Conceptio › Archive › Google Patents

Google Patentsopen access

Quantum computing machine learning module — Accenture Global Solutions Limited (US11803772B2)

Accenture Global Solutions Limited · Google Patents

Google Patents · Patents · License: Open Access

patent, google patents, intellectual property, US11803772B2, Accenture Global Solutions Limited, Carl Matthew Dukatz, en, 2023

ABSTRACT

Abstract

Methods, systems, and apparatus for training a machine learning model to route received computational tasks in a system including at least one quantum computing resource. In one aspect, a method includes obtaining a first set of data, the first set of data comprising data representing multiple computational tasks previously performed by the system; obtaining input data for the multiple computational tasks previously performed by the system, comprising data representing a type of computing resource the task was routed to; obtaining a second set of data, the second set of data comprising data representing properties associated with using the one or more quantum computing resources to solve the multiple computational tasks; and training the machine learning model to route received data representing a computational task to be performed using the (i) first set of data, (ii) input data, and (iii) second set of data.

Description

CROSS-REFERENCE TO RELATED APPLICATION

This application is a continuation of U.S. application Ser. No. 15/491,852, filed Apr. 19, 2017, now allowed, which is incorporated by reference in its entirety.

BACKGROUND

For some computational tasks, quantum computing devices may offer a computational speed up compared to classical devices. For example, quantum computers may achieve a speed up for tasks such as database search, evaluating NAND trees, integer factorization or the simulation of quantum many-body systems.

As another example, adiabatic quantum annealers may achieve a computational speed up compared to classical annealers for some optimization tasks. To perform an optimization task, quantum hardware may be constructed and programmed to encode the solution to a corresponding optimization task into an energy spectrum of a many-body quantum Hamiltonian characterizing the quantum hardware. For example, the solution is encoded in the ground state of the Hamiltonian.

SUMMARY

This specification describes a machine learning module that may be used to route received computational tasks to one or more quantum computing devices or one or more classical computing devices. The machine learning module uses machine learning techniques to determine when and how to leverage the power of quantum computing.

In general, one innovative aspect of the subject matter described in this specification can be implemented in a computer implement method for training a machine learning model to route received computational tasks in a system including at least one quantum computing resource, the method including the actions of: obtaining a first set of data, the first set of data comprising data representing multiple computational tasks previously performed by the system; obtaining a second set of data, the second set of data comprising data representing properties associated with using the one or more quantum computing resources to solve the multiple computational tasks; obtaining input data for the multiple computational tasks previously performed by the system, comprising data representing a type of computing resource the task was routed to; and training the machine learning model to route received data representing a computational task to be performed using the (i) first set of data, (ii) input data, and (iii) second set of data.

Other implementations of this aspect include corresponding computer systems, apparatus, and computer programs recorded on one or more computer storage devices, each configured to perform the actions of the methods. A system of one or more computers can be configured to perform particular operations or actions by virtue of having software, firmware, hardware, or a combination thereof installed on the system that in operation causes or cause the system to perform the actions. One or more computer programs can be configured to perform particular operations or actions by virtue of including instructions that, when executed by data processing apparatus, cause the apparatus to perform the actions.

The foregoing and other implementations can each optionally include one or more of the following features, alone or in combination. In some implementations the quantum computing resources comprise one or more of (i) quantum gate computers, (ii) adiabatic annealers, or (iii) quantum simulators.

In some implementations the system further comprises one or more classical computing resources.

In some implementations the computational tasks comprise optimization tasks.

In some implementations properties of using the one or more quantum computing resources to solve the multiple computational tasks comprise, for each computational task, one or more of: (i) approximate qualities of solutions generated by the one or more quantum computing resources; (ii) computational times associated with solutions generated by the one or more quantum computing resources; or (iii) computational costs associated with solutions generated by the one or more quantum computing resources.

In some implementations data representing properties of using the one or more quantum computing resources to solve multiple computational tasks further comprises, for each quantum computing resource, one or more of (i) a number of qubits available to the quantum computing resource; and (ii) a cost associated with using the quantum computing resource.

In some implementations the obtained input data for the multiple computational tasks previously performed by the system further comprises, for each computational task, one or more of: (i) data representing a size of an input data set associated with the computational task; (ii) data indicating whether an input data set associated with the computational task comprised static, real time or both static and real time input data; (iii) data representing an error tolerance associated with the computational task; and (iv) data representing a required level of confidence associated with the computational task.

In some implementations the obtained input data for the multiple computational tasks previously performed by the system further comprises data indicating a frequency of changes to the input data sets associated with each computational tasks.

In some implementations training the machine learning model to route received computational tasks comprises: generating a set of training examples using the (i) first set of data, (ii) input data, and (iii) second set of data, wherein each training example comprises a machine learning model input paired with a known machine learning model output; and training the machine learning model using the set of training examples.

In general, another innovative aspect of the subject matter described in this specification can be implemented in a computer implement method for obtaining a solution to a computational task, the method comprising: receiving data representing a computational task to be performed by a system including one or more quantum computing resources and one or more classical computing resources; processing the received data using a machine learning model to determine which of the one or more quantum computing resources or the one or more classical computing resources to route the data representing the computational task to, wherein the machine learning model has been configured through training to route received data representing computational tasks to be performed in a system including at least one quantum computing resource; and routing the data representing the computational task to the determined computing resource to obtain, from the determined computing resource, data representing a solution to the computational task.

In some implementations the computational tasks comprise optimization tasks.

In some implementations training the machine learning model to route received data representing computational tasks to be performed comprises training the machine learning model using (i) data representing multiple computational tasks previously performed by the system, and (ii) data representing a type of computing resource the task was routed to.

In some implementations the data representing multiple computational tasks previously performed by the system comprises, for each computational task, one or more of: (i) data representing a size of an input data set associated with the computational task; (ii) data indicating whether an input data set associated with the computational task comprised static, real time or both static and real time input data; (iii) data representing an error tolerance associated with the computational task; and (iv) data representing a required level of confidence associated with the computational task.

In some implementations the data representing multiple computational tasks previously performed by the system comprises data indicating a frequency of changes to input data sets associated with each computational tasks.

In some implementations the method further comprises training the machine learning model using data representing properties associated with using the one or more quantum computing resources to solve the multiple computational tasks.

In some implementations properties associated with using the one or more quantum computing resources to solve the multiple computational tasks comprise: for each computational task, one or more of: (i) approximate qualities of solutions generated by the one or more quantum computing resources; (ii) computational times associated with solutions generated by the one or more quantum computing resources; or (iii) computational costs associated with solutions generated by the one or more quantum computing resources, and for each quantum computing resource, one or more of (i) a number of qubits available to the quantum computing resource; and (ii) a cost associated with using the quantum computing resource.

The subject matter described in this specification can be implemented in particular ways so as to realize one or more of the following advantages.

For some optimization tasks, quantum computing devices may offer an improvement in computational speed compared to classical devices. For example, quantum computers may achieve an improvement in speed for tasks such as database search or evaluating NAND trees. As another example, quantum annealers may achieve an improvement in computational speed compared to classical annealers for some optimization tasks. For example, determining a global minimum or maximum of a complex manifold associated with the optimization task is an extremely challenging task. In some cases, using a quantum annealer to solve an optimization task can be an accurate and efficient alternative to using classical computing devices.

Conversely, for some optimization tasks, quantum computing devices may not offer an improvement compared to classical devices. For example, whilst quantum computing devices may offer computational speedups for some computational tasks, the costs associated with using the quantum devices to perform the computational tasks may be higher than the costs associated with using classical computing devices to perform the computational tasks. Such costs can include computational costs, i.e., the cost of resources required to build and use a computing device, and financial costs, i.e., monetary costs and fees of renting computational time on an external computing resource. Therefore, a tradeoff between the benefits of using quantum computing resources and classical computing resources exists.

A quantum computing machine learning module, as described in this specification, balances this tradeoff and learns optimal routings of computational tasks to classical or quantum computing resources. By learning when and how to utilize the power of quantum computing, a system implementing the quantum computing machine learning module may perform computational tasks more efficiently and/or accurately compared to systems that do not include quantum computing resources, or to systems that do not learn optimal routings of computational tasks to classical or quantum resources.

In addition, a quantum computing module, as described in this specification, can adapt overtime as more efficient quantum and classical systems are introduced. For example, while current implementations of quantum computations or current quantum hardware may include a significant classical overhead, e.g., leverages classical computing capabilities, well founded evidence supports that future quantum computing hardware will be able to perform exponentially more challenging tasks in less time that current quantum computing hardware or classical computing hardware. Conversely, as classical computers continue to be horizontally scalable, the additional capacity made available for computation will be factored into the machine learning to determine the best use of resources.

The details of one or more implementations of the subject matter of this specification are set forth in the accompanying drawings and the description below. Other features, aspects, and advantages of the subject matter will become apparent from the description, the drawings, and the claims.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 A depicts an example system for performing computational tasks.

FIG. 1 B depicts an example visualization of a global search space and local search space.

FIG. 2 depicts an example quantum computing machine learning module.

FIG. 3 is a flow diagram of an example process for training a machine learning model to route received computational tasks in a system including one or more quantum computing resources.

FIG. 4 is a flow diagram of an example process for obtaining a solution to a computational task using a system including one or more quantum computing resources.

Like reference numbers and designations in the various drawings indicate like elements.

DETAILED DESCRIPTION

For some computational tasks, quantum computing devices may offer a computational speed up compared to classical devices. For example, quantum computers may achieve a polynomial speed up for tasks such as database search or evaluating NAND trees. As another example, adiabatic quantum annealers may achieve a computational speed up compared to classical annealers for some optimization tasks.

However, such quantum computers or quantum annealers may not be universally applied to any computational taskâfor example, in the case of quantum annealers, a computational task is designed around the quantum annealer, rather than the quantum annealer being designed around the computational task. Furthermore, in some cases a computational task may be too complex to implement using quantum devices, or may require a large classical overhead, meaning that using the quantum device to solve the computational task is either slower or more costly than using a classical device to solve the computational task.

This specification provides systems and methods for determining when and how to leverage quantum computing devices when solving computational tasks. A system can receive computational tasks, e.g., optimization tasks, to be performed. For example, the system may be an optimization engine that is configured to receive input data and to generate, as output, an optimal solution to an optimization task based on the received input data. The received input data can include static and real-time data.

The system outsources computations associated with the received computational tasks to one or more exter

CROSS-REFERENCE TO RELATED APPLICATION

This application is a continuation of U.S. application Ser. No. 15/491,852, filed Apr. 19, 2017, now allowed, which is incorporated by reference in its entirety.

BACKGROUND

SUMMARY

In some implementations the system further comprises one or more classical computing resources.

In some implementations the computational tasks comprise optimization tasks.

The subject matter described in this specification can be implemented in particular ways so as to realize one or more of the following advantages.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 A depicts an example system for performing computational tasks.

FIG. 1 B depicts an example visualization of a global search space and local search space.

FIG. 2 depicts an example quantum computing machine learning module.

FIG. 3 is a flow diagram of an example process for training a machine learning model to route received computational tasks in a system including one or more quantum computing resources.

FIG. 4 is a flow diagram of an example process for obtaining a solution to a computational task using a system including one or more quantum computing resources.

Like reference numbers and designations in the various drawings indicate like elements.

DETAILED DESCRIPTION

The system outsources computations associated with the received computational tasks to one or more external devices. In some cases the system may preprocess the computational tasks before outsourcing the tasks, e.g., including separating a received computational task into one or more sub tasks. The external devices can include quantum computing devices, e.g., quantum annealers, quantum simulators or quantum gate computers, and classical computing devices, e.g., standard classical processors or supercomputers.

The system decides when and where to outsource computations associated with the received computational tasks. Such task routing may be a complex problem that is dependent on many factors. The system is trained to learn optimal routings of received computational tasks using a set of training data. The training data includes data from several sources, as described below, which may be used to generate multiple training examples. Each training example can include (i) input data relating to a previous computational task, e.g., data specifying the task, size/complexity of the task, restrictions for solving the task, error tolerance, (ii) information relating to which device was used to solve the task, or (iii) metrics indicating a quality of the solution obtained using the device, e.g., a level of confidence in the solution, computational time taken to generate the solution, or computational costs incurred.

Other data may be included in the training data, including an indication of the computational resources available when processing the previous computational tasks, a number of qubits used in the quantum devices, ability of algorithms running on the quantum devices or classical devices to process the problem, costs of using the classical or quantum devices, or reliability of classical or quantum devices.

In some cases the system learns to adapt the routings of computational tasks based on traffic flow in the system, e.g., if many computational tasks are received, the machine learning module may learn to prioritize certain tasks or subtasks, or learn when it is more efficient to wait for a particular device to become available again or when it is more efficient to route a task to a second-choice device.

Training can include applying conventional machine learning techniques, such as computing a loss function and backpropagating gradients. Once the machine learning module has been trained, it may be used at runtime for inference, e.g., to receive new computational tasks to be performed and to find an improved routing of the computational tasks in order to obtain solutions to the received tasks.

Example Operating Environment

FIG. 1 A depicts an example system 100 for performing computational tasks. The system 100 is an example of a system implemented as computer programs on one or more computers in one or more locations, in which the systems, components, and techniques described below can be implemented.

The system 100 for performing computational tasks is configured to receive as input data representing a computational task to be solved, e.g., input data 102 . For example, in some cases the system 100 may be configured to solve multiple computational tasks, e.g., including optimization tasks, simulation tasks, arithmetic tasks, database search, machine learning tasks, or data compression tasks. In these cases, the input data 102 may be data that specifies one of the multiple computational tasks. The input data 102 representing the computational task to be solved may specify one or more properties of the computational task. For example, in cases where the computational task is an optimization task, the input data 102 may include data representing parameters associated with the optimization task, e.g., parameters over which an objective function representing the optimization task is to be optimized, and one or more values of the parameters.

In some cases the input data 102 may include static input data and dynamic input data, e.g., real-time input data. As an example, the input data 102 may be data that represents the task of optimizing the design of a water network in order to optimize the amount of water distributed by the network. In this example, the input data 102 may include static input data representing one or more properties of the water network, e.g., a total number of available water pipes, a total number of available connecting nodes or a total number of available water tanks. In addition, the input data 102 may include data representing one or more parameters associated with the optimization task, e.g., level of water pressure in each pipe, level of water pressure at each connecting node, height of water level in each water tank, concentration of chemicals in the water throughout the network, water age or water source. Furthermore, the input data 102 may include dynamic input data representing one or more current properties or values of parameters of the water network, e.g., a current number of water pipes in use, a current level of water pressure in each pipe, a current concentration of chemicals in the water, or a current temperature of the water.

In some implementations, the input data 102 may further include data specifying one or more task objectives associated with the computational task. The task objectives may include local task objectives and global task objectives. Local task objectives may include local targets to be considered when solving the computational task, e.g., local objectives of a solution to the computational task. For example, local objectives may include constraints on values of subsets of computational task variables. Global task objectives may include global targets to be considered when solving the computational task, e.g., global objectives of a solution to the computational task.

For example, continuing the above example of the task of optimizing a water network, the input data 102 may further include data specifying local task objectives such as a constraint on the concentration of chemicals in the water, e.g., constraining the chemical concentration to between 0.2% and 0.5%, and on the number of water pipes in use, e.g., constraining the total number of water pipes to less than 1000. Another example local task objective may be to optimize a particular portion of the water network. In addition, the input data 102 may further include data specifying global task objectives such as one or more global targets, e.g., a target of keeping water wastage to below 2% or a target of distributing at least 10 million gallons of water per day.

In other implementations, data specifying one or more task objectives associated with the computational task may be stored in the system 100 for performing computational tasks, e.g., in task objective data store 112 . For example, as described above, the system 100 for performing computational tasks may be configured to solve multiple computational tasks and the input data 102 may be data that specifies one of the multiple computational tasks. In this example, the system 100 for performing computational tasks may be configured to store task objectives corresponding to each computational task that it is configured to perform. For convenience, data specifying one or more task objectives associated with the computational task is described as being stored in task objective data store 112 throughout the remainder of this document.

The system 100 for performing computational tasks is configured to process the received input data 102 to generate output data 104 . In some implementations, the generated output data 104 may include data representing a solution to the computational task specified by the input data 102 , e.g., a global solution to the computational task based on one or more global task objectives 112 b.

In other implementations or in addition, the output data 104 may include data representing one or more local solutions to the computational task, e.g., one or more initial solutions to the optimization task that are based on local task objectives 112 a and global task objectives 112 b . Local solutions to the optimization task may include solutions to sub tasks of the optimization task. For example, local solutions may include solutions that are optimal over a subset of the parameters associated with the optimization task, e.g., where the subset is specified by the local task objectives 112 a . That is, local solutions may include solutions that are optimal over a subspace, or local space, of a global search space or the optimization task. For example, a local space may be the result of a projection of a multi-dimensional spline representing the global search space to a two-dimensional base space. An example visualization of a global search space and local space 150 is shown in FIG. 1 B . In FIG. 1 B , multi-dimensional spline 152 represents a global search space, and two- dimensional base space 154 represents a local space.

As another example, in cases where the optimization task is a separable task, e.g., a task that may be written as the sum of multiple sub tasks, local solutions may include optimal solutions to each of the sub tasks in the sum of sub tasks, e.g., where the sub tasks are specified by the local task objectives 112 a.

For example, continuing the above example of the task of optimizing a water network, the output data 104 may include data representing a globally optimal configuration (with respect to global task objectives, e.g., wastage targets and distribution targets) of the above described parameters associated with the water network optimization task. Alternatively or in addition, the output data 104 may include data representing multiple local solutions to the water network optimization task, e.g., data specifying an optimal number of water pipes to use, an associated water pressure in each pipe, or a concentration of chemicals in the water flowing through the network. In some implementations, parameter values specified by local solutions may be the same as parameter values specified by a global solution. In other implementations, parameter values specified by local solutions may differ from parameter values specified by a global solution, e.g., a local solution may suggest a chemical concentration of 0.4%, whereas a global solution may suggest a chemical concentration of 0.3%.

The output data 104 may be used to initiate one or more actions associated with the optimization task specified by the input data 102 , e.g., actions 138 . For example, continuing the above example of the task of optimizing a water network, the output data 104 may be used to adjust one or more parameters in the water network, e.g., increase or decrease a current water chemical concentration, increase or decrease a number of water pipes in use, or increase or decrease one or more water pipe pressures.

Optionally, the system 100 for performing computational tasks may include an integration layer 114 and a broker 136 . The integration layer 114 may be configured to manage received input data, e.g., input data 102 . For example, the integration layer 114 may manage data transport connectivity, manage data access authorization, or monitor data feeds coming into the system 100 .

The broker 136 may be configured to receive output data 104 from the system 100 for performing optimization tasks and to generate one or more actions to be taken, e.g., actions 138 . The actions may include local actions, e.g., adjustments to a subset of optimization parameters, which contribute towards achieving local and global targets of the optimization task.

The system 100 for performing computational tasks includes a computation engine 106 . The computation engine 106 is configured to process the received data to obtain solutions to the computational task. The obtained solutions may include a global solution to the computational task that is based on one or more global task objectives 112 b . Alternatively or in addition, the obtained solutions may include one or more initial solutions to the optimization task that are based on the one or more local task objectives 112 a , e.g., one or more local solutions to the computational task. In some implementations, the computation engine 106 may process received input data to obtain one or more initial solutions to the optimization task that are based on local task objectives 112 a , then further process the one or more initial solutions to the optimization task to generate a global solution to the optimization task based on the global task objectives 112 b.

The computation engine 106 may be configured to process received data using one or more computing resources included in the computation engine 106 or otherwise included in the system 100 for performing computational tasks. In other implementations, the computation engine 106 may be configured to process received data using one or more external computing resources, e.g., additional computing resources 110 a - 110 d . For example, the computation engine 106 may be configured to analyze the received input data 102 representing the computational task to be solved and the data representing corresponding task objectives

112 a and 112 b , and outsource one or more computations associated with solving the computational task based on the task objectives

112 a and 112 b to the additional computing resources 110 a - 110 d.

The additional computing resources 110 a - 110 d may include quantum annealer computing resources, e.g., quantum annealer 110 a . A quantum annealer is a device configured to perform quantum annealingâa procedure for finding the global minimum of a given objective function over a given set of candidate states using quantum tunneling. Quantum tunneling is a quantum mechanical phenomenon where a quantum mechanical system overcomes localized barriers in the energy landscape which cannot be overcome by classically described systems.

The additional computing resources 110 a - 110 d may include one or more quantum gate processors, e.g., quantum gate processor 110 b . A quantum gate processor includes one or more quantum circuits, i.e., models for quantum computation in which a computation is performed using a sequence of quantum logic gates, operating on a number of qubits (quantum bits).

The additional computing resources 110 a - 110 d may include one or more quantum simulators, e.g., quantum simulator 110 c . A quantum simulator is a quantum computer that may be programmed to simulate other quantum systems and their properties. Example quantum simulators include experimental platforms such as systems of ultracold quantum gases, trapped ions, photonic systems or superconducting circuits.

The additional computing resources 110 a - 110 d may include one or more classical processors, e.g., classical processor 110 d . In some implementations, the one or more classical processors, e.g., classical processor 110 d , may include supercomputers, i.e., computers with high levels of computational capacity. For example, the classical processor 110 d may represent a computational system with a large number of processors, e.g., a distributed computing system or a computer cluster.

The system 100 for performing computational tasks includes a machine learning module 132 that is configured to learn which, if any, computations to route to the additional computing resources 110 a - 110 d . For example, the machine learning module 132 may include a machine learning model that may be trained using training data to determine when and where to outsource certain computations. The training data may include labeled training examples, e.g., a machine learning model input paired with a respective known machine learning model output, where each training example includes data from multiple resources, as described in more detail below. The machine learning model may process each machine learning model input to generate a respective machine learning model output, compute a loss function between the generated machine learning model output and the known machine learning model, and backpropagate gradients to adjust machine learning model parameters from initial values to trained values. An example machine learning module is described in more detail below with reference to FIG. 2 . Training a machine learning model to route received computations to one or more external computing resources is described in more detail below with reference to FIG. 4 .

The system 100 for performing computational tasks includes a cache 124 . The cache 124 is configured to store different types of data relating to the system 100 and to computational tasks performed by the system 100 . For example, the cache 124 may be configured to store data representing multiple computational tasks previously performed by the system 100 . In some cases the cache 124 may be further configured to store previously generated solutions to computational tasks that the system 100 has previously solved. In some cases this may include solutions to a same computational task, e.g., with different task objectives or different dynamic input data. In other cases this may include solutions to different computational tasks. The cache 124 may be configured to store previously generated solutions to previously received computational tasks from within a specified time frame of interest, e.g., solutions generated within the last 24 hours.

The cache 124 may also be configured to label previously generated solutions. For example, a previously generated solution may be labelled as a successful solution if the solution was generated within a predetermined acceptable amount of time, and/or if a cost associated with generating the solution was lower than a predetermine threshold. Conversely, a previously generated solution may be labelled as an unsuccessful solution if the solution was not generating within a predetermined acceptable amount of time, and/or if a cost associated with generating the solution was higher than a predetermine threshold. Labelling the solution as successful or unsuccessful may include storing data representing the cost associated with generating the solution or data representing a time taken to generate the solution. Such information may be provided for input into the machine learning module 132 . In some cases stored unsuccessful data may be cleaned from the cache 124 , e.g., to free storage space for data representing newly generated solutions.

The cache 124 may also be configured to store system input data associated with the multiple computational tasks previously performed by the system. For example, the input data may include data representing a type of computing resource that each computational task was routed to. In addition, the input data associated with the multiple computational tasks previously performed by the system may further include, for each computational task, one or more of: (i) data representing a size of an input data set associated with the computational task, (ii) data indicating whether an input data set associated with the computational task comprised static, real time or both static and real time input data, (iii) data representing an error tolerance associated with the computational task, or (iv) data representing a required level of confidence associated with the computational task. In some implementations, the cache 124 may further store data indicating a frequency of changes to input data sets associated with each computational tasks. Examples of the different types of input data stored in the cache 124 are described in more detail below with reference to FIG. 3 .

Optionally, the system 100 for performing computational tasks may include a monitoring module 128 . The monitoring module 128 is configured to monitor interactions between and transactions to and from the one or more additional computing resources 110 a - d . For example, the monitoring module 128 may be configured to detect failed or stuck calls to one or more of the additional computing resources 110 a - d . Example failures that can cause a call to one or more of the additional computing resources 110 a - d to fail or get stuck include issues with a transport layer included in the system 100 , i.e., issues with data being moved through the cloud, security login failures, or issues with the additional computing resources 110 a - d themselves such as performance or availability of the additional computing resources 110 a - d . The monitoring module 128 may be configured to process detected failed or stuck calls to one or more of the additional computing resources 110 a - d and determine one or more corrective actions to be taken by the system 100 in response to the failed or stuck calls. Alternatively, the monitoring module 128 may be configured to notify other components of the system 100 , e.g., the global optimization engine 106 or machine learning module 132 , of detected failed or stuck calls to one or more of the additional computing resources 110 a - d.

For example, if one or more computations are outsourced to a particular quantum computing resource, however the particular quantum computing resource suddenly becomes unavailable or is processing outsourced computations too slowly, the monitoring module 128 may be configured to notify relevant components of the system 100 , e.g., the machine learning module 132 . The machine learning module 132 may then be configured to determine one or more suggested corrective actions, e.g., instructing the system 100 to outsource the computation to a different computing resource or to retry the computation using the same computing resource. Generally, the suggested corrective actions may include actions that keep the system 100 successfully operating in real ti

CLAIMS

Claims ( 15 )

What is claimed is:

1. A computer implemented method for obtaining a solution to a computational task, the method comprising:

receiving data representing a computational task to be performed by a system including one or more quantum computing resources and one or more classical computing resources;

determining, by a model that processes at least a portion of the received data and that is configured through training to optimize computational times and computational costs associated with generating a solution to a computational task, a quantum computing resource of the one or more quantum computing resources to route the data to by (i) determining to route the computational task to one of the one or more quantum computing resources instead of one of the one or more classical computing resource and (ii) determining a type of quantum computing resource to route the computational task to, the quantum computing resource being of the type of quantum computing resource, wherein the model is trained on training data that comprises data representing properties associated with using the one or more quantum computing resources to solve computational tasks, the properties comprising numbers of physical qubits available to the quantum computing resources;

routing the data representing the computational task to the quantum computing resource;

processing, by the determined quantum computing resource, the data representing the computational task to generate data representing a solution to the computational task, wherein the determined quantum computing resource comprises a quantum gate computer, an adiabatic annealer, or a quantum simulator, wherein the quantum simulator is a quantum computer;

obtaining, from the quantum computing resource, data representing a solution to the computational task; and

determining one or more actions to be taken based on the data representing a solution to the computational task.

2. The method of claim 1 , wherein the quantum simulator comprises an experimental platform, the experimental platform comprising ultracold quantum gases, trapped ions, photonic systems or superconducting circuits.

3. The method of claim 1 , wherein the computational task comprises an optimization task.

4. The method of claim 1 , wherein the model comprises a machine learning model that has been configured through training to route received data representing computational tasks to be performed in a system including at least one quantum computing resource, the training comprising training the model using (i) data representing multiple computational tasks previously performed by the system, and (ii) data representing a type of computing resource to which the task was routed.

5. The method of claim 4 , wherein the data representing multiple computational tasks previously performed by the system comprises, for each computational task, one or more of:

(i) data representing a size of an input data set associated with the computational task;

(ii) data indicating whether an input data set associated with the computational task comprised static, real time or both static and real time input data;

(iii) data representing an error tolerance associated with the computational task; and

(iv) data representing a required level of confidence associated with the computational task.

6. The method of claim 4 , wherein the data representing multiple computational tasks previously performed by the system comprises data indicating a frequency of changes to input data sets associated with each computational tasks.

7. The method of claim 4 , further comprising training the machine learning model using data representing properties associated with using the one or more quantum computing resource to solve the multiple computational tasks.

8. The method of claim 7 , wherein properties associated with using the one or more quantum computing resource to solve the multiple computational tasks comprises:

for each computational task, one or more of:

(i) approximate qualities of solutions generated by the one or more quantum computing resource;

(ii) computational times associated with solutions generated by the one or more quantum computing resource; or

(iii) computational costs associated with solutions generated by the one or more quantum computing resource, and

for each quantum computing resource

a cost associated with using the quantum computing resource.

9. The method of claim 1 , further comprising routing another computational task to a classical computing resource.

10. A system comprising one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:

receiving data representing a computational task to be performed by a system including one or more quantum computing resources and one or more classical computing resources;

routing the data representing the computational task to the quantum computing resource;

obtaining, from the quantum computing resource, data representing a solution to the computational task; and

determining one or more actions to be taken based on the data representing a solution to the computational task.

11. The system of claim 10 , wherein the quantum simulator comprises an experimental platform, the experimental platform comprising ultracold quantum gases, trapped ions, photonic systems or superconducting circuits.

12. The system of claim 10 , wherein the model comprises a machine learning model that has been configured through training to route received data representing computational tasks to be performed in a system including at least one quantum computing resource, the training comprising training the model using (i) data representing multiple computational tasks previously performed by the system, and (ii) data representing a type of computing resource to which the task was routed.

13. The system of claim 12 , wherein the data representing multiple computational tasks previously performed by the system comprises, for each computational task, one or more of:

(i) data representing a size of an input data set associated with the computational task;

(ii) data indicating whether an input data set associated with the computational task comprised static, real time or both static and real time input data,

(iii) data representing an error tolerance associated with the computational task; and

(iv) data representing a required level of confidence associated with the computational task.

14. The system of claim 12 , wherein the data representing multiple computational tasks previously performed by the system comprises data indicating a frequency of changes to input data sets associated with each computational tasks.

15. The system of claim 10 , wherein the model is a machine learning model and wherein the operations further comprise training the machine learning model using data representing properties associated with using the one or more quantum computing resource to solve the multiple computational tasks.

US16/298,860

2017-04-19

2019-03-11

Quantum computing machine learning module

Active

2039-11-11

US11803772B2

( en )

Priority Applications (1)

Application Number

Priority Date

Filing Date

Title

US16/298,860

US11803772B2

( en )

2017-04-19

2019-03-11

Quantum computing machine learning module

Applications Claiming Priority (2)

Application Number

Priority Date

Filing Date

Title

US15/491,852

US10275721B2

( en )

2017-04-19

Quantum computing machine learning module

US16/298,860

US11803772B2

( en )

2017-04-19

2019-03-11

Quantum computing machine learning module

Related Parent Applications (1)

Application Number

Title

Priority Date

Filing Date

US15/491,852

Continuation

US10275721B2

( en )

2017-04-19

Quantum computing machine learning module

Publications (2)

Publication Number

Publication Date

US20190205790A1

US20190205790A1 ( en )

2019-07-04

US11803772B2

true

US11803772B2 ( en )

2023-10-31

Family

ID=61683604

Family Applications (2)

Application Number

Title

Priority Date

Filing Date

US15/491,852

Active

2037-06-18

US10275721B2

( en )

2017-04-19

Quantum computing machine learning module

US16/298,860

Active

2039-11-11

US11803772B2

( en )

2017-04-19

2019-03-11

Quantum computing machine learning module

Family Applications Before (1)

Application Number

Title

Priority Date

Filing Date

US15/491,852

Active

2037-06-18

US10275721B2

( en )

2017-04-19

Quantum computing machine learning module

Country Status (4)

Country

Link

( 2 )

US10275721B2

( en )

( 1 )

EP3392809B1

( en )

( 1 )

CN108734299B

( en )

( 1 )

CA2997970C

( en )

Families Citing this family (74)

* Cited by examiner, â Cited by third party

Publication number

Priority date

Publication date

Assignee

Title

US11797641B2

( en )

2015-02-03

2023-10-24

1Qb Information Technologies Inc.

Method and system for solving the lagrangian dual of a constrained binary quadratic programming problem using a quantum annealer

CA2881033C

( en )

2015-02-03

2016-03-15

1Qb Information Technologies Inc.

Method and system for solving lagrangian dual of a constrained binary quadratic programming problem

US10275721B2

( en )

2017-04-19

2019-04-30

Accenture Global Solutions Limited

Quantum computing machine learning module

US10977570B2

( en )

2017-06-19

2021-04-13

Rigetti & Co, Inc.

Distributed quantum computing system

US10817337B1

( en )

2017-11-15

2020-10-27

Amazon Technologies, Inc.

Cloud-based access to quantum computing resources

US11270220B1

( en )

2017-11-15

2022-03-08

Amazon Technologies, Inc.

Service for managing quantum computing resources

US11170137B1

( en )

2017-11-15

2021-11-09

Amazon Technologies, Inc.

Cloud-based simulation of quantum computing resources

US10592216B1

( en )

2017-11-15

2020-03-17

Amazon Technologies, Inc.

Development environment for programming quantum computing resources

WO2019104443A1

( en )

2017-12-01

2019-06-06

1Qb Information Technologies Inc.

Systems and methods for stochastic optimization of a robust inference problem

US11308248B2

( en )

2018-05-05

2022-04-19

Intel Corporation

Apparatus and method for quantum computing performance simulation

JP6992891B2

( en )

2018-05-25

2022-01-13

æ¥æ¬é»æ°æ ªå¼ä¼ç¤¾

Learning devices, information processing systems, learning methods, and learning programs

US20210264307A1

( en )

2018-06-26

2021-08-26

Nec Corporation

Learning device, information processing system, learning method, and learning program

US11568293B2

( en )

2018-07-18

2023-01-31

Accenture Global Solutions Limited

Quantum formulation independent solver

WO2020072819A1

( en )

2018-10-03

2020-04-09

Rigetti & Co, Inc.

Parcelled quantum resources

CN109559786B

( en )

2018-11-08

2022-04-08

ä¸å³¡å¤§å¦

Pilot compound discovery and synthesis method based on quantum group intelligent optimization

EP3891668A4

( en )

2018-12-06

2023-01-25

1QB Information Technologies Inc.

QUANTUM COMPUTING CONTROLLED BY ARTIFICIAL INTELLIGENCE

EP3667512A1

( en )

2018-12-11

2020-06-17

Siemens Aktiengesellschaft

A cloud platform and method for efficient processing of pooled data

CN111464154B

( en )

2019-01-22

2022-04-22

åä¸ºææ¯æéå¬å¸

Control pulse calculation method and device

CN109800883B

( en )

2019-01-25

2020-12-04

åè¥æ¬æºéåè®¡ç®ç§ææéè´£ä»»å¬å¸

Quantum machine learning framework construction method, device and quantum computer

US11586864B2

( en )

2019-02-15

2023-02-21

International Business Machines Corporation

Prior knowledge-based topological feature classification

US11580433B2

( en )

2019-03-09

2023-02-14

International Business Machines Corporation

Validating and estimating runtime for quantum algorithms

CN110135685B

( en )

2019-04-04

2022-08-16

ä¸è½¦å·¥ä¸ç ç©¶é¢æéå¬å¸

Method and device for distributing product design tasks

US11182800B2

( en )

2019-04-08

2021-11-23

Bank Of America Corporation

Controlling enterprise software policy compliance assessment processes based on quantum combinations of assessment elements

US11620569B2

( en )

2019-04-26

2023-04-04

International Business Machines Corporation

Machine learning quantum algorithm validator

WO2020227825A1

( en )

2019-05-13

2020-11-19

1Qb Information Technologies Inc.

Methods and systems for quantum computing enabled molecular ab initio simulations

CN110365473B

( en )

2019-05-31

2021-11-26

åäº¬é®çµå¤§å¦

Active feedback control method of quantum communication system based on machine learning

WO2020255076A1

( en )

2019-06-19

2020-12-24

1Qb Information Technologies Inc.

Method and system for mapping a dataset from a hilbert space of a given dimension to a hilbert space of a different dimension

US11336462B1

( en )

2019-09-10

2022-05-17

Wells Fargo Bank, N.A.

Systems and methods for post-quantum cryptography optimization

CN110674009B

( en )

2019-09-10

2022-10-28

å¹³å®æ®æ ä¼ä¸ç®¡çæéå¬å¸

Application server performance monitoring method and device, storage medium and electronic equipment

US11551129B2

( en )

2019-09-30

2023-01-10

International Business Machines Corporation

Quantum platform routing of a quantum application component

US11704715B2

( en )

2019-11-27

2023-07-18

Amazon Technologies, Inc.

Quantum computing service supporting multiple quantum computing technologies

US11605016B2

( en )

2019-11-27

2023-03-14

Amazon Technologies, Inc.

Quantum computing service supporting local execution of hybrid algorithms

US11605033B2

( en )

2019-11-27

2023-03-14

Amazon Technologies, Inc.

Quantum computing task translation supporting multiple quantum computing technologies

CA3157216A1

( en )

2019-12-03

2021-06-10

Pooya Ronagh

System and method for enabling an access to a physics-inspired computer and to a physics-inspired computer simulator

CA3161500A1

( en )

2019-12-10

2021-06-17

Paul Victor KLIMOV

Denoising qubit calibration data with deep learning

CN114868136B

( en )

2020-01-08

2025-08-29

åä¸ºææ¯æéå¬å¸

Method, device and system for marking data labels

US12126713B1

( en )

2020-01-17

2024-10-22

Wells Fargo Bank, N.A.

Systems and methods for quantum computing threat detection

US11366897B1

( en )

2020-01-17

2022-06-21

Wells Fargo Bank, N.A.

Systems and methods for layered quantum computing detection

US11334667B1

( en )

2020-01-17

2022-05-17

Wells Fargo Bank, N.A.

Systems and methods for disparate quantum computing threat detection

CN111241356B

( en )

2020-04-26

2020-08-11

è¾è®¯ç§æï¼æ·±å³ï¼æéå¬å¸

Data search method, device and equipment based on analog quantum algorithm

CA3179781A1

( en )

2020-05-27

2021-12-02

Silvan Shiwa KUTTIMALAI

Methods and systems for solving an optimization problem using a flexible modular approach

US11710060B2

( en )

2020-06-17

2023-07-25

The Boeing Company

Quantum processing system

US11539796B2

( en )

2020-07-08

2022-12-27

Bank Of America Corporation

System for intelligent sensor data transfer and device manipulation leveraging quantum optimization engine

CN112561069B

( en )

2020-12-23

2021-09-21

åäº¬ç¾åº¦ç½è®¯ç§ææéå¬å¸

Model processing method, device, equipment and storage medium

US20220283878A1

( en )

2021-03-02

2022-09-08

Zapata Computing, Inc.

Dependency-based data routing for distributed computing

US20230252501A1

( en )

2021-04-16

2023-08-10

Strong Force Vcn Portfolio 2019, Llc

Control-Tower-Based Digital Product Network System

JP2024515983A

( en )

2021-04-27

2024-04-11

ã¶ã»ã¦ããã¼ã·ãã£ã»ãªãã»ã·ã«ã´

System and method for optimizing quantum cloud computing in-queue

US12555046B2

( en )

2021-06-21

2026-02-17

International Business Machines Corporation

Quantum machine learning model feature space generation

CN113553028B

( en )

2021-07-20

2024-03-12

ä¸å½ç§å¦é¢åå¯¼ä½ç ç©¶æ

Problem solving optimization method and system based on probabilistic bit circuit

US12217090B2

( en )

2021-11-12

2025-02-04

Amazon Technologies, Inc.

On-demand co-processing resources for quantum computing

US11907092B2

( en )

2021-11-12

2024-02-20

Amazon Technologies, Inc.

Quantum computing monitoring system

US12333379B2

( en )

2021-12-02

2025-06-17

International Business Machines Corporation

Optimized selection of data for quantum circuits

US11436519B1

( en )

2021-12-23

2022-09-06

Quantum Computing Inc.

Machine learning mapping for quantum processing units

US20250117552A1

( en )

2022-02-11

2025-04-10

The Trustees Of Princeton University

Inverse system design for constrained multi-objective optimization

CN114358318B

( en )

2022-03-22

2022-06-21

åè¥æ¬æºéåè®¡ç®ç§ææéè´£ä»»å¬å¸

Machine learning framework-based classification method and related device

US20230325703A1

( en )

2022-04-11

2023-10-12

Red Hat, Inc.

Optimizing execution of quantum service definition files using a quantum optimization database

US20230368065A1

( en )

2022-05-10

2023-11-16

Mitsubishi Electric Research Laboratories, Inc.

Integrated Sensing and Communications Empowered by Networked Hybrid Quantum-Classical Machine Learning

CN115062721B

( en )

2022-07-01

2023-10-31

ä¸å½çµä¿¡è¡ä»½æéå¬å¸

Network intrusion detection methods and devices, computer readable media, electronic equipment

US12212988B2

( en )

2022-08-09

2025-01-28

T-Mobile Usa, Inc.

Identifying a performance issue associated with a 5G wireless telecommunication network

US12585975B2

( en )

2022-08-17

2026-03-24

Red Hat, LLC

State maps for quantum computing

US12617431B1

( en )

2022-09-30

2026-05-05

Zoox, Inc.

Risk determination for autonomous vehicle simulations

US20240160959A1

( en )

2022-11-11

2024-05-16

Dell Products L.P.

Self-learning quantum computing platform

US20240160490A1

( en )

2022-11-11

2024-05-16

Dell Products L.P.

Transpilation-oriented decisions in hybrid quantum-classic workload orchestration

US12505366B2

( en )

2022-11-30

2025-12-23

Nvidia Corporation

Simulating quantum computing circuits using Kronecker factorization

CN115576999B

( en )

2022-12-09

2023-03-10

Task data processing method, device and equipment based on cloud platform and storage medium

US20240378244A1

( en )

2023-01-30

2024-11-14

University Of Southern California

Real time search filters using quantum machine learning

CN116205301A

( en )

2023-01-31

2023-06-02

èå·æµªæ½®æºè½ç§ææéå¬å¸

Training frame construction method, device and system based on quantum machine learning

CN116167447B

( en )

2023-02-20

2024-05-14

åäº¬ç¾åº¦ç½è®¯ç§ææéå¬å¸

Quantum circuit processing method and device and electronic equipment

US20240394578A1

( en )

2023-05-22

2024-11-28

Dell Products L.P.

Orchestration for identifying a quantum annealer to solve combinatorial optimization problems

US12519856B1

( en )

2024-08-27

2026-01-06

At&T Intellectual Property I, L.P.

System and method for managing an enhanced performance network using role-based agents

CN119247515B

( en )

2024-09-25

2026-02-24

ä¸çµä¿¡éåä¿¡æ¯ç§æéå¢æéå¬å¸

Weather forecasting method based on quantum convolutional attention module

WO2026087913A1

( en )

2024-10-22

2026-04-30

Multiverse Computing S.L.

Method and system for enhancing performance of large language models using quantum circuits

WO2026087091A1

( en )

2024-10-22

2026-04-30

Multiverse Computing S.L.

Method and system for enhancing performance of large language models using quantum circuits

US12547466B1

( en )

2025-07-09

2026-02-10

Qubital LLC

Method and system for adaptive quantum backend selection

Citations (10)

* Cited by examiner, â Cited by third party

Publication number

Priority date

Publication date

Assignee

Title

US20090070402A1

( en )

2007-09-11

2009-03-12

Geordie Rose

Systems, methods, and apparatus for a distributed network of quantum computers

US20090228888A1

( en )

2008-03-10

2009-09-10

Sun Microsystems, Inc.

Dynamic scheduling of application tasks in a distributed task based system

US20120275690A1

( en )

2011-04-26

2012-11-01

Nec Laboratories America, Inc.

Distributed artificial intelligence services on a cell phone

CN103001875A

( en )

2013-01-07

2013-03-27

A dynamic routing method for quantum cryptography network

CN103003797A

( en )

2010-07-16

2013-03-27

è¯ºåºäºå¬å¸

Method and apparatus for distributing computation closures

US20140298343A1

( en )

2013-03-26

2014-10-02

Xerox Corporation

Method and system for scheduling allocation of tasks

US9152746B2

( en )

2013-03-26

2015-10-06

Microsoft Technology Licensing, Llc

Quantum annealing simulator

US9537953B1

( en )

2016-06-13

2017-01-03

1Qb Information Technologies Inc.

Methods and systems for quantum ready computations on the cloud

US10127499B1

( en )

2014-08-11

2018-11-13

Rigetti & Co, Inc.

Operating a quantum processor in a heterogeneous computing architecture

US10275721B2

( en )

2017-04-19

2019-04-30

Accenture Global Solutions Limited

Quantum computing machine learning module

2017

2017-04-19

US15/491,852

patent/US10275721B2/en

active

Active

2018

2018-03-12

CA2997970A

patent/CA2997970C/en

active

Active

2018-03-15

EP18161882.8A

patent/EP3392809B1/en

active

Active

2018-04-18

CN201810350991.4A

patent/CN108734299B/en

active

Active

2019

2019-03-11

US16/298,860

patent/US11803772B2/en

active

Active

Patent Citations (10)

* Cited by examiner, â Cited by third party

Publication number

Priority date

Publication date

Assignee

Title

US20090070402A1

( en )

2007-09-11

2009-03-12

Geordie Rose

Systems, methods, and apparatus for a distributed network of quantum computers

US20090228888A1

( en )

2008-03-10

2009-09-10

Sun Microsystems, Inc.

Dynamic scheduling of application tasks in a distributed task based system

CN103003797A

( en )

2010-07-16

2013-03-27

è¯ºåºäºå¬å¸

Method and apparatus for distributing computation closures

US20120275690A1

( en )

2011-04-26

2012-11-01

Nec Laboratories America, Inc.

Distributed artificial intelligence services on a cell phone

CN103001875A

( en )

2013-01-07

2013-03-27

A dynamic routing method for quantum cryptography network

US20140298343A1

( en )

2013-03-26

2014-10-02

Xerox Corporation

Method and system for scheduling allocation of tasks

US9152746B2

( en )

2013-03-26

2015-10-06

Microsoft Technology Licensing, Llc

Quantum annealing simulator

US10127499B1

( en )

2014-08-11

2018-11-13

Rigetti & Co, Inc.

Operating a quantum processor in a heterogeneous computing architecture

US9537953B1

( en )

2016-06-13

2017-01-03

1Qb Information Technologies Inc.

Methods and systems for quantum ready computations on the cloud

US10275721B2

( en )

2017-04-19

2019-04-30

Accenture Global Solutions Limited

Quantum computing machine learning module

Non-Patent Citations (8)

* Cited by examiner, â Cited by third party

Title

âarxiv.orgâ [online]. " Quantum Computation by Adiabatic Evolution, " Jan. 2000 [retrieved on Aug. 28, 2017]. Retrieved from the Internet: URL <https://arxiv.org/abs/quant-ph/0001106>.

âRedcedartech.comâ [online]. " How to Select the Right Optimization Method for Your Problem, " Aug. 21, 2010 [retrieved on Aug. 28, 2017]. Retrieved from the Internet: URL <http://www.redcedartech.com/pdfs/Select_Optimization_Method.pdf>. 2 pages.

âResearch.cs.wisc.eduâ [online]. " Adaptive Scheduling for Master-Worker Applications on the Computational Grid, " [retrieved on Aug. 28, 2017]. Retrieved from the Internet: URL http://research.cs.wisc.edu/htcondor/doc/camera.pdf.

Ahamed, Quantum Recurrent Neural Networks for Filtering, Doctoral Thesis, Department of Computer Science, University of Hull, 2009, pp. 1-148.

CN Office Action in Chinese Appln. No. 201810350991.4, dated May 8, 2021, 17 pages (with English translation).

EP Office Action in European Appln No. 18161882.8, dated Feb. 7, 2020, 5 pages.

EP Search Report and Written Opinion in European Application No. 18161882.8, dated Sep. 6, 2018, 10 pages.

Hongzai Mao et al., " Resource management with deep reinforcement learning ", 15th ACM Workshop on Hot Topics in Networks, Hotnets '16, Nov. 9, 2016, 7 pages.

Also Published As

Publication number

Publication date

CA2997970A1

( en )

2018-10-19

US20190205790A1

( en )

2019-07-04

CA2997970C

( en )

2020-03-24

EP3392809B1

( en )

2021-11-17

EP3392809A1

( en )

2018-10-24

US20180308000A1

( en )

2018-10-25

CN108734299A

( en )

2018-11-02

US10275721B2

( en )

2019-04-30

CN108734299B

( en )

2022-05-10

Quantum computing machine learning module — Accenture Global Solutions Limited (US11803772B2)

Related documents