in various embodiments. For example, one or more of the steps / operations shown in Figure 3 may be performed as part of step / operation 202 in Figure 2 in various embodiments.

[0085] Starting from step / operation 302, circuit execution data generated during the operation of a particular quantum processor is received. For example, in various embodiments, the circuit execution data is received by the controller 30 via the A / D converter 825 and / or the communication interface 820. For example, in various embodiments, the circuit execution data is received by the computing entity 10 via the network interface 920 and / or the receiver 906.

[0086] Circuit execution data is generated and / or captured by one or more sensors coupled to a specific quantum processor and configured to capture various measurements related to the operation of that particular quantum processor (e.g., electric fields generated in response to a series of voltage signals applied to potential generating elements (e.g., electrodes), optical power along a specific optical path, fluorescence of qubits or spectator objects, etc.). In various embodiments, the circuit execution data is stored in memories 810, 922, and 924.

[0087] In step / operation 304, calibration is triggered. For example, calibration may be triggered periodically, in response to a determination that an element of circuit execution data is outside a specified range. For example, the controller 30 and / or computing entity 10 may trigger calibration.

[0088] In various embodiments, triggering a calibration involves initiating a calibration process. For example, the controller 30 and / or computing entity 10 may initiate one or more calibration processes (e.g., suspending at least a portion of the execution of a quantum circuit, executing routine and / or scripted calibration processes, and / or generating corresponding calibration data). For example, the controller 30 may determine that a significant shift in qubit frequency has occurred since the last calibration cycle. This may indicate a change in the magnetic field in at least a portion of the confinement device 120, which may be used to trigger one or more calibration processes (e.g., to determine whether and / or the extent of the change in the magnetic field).

[0089] In various embodiments, performing a calibration process involves performing the operation multiple times so that the probability distribution and / or statistical analysis of the results of the operation may be determined. In another example, one or more environmental properties may be checked to detect changes in environmental properties over a period of time. For example, the magnetic field at one or more locations within the confinement device may be checked to determine whether the magnetic field has changed.

[0090] In step / operation 306, as a result of triggering calibration, the controller 30 and / or computing entity 10 generate and / or acquire calibration data. For example, one or more sensors coupled to a particular quantum processor acquire calibration data during and / or as part of one or more calibration processes. In exemplary embodiments, the calibration data is received by the controller 30 via the A / D converter 825 and / or the communication interface 820 in various embodiments. In exemplary embodiments, the calibration data is received by the computing entity 10 via the network interface 920 and / or the receiver 906 in various embodiments. In various embodiments, the calibration data is stored in memories 810, 922, and 924.

[0091] In various embodiments, spectator object data is collected as part of one or more calibration processes. For example, the controller 30 may be configured to trigger the execution of one or more calibration processes by the quantum processor 115, which include the acquisition of spectator object data. In step / operation 308, as a result of triggering the calibration, the controller 30 and / or computing entity 10 cause the spectator object data to be generated and / or acquired. For example, one or more sensors coupled to a particular quantum processor acquire spectator object data during and / or as part of one or more calibration processes. In exemplary embodiments, the spectator object data is received by the controller 30 via the A / D converter 825 and / or the communication interface 820 in various embodiments. In exemplary embodiments, the spectator object data is received by the computing entity 10 via the network interface 920 and / or the receiver 906 in various embodiments. In various embodiments, the spectator object data is stored in memories 810, 922, and 924.

[0092] In various embodiments, operational data (e.g., circuit execution data, calibration data, and / or spectator object data) may be obtained by accessing operational data from memories 810, 922, and 924.

[0093] Exemplary use of a quantum noise decoder to determine a noise model for a specific quantum processor. In various embodiments, the quantum noise decoder is configured to take operational data corresponding to and / or captured during the operation of a particular quantum processor 115 as input and to provide an output containing a noise model that characterizes the noise present in the computation performed by the particular quantum processor 115. In various embodiments, the quantum noise decoder includes a quantum error decision model, which is a model trained by machine learning. At least a portion of the training data used to train the machine learning-trained quantum error decision model is empirical operational data corresponding to the operation of a particular quantum processor 115. Thus, the quantum error decision model is configured and / or trained in particular to determine the noise profile of a particular quantum processor 115 and / or to identify (potential) noise sources and / or inducers.

[0094] In various embodiments, the quantum error detection model includes one or more neural networks. In various embodiments, the quantum error detection model includes one or more deep neural networks (DNNs). In various embodiments, the quantum error detection model is and / or includes one or more of the following neural networks: classifier DNNs, convolutional neural networks (CNNs), recurrent neural networks (RNNs), long short-term memory networks (LSTMs), modular neural networks, and / or other architectures. In various embodiments, the quantum error detection model includes support vector machines, kernel-based models (for example, one-class support vectors configured to distinguish between "normal" and "abnormal" operation of a particular quantum processor 115), etc. In exemplary embodiments, the quantum error detection model is trained using supervised machine learning techniques.

[0095] In various embodiments, the quantum noise decoder further includes a noise model generation module. In various embodiments, the noise model generation module is configured to convert, transform, format, compile, and / or configure the output of a quantum error judgment model into a noise model understandable to the controller 30 and / or computing entity 10 of the quantum computing system 100. In various embodiments, the noise model generation module is operated by and / or by the execution of classically programmed computer-executable instructions (e.g., via processing devices 805, 908). In various embodiments, the noise model generation module may optionally include, in addition to classically programmed computer-executable instructions, one or more machine learning-trained models (e.g., neural networks).

[0096] Figure 4B shows at least some exemplary architectures of the exemplary quantum noise decoder 400, and Figure 4A provides flowcharts illustrating various processes, procedures, and operations that use the quantum noise decoder 400 to generate and / or provide noise models for a particular quantum processor 115.

[0097] As shown in Figure 4B, the exemplary quantum noise decoder 400 includes a quantum error-decision model 420 and a noise model generation module 430. The quantum error-decision model 420 is configured to receive an input 442 (including, for example, operation data corresponding to the operation of a particular quantum processor 115). In an exemplary embodiment, the quantum error-decision model 420 includes one or more neural networks (e.g., DNNs) and is configured to receive the input 442 through one or more input layers of the one or more neural networks. For example, the quantum error-decision model 420 includes, for each of its DNNs, an input layer, one or more hidden layers, and an output layer. The nodes of the input layer of each DNN of the quantum error-decision model are linked by their respective weights to the nodes of the first hidden layer of each DNN. The nodes of the first hidden layer are linked by their respective weights to the nodes of subsequent hidden layers of each DNN, and the nodes of the last hidden layer are linked by their respective weights to the nodes of the output layer of each DNN. Each weight is determined by machine learning techniques and / or processes. In an exemplary embodiment, the machine learning technique and / or process is iterative, so as new (empirical) behavioral data is generated (e.g., by the operation of a particular quantum processor) and / or provided to the quantum noise decoder 400, the quantum error judgment model 420 is continuously trained.

[0098] The quantum error judgment model 420 is configured to provide a raw noise model 444 through one or more output layers of its one or more neural networks. The noise model generation module 430 receives the raw noise model 444 and is configured to convert, transform, format, compile, and / or configure the raw noise model 444 into a noise model 446 that is understandable to the controller 30 and / or computing entity 10.

[0099] The quantum noise decoder 400 then provides an output including a noise model 446. This output may be received by one or more applications, programs, modules, etc., running on the controller 30 and / or computing entity 10.

[0100] For example, as shown in Figure 4A, the generation of a noise model 446 for a particular quantum processor 115, in an exemplary embodiment, includes training a quantum error-decision model 420 using operation data in step / operation 402. For example, a quantum noise decoder 400 may receive an input 442 containing (empirical) operation data corresponding to the operation of a particular quantum processor 115. The quantum error-decision model 420 is then trained using at least a portion of the input 442 containing (empirical) operation data corresponding to the operation of a particular quantum processor 115.

[0101] For example, machine learning techniques may be used to train a quantum error-decision model 420 using training data that includes empirical operational data corresponding to the operation of a particular quantum processor 115. In various embodiments, training may be an initial training of a quantum error-decision model in which the initial weights of one or more DNNs are set randomly or to selected (e.g., untrained) values. In various embodiments, training may be a continuation of training an already trained quantum error-decision model 420 (e.g., using a new batch of training data that includes empirical operational data) in which the initial weights of one or more DNNs are set to previously trained values. For example, the quantum error-decision model 420 is trained iteratively in an exemplary embodiment.

[0102] Once the training criteria for the quantum error judgment model 420 are met (for example, when the loss function used in the machine learning technique is minimized), the raw noise model 444 is read from and / or extracted from the output layer of the quantum error judgment model 420.

[0103] In step / operation 404, the noise model generation module 430 is executed to generate a noise model 446 based on the raw noise model 444 read from and / or extracted from the output layer of the quantum error judgment model 420. For example, the noise model generation module 430 converts, deforms, formats, compiles, and / or configures the raw noise model 444 into a noise model 446 that is understandable to the controller 30 and / or computing entity 10.

[0104] In step / operation 406, the quantum noise decoder 400 provides an output containing a noise model 446 for a particular quantum processor 115. For example, an output containing a noise model 446 for a particular quantum processor 115 may be provided by the quantum noise decoder 400 via an API call or API response (for example, when the output is provided in response to an API call providing an input 442) to an application, program, module, etc., being executed by processing devices 805, 908.

[0105] In various embodiments, the quantum noise decoder 400 may have various architectures. In an exemplary embodiment, the quantum noise decoder 400 includes a generative adversarial network (GAN) and / or GAN machine learning techniques are used to train a quantum error judgment model 420. For example, Figure 5B shows an exemplary quantum noise decoder 500 that uses a GAN architecture to generate a noise model of noise present in a computation performed by a particular quantum processor 115. Figure 5A provides a flowchart illustrating various processes, procedures, operations, etc., using the quantum noise decoder 500 to generate and / or provide a noise model for a particular quantum processor 115.

[0106] In the shown embodiment, the quantum noise decoder 500 includes two or more DNNs of a GAN architecture. For example, the quantum noise decoder 500 includes a generator 520 and a discriminator 540. The generator 520 includes a simulated noise model 525 of a particular quantum processor 115 and is configured to generate simulated operating data of the particular quantum processor 115 based at least in part on the simulated noise model 525. The simulated operating data 554 of the particular quantum processor 115 is provided to the discriminator 540.

[0107] The classifier 540 is configured to receive and / or acquire empirical operation data 552 (for example, from processing devices 805, 908 and / or programs, applications, modules, etc., running on the processing devices). The classifier 540 is further configured to receive and / or acquire simulated operation data 554 generated by the generator 520 based at least in part on the simulated noise model 525. The classifier 540 is configured to perform a blind analysis, processing, and / or comparison of the simulated operation data 554 and the empirical operation data 552 of a particular quantum processor to determine which dataset is the simulated operation data 554 and which dataset is the empirical operation data 552.

[0108] In various embodiments, the discriminator 540 includes a quantum error determination model 545. In various embodiments, the quantum error determination model 545 is trained and / or configured to characterize the noise of a particular quantum processor based on operation data corresponding to the operation of that particular quantum processor. For example, the quantum error determination model 545 is configured to analyze, process, and / or compare simulated operation data 554 of a particular quantum processor with empirical operation data 552 of a particular quantum processor. The quantum error determination model 545 may use the analysis, processing, and / or comparison of the simulated operation data 554 and the empirical operation data 552 of a particular quantum processor to characterize the noise of a particular quantum processor.

[0109] In various embodiments, the simulation noise model 525 of the generator 520 and the quantum error determination model 545 of the discriminator 540 are trained using GAN machine learning techniques. For example, the training module 560 receives from the discriminator 540 a determination and / or selection of which datasets consist of simulation data and which datasets consist of empirical data.

[0110] Based on whether the determination and / or selection from the discriminator 540 is correct, the training module 560 trains the generator 520 to produce simulated behavior data that is closer to empirical behavior data. For example, the training module 560 may adjust and modify the simulated noise model 525 to better approximate and / or reflect the noise present in the computation performed by a particular quantum processor 115.

[0111] Based on whether the decisions and / or selections from the classifier 540 are correct, the training module 560 trains the classifier 540 to be able to better distinguish between empirical operational data and simulated operational data. For example, the quantum error judgment model 545 may be trained, modified, or adjusted to better characterize noise present in the empirical operational data.

[0112] Once the simulated noise model 525 of the generator 520 and the quantum error decision model 545 of the discriminator 540 are trained to satisfy the convergence requirements, the noise model generation module 530 extracts a raw noise model 556 from the generator 520. In an exemplary embodiment, the raw noise model 556 is substantially similar to and / or a copy of the trained simulated noise model 525.

[0113] The noise model generation module 530 receives a raw noise model 556 and is configured to convert, transform, format, compile, and / or configure the raw noise model 556 into a noise model 558 that is understandable to the controller 30 and / or computing entity 10.

[0114] The quantum noise decoder 500 then provides an output including a noise model 558. The output may be received by one or more applications, programs, modules, etc., running on the controller 30 and / or computing entity 10.

[0115] For example, as shown in Figure 5A, starting from step / operation 502, the computing entity 10 and / or controller 30 cause the generator 520 to generate simulated operation data 554 based at least partially on the simulated noise model 525. The generator 520 then provides the simulated operation data 554 to the discriminator 540.

[0116] The discriminator 540 receives simulated operation data 554 and empirical operation data 552 (for example, provided to the quantum noise decoder 500 in step / operation 204).

[0117] In step / operation 504, the computing entity 10 and / or the controller 30 cause the classifier 540 to analyze, process, and / or compare simulated operation data 554 of a particular quantum processor with empirical operation data 552 of a particular quantum processor. For example, the classifier 540 receives the simulated operation data 554 and the empirical operation data 552 as a blind data set. For example, the classifier 540 receives two data sets, one containing the simulated operation data 554 and the other containing the empirical operation data 552. However, the classifier 540 receives the two data sets in such a way that it does not know which of the two data sets is the simulated operation data 554 and which of the two data sets is the empirical operation data 552.

[0118] The discriminator 540 uses a quantum error judgment model 545 to select one of the datasets as simulated behavioral data and the other as empirical behavioral data.

[0119] In step / operation 506, the computing entity 10 and / or controller 30 cause the training module 560 to adjust the training of the simulated noise model 525, the generator 520, the quantum error decision model 545, and / or the classifier 540 based on whether the classifier 540 correctly identified the simulated and / or empirical operational data. For example, the training module 560 may adjust and / or modify one or more weights and / or parameters of the simulated noise model 525, the generator 520, the quantum error decision model 545, and / or the classifier 540 using a loss function or the like.

[0120] In various embodiments, the training module 560 is configured to cause the generator 520 to generate simulation behavior data that better approximates empirical behavior data. For example, the training module 560 is configured to adjust and / or modify the simulation noise model 525 to better reflect and / or approximate the noise present in computations performed by a particular quantum processor 115. In various embodiments, the training module 560 is configured to cause the discriminator 540 to better distinguish between simulation behavior data and empirical behavior data. For example, the training module 560 is configured to cause the quantum error judgment model 545 to better characterize the noise present in computations performed by a particular quantum processor 115.

[0121] In step / operation 508, it is determined whether the training criteria are met. For example, the computing entity 10 and / or the controller 30 (possibly using the training module 560) determine whether the training criteria are met. For example, when the discriminator 540 correctly selects the simulation operation data and / or empirical operation data a threshold number of times consecutively, when the simulation noise model 525 and / or the quantum error decision model 545 converge, or when the loss function of the generator 520 satisfies the threshold criteria.

[0122] If, in step / action 508, it is determined that the training criteria are not met, the process returns to step / action 502, and another round of simulated action data is generated by the generator so that further training can be performed.

[0123] If it is determined in step / action 508 that the training criteria are met, the process proceeds to step / action 510.

[0124] In step / operation 510, the computing entity 10 and / or the controller 30 cause the noise model generation module 530 to extract the raw noise model 556 from the generator 520 and generate a noise model 558 based on the raw noise model 556. In an exemplary embodiment, the raw noise model is generated based on the output of the quantum error judgment model 545. In an exemplary embodiment, the noise model generation module 530 converts, transforms, formats, compiles, and / or configures the raw noise model into a noise model 558 that is understandable to the controller 30 and / or the computing entity 10.

[0125] In step / operation 512, the quantum noise decoder 500 provides an output containing a noise model 558 for a particular quantum processor 115. For example, an output containing a noise model 558 for a particular quantum processor 115 may be provided by the quantum noise decoder 500 via an API call or API response (for example, when the output is provided in response to an API call providing input empirical operation data 552) to an application, program, module, etc., being executed by processing devices 805, 908.

[0126] Exemplary noise models for specific quantum processors. Figure 6 provides a flowchart illustrating various processes, operations, and / or procedures performed by, for example, the controller 30 and / or computing entity 10 to provide a noise model so that at least one component and / or parameter of the quantum processor is modified, changed, tuned, etc., by a human engineer based on the noise model, according to various embodiments. In various embodiments, the processes, procedures, and / or operations in Figure 6 are performed as part of step / operation 208.

[0127] Beginning in step 602, a graphical representation of at least a portion of the noise model is generated. For example, the controller 30 (e.g., via processing device 805) and / or the computing entity 10 (e.g., via processing device 908) generate a graphical representation of at least a portion of the noise model. For example, memories 810, 922, and 924 may contain instructions that are executable by a computer configured to process the noise model and generate a graphical representation of the noise model when executed by processing devices 805 and 908. In various embodiments, the graphical representation of at least a portion of the noise model is configured to communicate and / or show to a human user information corresponding to the noise model, trends associated with noise identified in the behavioral data, etc. For example, the graphical representation of at least a portion of the noise model is configured to make at least a portion of the noise model readable and / or understandable to a human.

[0128] In various embodiments, the graphical representation may provide plots showing, as indicated by the noise model, the frequency profile of noise present in the electrical signals applied to the potential-generating elements of the confinement device 120; the wavelength / frequency fluctuations, phase shifts, and / or optical power fluctuations of various operating signals; the magnitude, direction, and / or frequency profiles of magnetic field fluctuations at one or more locations within the confinement device 120; and fluctuations in the quantum state indications of a collection of physical qubits used as logical qubits and / or a collection of spectator objects. The graphical representation may include plots showing and / or illustrating trends of various noise types and / or inducers over time, as indicated by the noise model.

[0129] In various embodiments, a graphical representation of a portion of a noise model may indicate that a given plot represents a portion of the quantum processor corresponding to that portion. For example, a set of plots showing the magnitude, direction, and / or frequency profiles of fluctuations in the magnetic field at one or more locations within the confinement device 120 when a two-qubit gate operation signal is applied to one or more locations may include indications that the set of plots corresponds to the application of a two-qubit gate operation signal, identify one or more locations, and so on.

[0130] In step / operation 604, the controller 30 and / or computing entity 10 cause a graphical representation of the noise model to be displayed via a GUI on a display (e.g., display 916). A human technician may examine and / or analyze the graphical representation displayed (e.g., via the GUI on display 916) and, at least in part thereon, modify, adjust, or change at least one component and / or parameter of the quantum processor 115 in a manner that is expected to reduce, and / or attempts to reduce, the noise of the computations performed by that particular quantum processor. For example, based on the noise model's contribution to the noise of that particular quantum processor, the human technician may replace a burnt-out optical fiber, adjust the alignment of a mirror or lens, and so on.

[0131] Figure 7 provides a flowchart illustrating various processes, operations, and / or procedures performed by, for example, the controller 30 and / or computing entity 10 to provide a noise model such that at least one component and / or parameter of the quantum processor is automatically modified, changed, adjusted, etc., based on the noise model, according to various embodiments. In various embodiments, the processes, procedures, and / or operations in Figure 7 are performed as part of step / operation 208.

[0132] Starting from step / operation 702, the noise model is processed to determine whether any components and / or parameters may be automatically modified, adjusted, or changed in an attempt to reduce the noise afflicted by a particular quantum processor 115, characterized by the noise model (e.g., reducing the amplitude of the noise in one or more noise profiles provided by the noise model, reducing the presence of certain features present in one or more noise profiles provided by the noise model, etc.), and / or to identify whether any components and / or parameters may be automatically modified, adjusted, or changed. In an exemplary embodiment, the controller 30 and / or computing entity 10 process the noise model to determine whether there are any components and / or parameters that may be automatically modified, adjusted, or changed in an attempt to reduce the noise afflicted by a particular quantum processor 115 (e.g., reducing the amplitude of the noise in one or more noise profiles provided by the noise model, reducing the presence of certain features present in one or more noise profiles provided by the noise model, etc.).

[0133] For example, a quantum error judgment model and / or noise model generation module trained by machine learning is configured, in an exemplary embodiment, to identify likely noise sources with respect to a particular subsystem of the quantum processor 115. In such an embodiment, the noise model identifies the identified likely noise sources. For example, the noise model may include indications that an optical fiber or waveguide along a particular optical path may be burnt out, or that the alignment of an optical element along a particular optical path may need to be addressed. The noise model may then be processed with knowledge of what corrections, adjustments, modifications, etc., may be performed automatically and which require human technician intervention. For example, a human technician may be required to replace a burnt-out optical fiber. However, an automated alignment process may be defined and / or programmed so that the controller 30 can perform automated alignment of a particular optical path (or at least a portion thereof). In another example, one or more software components and / or modifications may be performed automatically (e.g., updating parameters of a calibration process, providing a noise profile to a real-time quantum error decoder).

[0134] When modifications, adjustments, or changes to at least one component and / or parameter are likely and / or expected to reduce noise present in a computation performed by a particular quantum processor 115 (for example, reducing the amplitude of noise in one or more noise profiles provided by a noise model, or reducing the presence of certain features present in one or more noise profiles provided by a noise model), the controller 30 and / or computing entity 10 may trigger the execution of one or more such automated modifications, adjustments, or changes. For example, the controller 30 of a quantum computer may modify, adjust, or change at least one component and / or parameter of the quantum processor based on a noise model (for example, adjusting hardware components and / or parameters, software components and / or parameters, and / or calibration components and / or parameters).

[0135] For example, in step / operation 704, the controller 30 and / or computing entity 10 may modify, adjust, change, etc., at least one component and / or parameter of the real-time quantum error decoder based on a noise model. For example, the real-time quantum error decoder may be used to perform real-time quantum error correction during the operation of a particular quantum processor 115. At least one component and / or parameter of the real-time quantum error decoder may be modified, adjusted, changed, etc., based on a noise model, so that the real-time quantum error decoder can more accurately determine, consider, and / or correct quantum errors during the operation of a particular quantum processor 115. For example, the real-time quantum error decoder may be used to determine the phase shift of a qubit that needs to be considered during the execution of a quantum circuit. Thus, in an exemplary embodiment, at least one component and / or parameter of the real-time quantum error decoder may be modified, adjusted, changed, etc., based on a noise model, for example, so that a more accurate phase shift of a qubit is determined.

[0136] In another example, in step / operation 706, the controller 30 and / or computing entity 10 may modify, adjust, or change at least one component and / or parameter of the calibration process based on the noise model. For example, a particular calibration process may be performed more frequently, a new calibration process may be developed and used, or the parameters used in the calibration process may be updated.

[0137] In another example, in step / operation 708, the controller 30 and / or computing entity 10 may modify, adjust, or change at least one component and / or parameter of the driver controller element 815 based at least partially on the noise model. For example, if the noise model indicates that the noise of a voltage signal provided by a particular voltage source 50 is particularly high, the corresponding driver controller element 815 may be modified, adjusted, or changed to cause filtering of the voltage signal provided by the particular voltage source in a way that reduces the noise observed in the voltage signal. In another example, the technique for performing a function of the quantum computer (e.g., performing a 1 or 2 qubit gate, performing a transfer operation, performing a read operation, etc.) may be modified, updated, or changed based on the noise model and / or the results of processing and / or analyzing the noise model to reduce the noise present in the computation performed by a particular quantum processor 115. For example, the components and / or parameters of the driver controller element 815 may be modified, adjusted, or changed so that a particular operation source 60 may be driven in a slightly different way during the execution of a function of the quantum computer.

[0138] Technical advantages Large-scale quantum computers are expected to solve problems in fields such as chemistry, materials science, and biology that are currently beyond the capabilities of modern technology. Solving such problems involves computations employing quantum algorithms implemented using deep quantum circuits. Achieving the required level of precision in these deep circuits requires a high level of reliability in quantum operations. To achieve such reliability, quantum error correction (QEC) is employed during computation to suppress noise to the required level. However, understanding the noise present in a particular quantum processor is helpful in suppressing the noise present within that particular quantum processor to the required level.

[0139] When used herein, a particular quantum processor corresponds to a particular instance of hardware and its configuration for providing a particular quantum processor. For example, in the field of quantum charge-coupled device (QCCD) based quantum computing, a particular quantum processor corresponds to a particular ion trap, a magnetic field generating component, an operating source (e.g., a laser), and an optical path defined to provide operating signals (e.g., a laser beam) to each of the particular ion traps. For example, if a particular mirror or lens in the optical path is slightly misaligned, or if the optical fiber defining part of the optical path is nearly burnt out, the operating signals provided along the optical path may carry less optical power than expected and / or result in an uncompensated shift in the optical phase of the operating signals. Thus, the function performed using the optical path contributes to the noise of the particular quantum processor. However, a second quantum processor of a similar design does not suffer its particular contribution to the noise of the second quantum processor.

[0140] Current techniques for managing noise in quantum processors include the use of real-time quantum error decoders. However, due to the time constraints of performing real-time quantum error correction while running quantum circuits, current real-time quantum error decoders tend to be simple programs that rely on algorithms such as the Blossom algorithm or Dijkstra's algorithm. These real-time quantum error decoders generally cannot provide a broader characterization of the noise of a particular quantum processor and may rely on general noise models that cannot characterize the noise contributions of different and / or specific quantum processors to which the real-time quantum error decoder is associated. Thus, there are technical problems in the field of characterizing the noise of quantum processors and performing quantum error correction (including real-time quantum error correction) of quantum processors using noise models that characterize the noise of quantum processors.

[0141] Various embodiments provide technical solutions to these technical problems. In particular, various embodiments provide a quantum noise decoder that includes a machine learning-based quantum error judgment model, which is trained using machine learning techniques to characterize the noise of a particular quantum processor. The quantum noise decoder (e.g., the machine learning-based quantum error judgment model) is trained using operation data (e.g., empirical operation data) corresponding to and / or generated during the operation of a particular quantum processor. The quantum noise decoder is then used to generate a noise model that characterizes the noise of a particular quantum processor.

[0142] Based on noise related to a particular quantum processor, at least one component and / or parameter of the quantum processor may be modified, tuned, or altered to reduce noise in computations performed by that particular quantum processor. For example, a graphical representation of at least a portion of the noise model for a particular quantum processor may be displayed by a graphical user interface (GUI) provided via a display of the computing entity, so that a human technician may modify, tune, or alter at least one component and / or parameter. For example, a human technician may replace a burnt-out optical fiber, adjust the alignment of a mirror or lens, etc., based on the noise contribution of a particular quantum processor identified and / or indicated by the noise model. For example, a controller of a quantum computer may modify, tune, or alter at least one component and / or parameter of the quantum processor based on the noise model (for example, it may tune hardware components and / or parameters, software components and / or parameters, and / or calibration components and / or parameters). In an exemplary embodiment, the noise model is provided to a real-time quantum error decoder for use when performing real-time quantum error correction for a particular quantum processor.

[0143] Accordingly, various embodiments provide methods, apparatus, systems, computer program products, etc., for determining a noise model that characterizes the noise of a particular quantum processor. Various embodiments also provide methods, apparatus, systems, computer program products, etc., for reducing the noise of a particular quantum processor based on the determined noise model that characterizes the noise of a particular quantum processor. Accordingly, various embodiments provide practical applications that offer technical solutions and technical advantages to quantum computing, including areas such as quantum error correction, real-time quantum error correction, and quantum processor noise reduction.

[0144] Example Controller In various embodiments, the controller 30 of the quantum computer 110 is configured to control the operation of various components, elements, assemblies, etc., of the quantum processor 115. For example, in various embodiments, the controller 30 is configured to control a voltage source 50, a cryostat system and / or vacuum system that controls the temperature and pressure within the cryostat and / or vacuum chamber 40, an operating source 60, and / or other systems that control various environmental conditions (e.g., temperature, pressure, magnetic field, etc.) within the cryostat and / or vacuum chamber 40, and / or is configured to operate and / or induce a controlled evolution of the quantum state of one or more atomic objects within the confinement device. In various embodiments, the controller 30 is configured to trigger the execution of one or more calibration processes to generate calibration data corresponding to the operation of the quantum processor 115, and / or to modify, adjust, change, etc., one or more components and / or parameters of the quantum processor 115, at least in part based on a noise model for a particular quantum processor 115.

[0145] As shown in Figure 8, in various embodiments, the controller 30 includes various controller elements, including a processing device 805, memory 810, driver controller element 815, communication interface 820, analog-to-digital converter element 825, and so on. For example, the processing device 805 may include one or more processing elements such as a programmable logic device (CPLD), microprocessor, coprocessing entity, application-specific instruction-set processor (ASIP), integrated circuit, application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), programmable logic array (PLA), hardware accelerator, and other processing devices and / or circuits and / or controllers. The term "circuit" may refer to an entirely hardware embodiment or a combination of hardware and computer program products. In an exemplary embodiment, the processing device 805 of the controller 30 includes and / or communicates with a clock.

[0146] For example, memory 810 may include non-temporary memory such as volatile and / or non-volatile memory storage, such as one or more of the following: hard disk, ROM, PROM, EPROM, EEPROM, flash memory, MMC, SD memory card, memory stick, CBRAM, PRAM, FeRAM, RRAM, SONOS, racetrack memory, RAM, DRAM, SRAM, FPM DRAM, EDO DRAM, SDRAM, DDR SDRAM, DDR2 SDRAM, DDR3 SDRAM, RDRAM, RIMM, DIMM, SIMM, VRAM, cache memory, register memory, etc. In various embodiments, memory 810 may store qubit records corresponding to the qubits of a quantum computer (e.g., in a qubit record datastore, qubit record database, qubit record table, etc.), calibration tables, executable queues, computer program code (e.g., one or more computer languages, special controller languages, etc.). In an exemplary embodiment, the execution of at least a portion of computer program code stored in memory 810 (for example, by processing device 805) causes controller 30 to perform one or more steps, operations, processes, procedures, etc. described herein to track the phases of atomic objects in an atomic system and cause phase adjustments of one or more manipulated sources and / or signals generated by the manipulated sources.

[0147] In various embodiments, the driver controller element 815 may include one or more drivers and / or controller elements, each configured to control one or more drivers. In various embodiments, the driver controller element 815 may include drivers and / or driver controllers. For example, a driver controller may be configured to operate one or more corresponding drivers according to executable instructions, commands, etc., scheduled and executed by the controller 30 (e.g., by the processing device 805). In various embodiments, the driver controller element 815 may enable the controller 30 to operate the operating source 60. In various embodiments, the drivers may be laser drivers, vacuum component drivers, drivers for controlling the flow of current and / or voltage of electrical signals applied to potential generating elements (e.g., electrodes) of the confinement device 120 (e.g., voltage source 50), cryogenic and / or vacuum system component drivers, etc.

[0148] In various embodiments, the controller 30 includes means for transmitting and / or receiving signals from one or more photodetector components, such as cameras, MEMS cameras, CCD cameras, photodiodes, and photomultiplier tubes. For example, the controller 30 may include one or more analog-to-digital converter elements 825 configured to receive signals from one or more photodetector components, calibration sensors, and the like.

[0149] In various embodiments, the controller 30 includes a communication interface 820 for interfaceing with and / or communicating with one or more computing entities 10. For example, the controller 30 may include a communication interface 820 for receiving executable instructions, command sets, noise models, etc., from the computing entities 10 and for providing the computing entities 10 with outputs received from the quantum computer 110 (e.g., from the optical acquisition system 70) and / or the results of processing those outputs. In various embodiments, the computing entities 10 and the controller 30 may communicate directly via wired and / or wireless connections and / or via one or more wired and / or wireless networks 20.

[0150] Exemplary Computing Entity Figure 9 provides a descriptive schematic diagram of an exemplary computing entity 10 that may be used in conjunction with embodiments of the present disclosure. In various embodiments, the computing entity 10 is a classical (e.g., semiconductor-based) computer configured to allow a user to provide input to a quantum computer 110 (e.g., through a user interface of the computing entity 10) and to receive, display, analyze, and so on outputs from the quantum computer 110.

[0151] As shown in Figure 9, the computing entity 10 may include an antenna 912, a (e.g., wireless) transmitter 904, a (e.g., wireless) receiver 906, and a processing device 908 that provides a signal to the transmitter 904 and receives a signal from the receiver 906, respectively. In various embodiments, the processing device 908 may include one or more processing elements such as a programmable logic device (CPLD), a microprocessor, a coprocessing entity, an application-specific instruction set processor (ASIP), an integrated circuit, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a programmable logic array (PLA), a hardware accelerator, and / or other processing devices and / or circuits and / or controllers. The term "circuit" may refer to an entirely hardware embodiment or a combination of hardware and computer program products.

[0152] The signals provided from the processing device 908 to the transmitter 906 and the signals received by the processing device 908 from the receiver 906 may include signaling information / data in accordance with applicable wireless system radio interface standards for communication with various entities such as the controller 30 and other computing entities 10.

[0153] In this regard, the computing entity 10 may operate using one or more wireless interface standards, communication protocols, modulation types, and access types. For example, the computing entity 10 may be configured to receive and / or provide communications using wired data transmission protocols such as Fiber Optic Distributed Data Interface (FDDI), Digital Subscriber Line (DSL), Ethernet, Asynchronous Transfer Mode (ATM), Frame Relay, Data Over Cable Service Interface Specification (DOCSIS), or any other wired transmission protocol. Similarly, Computing Entity 10 supports General-Purpose Packet Radio Services (GPRS), Universal Mobile Telecommunications System (UMTS), Code Division Multiple Access 2000 (CDMA2000), CDMA2000 1X (1xRTT), Wideband Code Division Multiple Access (WCDMA®), Global System for Mobile Communications (GSM), Enhanced Data Rates for GSM Evolution (EDGE), Time Division Synchronous Code Division Multiple Access (TD-SCDMA), Long-Term Evolution (LTE), Evolutionary Universal Terrestrial Radio Access Network (E-UTRAN), Evolution-Data Optimized (EVDO), High-Speed ​​Packet Access (HSPA), High-Speed ​​Downlink Packet Access (HSDPA), IEEE 802.11 (Wi-Fi), and Wi-Fi. It may be configured to communicate over a wireless external communication network using any of the following protocols: Direct, 802.16 (WiMAX), Ultra Wideband (UWB), Infrared (IR) protocol, Near Field Communication (NFC) protocol, Wibree, Bluetooth protocol, Wireless Universal Serial Bus (USB) protocol, and / or any other wireless protocol.Computing entity 10 may communicate using such protocols and standards, including Border Gateway Protocol (BGP), Dynamic Host Configuration Protocol (DHCP), Domain Name System (DNS), File Transfer Protocol (FTP), Hypertext Transfer Protocol (HTTP), HTTP, HTTP over TLS / SSL / Secure, Internet Message Access Protocol (IMAP), Network Time Protocol (NTP), Simple Mail Transfer Protocol (SMTP), Telnet, Transport Layer Security (TLS), Secure Sockets Layer (SSL), Internet Protocol (IP), Transmission Control Protocol (TCP), User Datagram Protocol (UDP), Datagram Congestion Control Protocol (DCCP), Stream Control Transmission Protocol (SCTP), and Hypertext Markup Language (HTML).

[0154] Through these communication standards and protocols, computing entity 10 can communicate with various other entities using concepts such as Unstructured Supplementary Service information / data (USSD), Short Message Service (SMS), Multimedia Messaging Service (MMS), Dual-Tone Multi-Frequency Signaling (DTMF), and / or Subscriber Identification Module Dialer (SIM dialer). Computing entity 10 can also download changes, add-ons, and updates to its firmware, software (including executable instructions, applications, and program modules), and operating system, for example.

[0155] The computing entity 10 may also include user interface devices, including one or more user input / output interfaces (for example, a display 916 and / or speaker / speaker driver coupled to the processing device 908, as well as a touchscreen, keyboard, mouse, and / or microphone coupled to the processing device 908). For example, a user output interface may be configured to provide applications, browsers, user interfaces, interfaces, dashboards, screens, web pages, pages, and / or similar words used herein to be interchangeable, to trigger the display or audible presentation of information / data, and for interaction with that information / data via one or more user input interfaces. A user input interface may include any of several devices that enable the computing entity 10 to receive data, such as a keypad 918 (hard or soft), a touch display, a voice / speech or motion interface, a scanner, a reader, or other input device. In embodiments including a keypad 918, the keypad 918 may include (or trigger the display of) conventional numerals (0-9) and associated keys (#, *), as well as other keys used to operate the computing entity 10, and may include a set of keys that can be operated to provide a complete set of alphabet keys or a complete set of alphanumeric keys. In addition to providing input, the user input interface may be used to activate or deactivate certain functions, such as a screen saver and / or sleep mode. Through such input, the computing entity 10 can collect information / data, user interactions / input, etc.

[0156] The computing entity 10 may also include volatile storage or memory 922 and / or non-volatile storage or memory 924, which may be embedded and / or removable. For example, non-volatile memory may be ROM, PROM, EPROM, EEPROM, flash memory, MMC, SD memory card, memory stick, CBRAM, PRAM, FeRAM, RRAM, SONOS, racetrack memory, etc. Volatile memory may be RAM, DRAM, SRAM, FPM DRAM, EDO DRAM, SDRAM, DDR SDRAM, DDR2 SDRAM, DDR3 SDRAM, RDRAM, RIMM, DIMM, SIMM, VRAM, cache memory, register memory, etc. The volatile and non-volatile storage or memory may store databases, database instances, database management system entities, data, applications, programs, program modules, scripts, source code, object code, bytecode, compiled code, interpreted code, machine code, executable instructions, etc., for implementing the functions of the computing entity 10.

[0157] conclusion Many modifications and other embodiments of the invention described herein will be conceivable to those skilled in the art in which the invention relates, benefiting from the teachings shown in the above description and the accompanying drawings. It should be understood that the invention should not be limited to the specific embodiments disclosed, and that modifications and other embodiments are intended to be included within the scope of the appended claims. While specific terminology is used herein, these terms are used only in a general and descriptive sense and not for limiting purposes. [Explanation of Symbols]

[0158] 10 Computing Entities 30 controllers 40 Cryostat and / or vacuum chamber 50 Voltage source 60 Operation source 70 Optical Acquisition System 100 Quantum Computing Systems 110 Quantum Computers 115 Quantum Processors 120 Atomic Object Confinement Device 400 Quantum Noise Decoder 420 Quantum Error Judgment Models 430 Noise Model Generation Module 442 inputs 444 Raw Noise Model 446 Noise Model 500 Quantum Noise Decoder 520 Generator 525 Simulation Noise Model 530 Noise Model Generation Module 540 Classifiers 545 Quantum Error Judgment Model 552 Empirical Operational Data 554 Simulated operation data 556 Raw Noise Model 558 Noise Model 560 training modules 805 Processing Device 810 memory 815 Driver Controller Element 820 Communication Interface 825 A / D Converter 904 Transmitter 906 Receiver 908 Processing Devices 912 Antenna 916 displays 918 Keypad 920 Network Interfaces 922 memory 924 memory

Claims

1. The steps include training a quantum noise decoder, which includes a machine learning-trained quantum error judgment model, using training data that includes operational data captured by one or more processors based at least partially on the operation of a particular quantum processor, and The steps include generating a noise model for a particular quantum processor based on a quantum error judgment model trained by machine learning using one or more of the aforementioned processors, A step of providing the noise model by one or more processors, comprising at least one of the following: (a) providing a graphical representation of at least a portion of the noise model via a display of a computing entity so that at least one component or parameter of the particular quantum processor is modified or changed based on a graphical representation of at least a portion of the noise model; or (b) providing the at least portion of the noise model as input related to executable instructions to be executed by a controller of the particular quantum processor or a computing entity communicating with the controller of the particular quantum processor, so that at least one component or parameter of the particular quantum processor is modified or changed based on at least a portion of the noise model. Methods that include...

2. The operation data includes calibration data generated through the operation of the particular quantum processor, and The operation data includes spectator object data acquired by direct or indirect observation of one or more spectator objects controlled by the specific quantum processor, wherein the one or more spectator objects are controlled independently of the quantum algorithm being executed by the specific quantum processor. The method according to claim 1, wherein at least one of the following:

3. The quantum noise decoder includes a generative adversarial network (GAN) comprising a generator and a discriminator, wherein the generator is configured to generate simulated behavioral data, and The quantum noise decoder includes a noise model generation module configured to generate the noise model for the particular quantum processor based at least partially on the output of the machine learning-trained quantum error judgment model. The method according to claim 1, wherein at least one of the following:

4. The method according to claim 1, wherein the noise model characterizes the noise present in the operation data of the particular quantum processor.

5. The at least one component or parameter is part of a real-time quantum error decoder for correcting quantum errors during the operation of the particular quantum processor, or is used by the real-time quantum error decoder. The at least one component or parameter is a hardware component or physical parameter of the particular quantum processor, and The at least one component or parameter corresponds to the recalibration of the hardware components of the particular quantum processor or the software process of the controller of the particular quantum processor. The method according to claim 1, wherein at least one of the following:

6. The method according to claim 1, wherein the quantum noise decoder is a real-time quantum error decoder configured to cause quantum error correction during the operation of the particular quantum processor.

Images