DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim(s) 1-21 is/are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 1 (similarly claims 11 and 21) recite: “their” constituent information. The examiner is unclear what is being referred to. Nodes, edges, computational processes, and/or any combination, etc.
Claim 1 (similarly claims 11 and 21) recite: “constituent information”. The examiner is unclear how the term “constituent” should be interpreted. The specification does not provide any disclosures as to what a constituent information is.
Claim 1 (similarly claims 7, 10, 11, 17, 20 and 21) recite: “classical computing”. The term classical is a relative term. Therefore, the examiner is unclear what specific type of computing resources would be considered classical vs non-classical.
Claim 5 (similarly claim 15) recite: “data element is respected”. The examiner is unclear how the term “respected” should be interpreted. Although, PGPub paragraph 43 provides an example: “For example, if a user specifies a data element v representing a measured value known at the start of a process (based on knowledge of the system), this constraint ensures that this data element is true when starting the optimization.”
The examiner has to consider the provided description is a mere example thus alternative interpretation can exist. Therefore, the term “respected” is ambiguous.
Claim 6 (similarly claim 16) recites the limitation "the start of the optimization". There is insufficient antecedent basis for this limitation in the claim. The examiner is unclear if “the start of the optimization” is referring to the network optimization (preamble) or the solving of the optimization problem.
Claim 7 (similarly claim 17) recite: “within equipped resources”. The examiner is unclear how this should be interpreted.
Claims 2-10 and 12-20 are rejected based on rejection of its corresponding dependent claim.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1, 9-11 and 19-21 is/are rejected under 35 U.S.C. 103 as being unpatentable over Pinho et al. (Pub 20230229514) (hereafter Pinho) in view of Kuttimalai et al. (Pub 20230077665) (hereafter Kuttimalai).
As per claim 1, Pinho teaches:
A computer-implemented method of network optimization for arranging computational sub-tasks in a hybrid-computing environment, the method comprising: ([Paragraph 17], Embodiments of the present invention generally relate to execution of hybrid quantum-classical workloads. More particularly, at least some embodiments of the invention relate to systems, hardware, software, computer-readable media, and methods for constructing and operating a framework that is able to determine an amount of resources for, and to optimize the execution of, hybrid quantum-classical computing workloads. [Paragraph 28], A follow-up of VQAs is the combination of several quantum and classical computing modules to solve more complex problems, as illustrated in the example hybrid quantum-classic computation graph 200 and workflow that is disclosed in FIG. 2. As shown, the example graph 200 includes a combination of classic computing modules, represented by nodes 202, and quantum computing modules, represented by nodes 204.)
receiving input of a network of nodes and edges representing computational processes and their constituent information, wherein the nodes are grouped according to whether the nodes use classical computing resources or quantum computing resources; ([Paragraph 28], A follow-up of VQAs is the combination of several quantum and classical computing modules to solve more complex problems, as illustrated in the example hybrid quantum-classic computation graph 200 and workflow that is disclosed in FIG. 2. As shown, the example graph 200 includes a combination of classic computing modules, represented by nodes 202, and quantum computing modules, represented by nodes 204. The various nodes 202/204 may be connected to each other in different arrangements with one or more edges 206 that represent interdependences between/among the connected nodes 202/204. Note, with respect to the example of FIG. 2, that the same set of nodes 202/204 may be connected in any of a variety of different ways, depending on the problem(s) to be solved. In purely quantum computing implementations, such problems would probably require intractable circuit depths. By building, according to example embodiments of the invention, a hybrid workflow in which a complex solution may be developed from a combination of simpler quantum and classic computation steps, the quantum computing pieces may even be executed in simulation engines running on classical devices. [Paragraph 54], receiving a computation workflow defined by a graph that includes a group of computing nodes comprising quantum computing nodes; [Paragraph 58], Embodiment 5. The method as recited in any of embodiments 1-4, wherein the group of nodes comprises classic computing nodes. [Paragraph 31], The pipeline 300 may receive, as input, the workflow graph 302 and a catalogue 304 of all quantum computing instances available in a hybrid quantum-classic computation infrastructure. Such computation instances may be simulation engines or quantum computing devices. The latter may be available on-premises and/or in a cloud computing environment.)
generating workflow constraints;
generating scheduling constraints;
generating computing resource assignment constraints;
generating an objective function; ([Paragraph 18], Example embodiments may leverage any available on-premises classical hardware for both classical computation and quantum simulation, and leverage any available quantum hardware for quantum computations to satisfy performance constraints and SLAs (service level agreements). Some embodiments employ an orchestration pipeline that receives a computation workflow in the form of a graph, and then alters the graph in such a way that each node of the graph contains the resource allocation configuration that is needed to execute the workflow while also satisfying SLA (service level agreement) constraints. [Paragraph 34], In the OptimizeWorkflow stage, which may be carried out by an optimizer module 308, an optimization algorithm is executed to transform the interim workflow graph 302a to obtain a specific resource allocation configuration for each node so that the provided global SLA constraint is satisfied. That is, the transformation of interim workflow 302a (custom-character), so as to generate the final graph 302b custom-character with the selected resource allocations, may comprise selecting, for each node, a specific resource allocation for that node. The final graph 302b custom-character with the selected resource allocations may then be provided as an input to the Executestage, which may be carried out by a module 310. Some embodiments may employ an execution scheduler, of the type used in a containerization engine for example, to perform the execution of the workflow using the specified resources. [Paragraph 28], Note, with respect to the example of FIG. 2, that the same set of nodes 202/204 may be connected in any of a variety of different ways, depending on the problem(s) to be solved. In purely quantum computing implementations, such problems would probably require intractable circuit depths. By building, according to example embodiments of the invention, a hybrid workflow in which a complex solution may be developed from a combination of simpler quantum and classic computation steps....)
solving an optimization problem according to the objective function and all said constraints, wherein the solution determines a best computational objective achieved, a selected computational workflow through the nodes, compute job scheduling, and assignment of the computational processes among the classical computing resources and quantum computing resources; and ([Paragraph 18], In general, example embodiments of the invention may bring together both classical and quantum computing in a framework that is operable to decide the amount of resources for, and to optimize the execution of, hybrid quantum-classic computation problems. Example embodiments may leverage any available on-premises classical hardware for both classical computation and quantum simulation, and leverage any available quantum hardware for quantum computations to satisfy performance constraints and SLAs (service level agreements). Some embodiments employ an orchestration pipeline that receives a computation workflow in the form of a graph, and then alters the graph in such a way that each node of the graph contains the resource allocation configuration that is needed to execute the workflow while also satisfying SLA (service level agreement) constraints.)
executing the computational workflow to achieve the best computational objective according to the computed job scheduling and assignment of computational processes among the classical computing resources and quantum computing resources. ([Paragraph 18], In general, example embodiments of the invention may bring together both classical and quantum computing in a framework that is operable to decide the amount of resources for, and to optimize the execution of, hybrid quantum-classic computation problems. Example embodiments may leverage any available on-premises classical hardware for both classical computation and quantum simulation, and leverage any available quantum hardware for quantum computations to satisfy performance constraints and SLAs (service level agreements). Some embodiments employ an orchestration pipeline that receives a computation workflow in the form of a graph, and then alters the graph in such a way that each node of the graph contains the resource allocation configuration that is needed to execute the workflow while also satisfying SLA (service level agreement) constraints. [Paragraph 34], In the OptimizeWorkflow stage, which may be carried out by an optimizer module 308, an optimization algorithm is executed to transform the interim workflow graph 302a to obtain a specific resource allocation configuration for each node so that the provided global SLA constraint is satisfied. That is, the transformation of interim workflow 302a (custom-character), so as to generate the final graph 302b custom-character with the selected resource allocations, may comprise selecting, for each node, a specific resource allocation for that node. The final graph 302b custom-character with the selected resource allocations may then be provided as an input to the Executestage, which may be carried out by a module 310. Some embodiments may employ an execution scheduler, of the type used in a containerization engine for example, to perform the execution of the workflow using the specified resources.)
Although Pinho discloses determining of an optimal computational objective.
Pinho does not explicitly disclose solution determines a best computational objective.
Kuttimalai discloses solution determines a best computational objective. ([Paragraph 206], According to the processing operation 220, the result is obtained. The result represents the best feasible solution among the solutions collected in processing operation 218. [Paragraph 101], A hybrid architecture (e.g., computing system) comprising a classical computer and a quantum computer can be very efficient for addressing complex computational tasks… [Paragraph 47], A feasible configuration may be a configuration which satisfies each of the set of constraints. In some cases, a feasible configuration may be a configuration x which satisfies problem constraints in the set C(x).)
Kuttimalai also discloses generating workflow constraints; generating scheduling constraints; generating computing resource assignment constraints;
generating an objective function; ([Paragraph 44, 55, 82-94 100])
It would have been obvious to a person with ordinary skill in the art, before the effective filing date of the invention, to combine the teachings of Pinho wherein input representing (i.e. nodes and edges) computational processes are received, nodes are grouped according to quantum or classical computing resources, constraints and objective functions are generated to provide an optimal solution for a workflow to achieve optimal solution for the generated objective function to be executed via the workflow, into teachings of Kuttimalai wherein the optimal solution is the best feasible solution generated based on plurality of constraints, because this would enhance the teachings of Pinho wherein the best feasible solution can be generated to achieve the objective function based on the constraints. [Kuttimalai paragraph 47]
As per claim 9, rejection of claim 1 is incorporated:
Pinho teaches wherein the scheduling constraints comprise: scheduling duration constraints; and temporal discretization constraints. ([Paragraph 36], Through their experimentation, the inventors have discovered that, in quantum simulations, the amount of resources used by a quantum algorithm is related to the number of qubits and the depth of the associated quantum circuit. By executing the same quantum algorithms several times with different resource allocation configurations, the impact of each configuration on SLA metrics such as execution time, and cost, for example, can be measured. These executions of the quantum algorithms generate large amounts of data that can be used to learn a function m=f.sub.sim(Q, D, R|QC.sub.sim), where m is an SLA metric of interest, Q is the number of qubits of the quantum circuit of the algorithm, D is the circuit depth, and R is the resource allocation configuration for a given quantum simulation instance QC.sub.sim, that is, for a single given run of the quantum algorithm. The configuration may translate into, but is not limited to, a grouping of specific numbers and types of resources such as, for example, the number of CPU cores, the amount of memory, the amount of disk allocated to the process, container, or virtual machine, that runs the quantum algorithm. FIG. 4 shows example graphs 402 and 404 of data collected from the execution of a quantum algorithm on a simulation environment. More specifically, FIG. 4 discloses collected data from quantum algorithm executions on a simulation environment. The graphs 402 and 404 respectively show, for a given {qubits, depth} quantum circuit configuration, the SLA metric “execution time” as a function of the allocated memory, and of the number of CPU cores assigned to the simulation engine execution. [Paragraph 37], The case with estimating SLA metrics for a true quantum device is somewhat different that just described. Although SLA metrics in quantum devices are also affected by the number of qubits and depth of algorithm, the concept of resource allocation above does not directly apply. Other factors, such as queuing time, data loading times and required number of measurements of outputs affect execution times and costs. Still, data such as that in FIG. 4 can also be collected to model a function m=f.sub.QC(Q, D|QC.sub.dev), which measures an SLA metric as a function of the execution of a quantum circuit on a given quantum device QC.sub.dev.)
As per claim 10, rejection of claim 1 is incorporated:
Pinho teaches wherein the computing resource assignment constraints comprise activity constraints that ensure the activity of machines running the classical and quantum computing resources matches in time to startings and endings of computational processes assigned to those machines. ([Paragraph 33], In the first stage of the pipeline 300, which may be carried out by a module 306, of the pipeline, EstimateWorkflowResources, each graph node (see example nodes 202 and 204 in FIG. 2) of the input workflow graph 302 is assigned all candidate resource allocation configurations, given quantum circuit parameters, and estimations of SLA metrics for the execution of the graph node with the candidate configuration. Thus, the output of the stage 306 is an interim workflow graph 302a (custom-character), which is a transformation of the input workflow graph 302 (custom-character). That is, the interim workflow graph 302a has been transformed, at the first stage, to include candidate resource allocations for each node of the workflow graph 302. [Paragraph 36], Through their experimentation, the inventors have discovered that, in quantum simulations, the amount of resources used by a quantum algorithm is related to the number of qubits and the depth of the associated quantum circuit. By executing the same quantum algorithms several times with different resource allocation configurations, the impact of each configuration on SLA metrics such as execution time, and cost, for example, can be measured. These executions of the quantum algorithms generate large amounts of data that can be used to learn a function m=f.sub.sim(Q, D, R|QC.sub.sim), where m is an SLA metric of interest, Q is the number of qubits of the quantum circuit of the algorithm, D is the circuit depth, and R is the resource allocation configuration for a given quantum simulation instance QC.sub.sim, that is, for a single given run of the quantum algorithm. The configuration may translate into, but is not limited to, a grouping of specific numbers and types of resources such as, for example, the number of CPU cores, the amount of memory, the amount of disk allocated to the process, container, or virtual machine, that runs the quantum algorithm. FIG. 4 shows example graphs 402 and 404 of data collected from the execution of a quantum algorithm on a simulation environment. More specifically, FIG. 4 discloses collected data from quantum algorithm executions on a simulation environment. The graphs 402 and 404 respectively show, for a given {qubits, depth} quantum circuit configuration, the SLA metric “execution time” as a function of the allocated memory, and of the number of CPU cores assigned to the simulation engine execution. [Paragraph 37], The case with estimating SLA metrics for a true quantum device is somewhat different that just described. Although SLA metrics in quantum devices are also affected by the number of qubits and depth of algorithm, the concept of resource allocation above does not directly apply. Other factors, such as queuing time, data loading times and required number of measurements of outputs affect execution times and costs. Still, data such as that in FIG. 4 can also be collected to model a function m=f.sub.QC(Q, D|QC.sub.dev), which measures an SLA metric as a function of the execution of a quantum circuit on a given quantum device QC.sub.dev.)
As per claims 11, 19 and 20, these are system claims corresponding to the method claims 1, 9 and 10. Therefore, rejected based on similar rationale.
As per claim 20, this is a computer-readable storage medium claim corresponding to the method claim 1. Therefore, rejected based on similar rationale.
Claim(s) 2-8 and 12-18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Pinho in view of Kuttimalai and further in view of Gunnels et al. (Pub 20200285947) (hereafter Gunnels).
As per claim 2, rejection of claim 1 is incorporated:
Kuttimalai teaches wherein the workflow constraints comprise: node and edge activation constraints; node and edge relationship constraints; domain input constraints; domain output constraints; resource constraints; time definition constraints; and total cost constraints. ([Paragraph 6], In an aspect, a method for solving an optimization problem inputted by a user is provided. The method may comprise: (a) using an interface of a digital computer to receive an indication of (i) said optimization problem, (ii) one or more optimization problem specifications of a plurality of optimization problem specifications, (iii) at least one generation method and at least one evaluation method, (iv) one or more algorithms of a plurality of algorithms, wherein said one or more algorithms is configured to solve said optimization problem subject to at least said one or more optimization problem specifications using said at least one generation method and said at least one evaluation method, and (v) one or more computational platforms of a plurality of computational platforms, wherein said one or more computational platforms is configured to execute said one or more algorithms; (b) transmitting a request to said one or more computational platforms to execute said one or more algorithms, wherein upon execution of said one or more algorithms by said one or more computational platforms said one or more computation platforms yields a solution to said optimization problem; and (c) providing said solution to said optimization problem from said computational platform at said interface. [Paragraph 44], An optimization problem may be denoted by P herein. Each constraint may represent a relationship among decision variables x. In a general optimization problem, there may be no restriction on decision variables' type and value. In a general optimization problem, there may be no restriction on the form of the objective functions and constraints. In some cases, the definition may include finding a near optimal solution or a number of near optimal solutions. [Paragraph 47], A feasible configuration may be a configuration which satisfies each of the set of constraints. In some cases, a feasible configuration may be a configuration x which satisfies problem constraints in the set C(x). [Paragraph 67], In some cases, a module for selecting a computational platform comprises or is operably connected to a hardware interface. The hardware interface may facilitate connection to various hardware types, including classical and non-classical computational platforms, to facilitate implementation of various types of optimization problems. The hardware interface may be configured to communicate with the plurality of computational platforms over a network. In some cases, the network is a local network. In some cases, the computational platform is remote to the modular interface. For example, a user interface may be on machine local to a user, a modular interface may be implemented via a networked server which is local or remote to a user, and a computational platform may be local or remote to each of the networked server or the user. [Paragraph 82], Some examples of minimizing an objective function include the total time and distance traveled in routing applications, the total time to process jobs in scheduling applications, the number of resources used in resource planning applications, the operational cost in healthcare and service applications, etc. Some examples of maximizing an objective function include the profit of investment in finance applications, the reliability in asset management applications, and the similarity in clustering applications. [Paragraph 100], However, quantum computation resources may be rare and expensive, and may involve a certain level of expertise to be used efficiently or effectively (e.g., cost-efficiently, or cost-effectively). A number of parameters may be tuned in order for a quantum computer to deliver its potential computational power.)
Pinho also teaches ([Paragraph 36], Through their experimentation, the inventors have discovered that, in quantum simulations, the amount of resources used by a quantum algorithm is related to the number of qubits and the depth of the associated quantum circuit. By executing the same quantum algorithms several times with different resource allocation configurations, the impact of each configuration on SLA metrics such as execution time, and cost, for example, can be measured. These executions of the quantum algorithms generate large amounts of data that can be used to learn a function m=f.sub.sim(Q, D, R|QC.sub.sim), where m is an SLA metric of interest, Q is the number of qubits of the quantum circuit of the algorithm, D is the circuit depth, and R is the resource allocation configuration for a given quantum simulation instance QC.sub.sim, that is, for a single given run of the quantum algorithm. The configuration may translate into, but is not limited to, a grouping of specific numbers and types of resources such as, for example, the number of CPU cores, the amount of memory, the amount of disk allocated to the process, container, or virtual machine, that runs the quantum algorithm. FIG. 4 shows example graphs 402 and 404 of data collected from the execution of a quantum algorithm on a simulation environment. More specifically, FIG. 4 discloses collected data from quantum algorithm executions on a simulation environment. The graphs 402 and 404 respectively show, for a given {qubits, depth} quantum circuit configuration, the SLA metric “execution time” as a function of the allocated memory, and of the number of CPU cores assigned to the simulation engine execution.)
However, Pinho and Kuttimalai do not explicitly disclose node and edge activation constraints.
Gunnels teaches node and edge activation constraints. ([Paragraph 73], FIG. 3B illustrates identifies a subnetwork component for quantum feature enhancement by determining that a component including a first classical neural node 302A and a second classical neural node 302B of neural network 300 is suitable for replacement by a quantum component such as a quantum kernel component. FIG. 3C illustrates replacement of first classical neural node 302A and second classical neural node 302B with a quantum component 304 such as a quantum kernel component to produce a hybrid classical-quantum neural network for classification of input data. In one or more embodiments, quantum component 304 is implemented by a quantum processor such as quantum processor 142 of quantum processing system 140. [Paragraph 74], With reference to FIG. 4, this figure depicts a block diagram of an example process 400 for training a hybrid classical-quantum neural network in accordance with an illustrative embodiment. Process 400 includes a training operation 402 configured to receive an input including a user-defined structure 404 for the neural network, and a dataset 406 including training data for training the neural network. In particular embodiments, the user-defined structure for the neural network includes a number of layers, a neuron population in each layer, and activation functions for the neural network. In one or more embodiments, one or more portions of training operation 402 are implemented using an application within a classical processing system such as application 105 of FIG. 1. )
It would have been obvious to a person with ordinary skill in the art, before the effective filing date of the invention, to combine the teachings of Pinho and Kuttimalai wherein input representing (i.e. nodes and edges) computational processes are received, nodes are grouped according to quantum or classical computing resources, constraints and objective function(s) is/are generated to provide an optimal solution for a workflow to achieve a best solution for the generated objective function to be executed via the workflow, into teachings of Gunnels wherein hybrid-classical-quantum neural network of nodes and edges includes an activation function (i.e. activation constraints), because this would enhance the teachings of Pinho and Kuttimalai wherein by including the activation function, it allows plurality of connected nodes and edges to be activated based on user provided constrains and to aid in training of the machine learning neural network. [Gunnels paragraph 77]
As per claim 3, rejection of claim 2 is incorporated:
Pinho teaches wherein the node and edge activation constraints, responsive to selection of a process as part of the computational workflow, ensure corollary input data elements and output data elements are also activated as part of the workflow. ([Paragraph 18], Example embodiments may leverage any available on-premises classical hardware for both classical computation and quantum simulation, and leverage any available quantum hardware for quantum computations to satisfy performance constraints and SLAs (service level agreements). Some embodiments employ an orchestration pipeline that receives a computation workflow in the form of a graph, and then alters the graph in such a way that each node of the graph contains the resource allocation configuration that is needed to execute the workflow while also satisfying SLA (service level agreement) constraints… [Paragraph 33], In the first stage of the pipeline 300, which may be carried out by a module 306, of the pipeline, EstimateWorkflowResources, each graph node (see example nodes 202 and 204 in FIG. 2) of the input workflow graph 302 is assigned all candidate resource allocation configurations, given quantum circuit parameters, and estimations of SLA metrics for the execution of the graph node with the candidate configuration. Thus, the output of the stage 306 is an interim workflow graph 302a (custom-character), which is a transformation of the input workflow graph 302 (custom-character). That is, the interim workflow graph 302a has been transformed, at the first stage, to include candidate resource allocations for each node of the workflow graph 302.
Gunnels teaches node and edge activation constraints. ([Paragraph 73], FIG. 3B illustrates identifies a subnetwork component for quantum feature enhancement by determining that a component including a first classical neural node 302A and a second classical neural node 302B of neural network 300 is suitable for replacement by a quantum component such as a quantum kernel component. FIG. 3C illustrates replacement of first classical neural node 302A and second classical neural node 302B with a quantum component 304 such as a quantum kernel component to produce a hybrid classical-quantum neural network for classification of input data. In one or more embodiments, quantum component 304 is implemented by a quantum processor such as quantum processor 142 of quantum processing system 140. [Paragraph 74], With reference to FIG. 4, this figure depicts a block diagram of an example process 400 for training a hybrid classical-quantum neural network in accordance with an illustrative embodiment. Process 400 includes a training operation 402 configured to receive an input including a user-defined structure 404 for the neural network, and a dataset 406 including training data for training the neural network. In particular embodiments, the user-defined structure for the neural network includes a number of layers, a neuron population in each layer, and activation functions for the neural network. In one or more embodiments, one or more portions of training operation 402 are implemented using an application within a classical processing system such as application 105 of FIG. 1. )
As per claim 4, rejection of claim 2 is incorporated:
Gunnels teaches wherein the node and edge relationship constraints, responsive to activation of one of a pair of nodes, ensure that an edge connecting the pair of nodes is activated. ([Paragraph 73], FIG. 3B illustrates identifies a subnetwork component for quantum feature enhancement by determining that a component including a first classical neural node 302A and a second classical neural node 302B of neural network 300 is suitable for replacement by a quantum component such as a quantum kernel component. FIG. 3C illustrates replacement of first classical neural node 302A and second classical neural node 302B with a quantum component 304 such as a quantum kernel component to produce a hybrid classical-quantum neural network for classification of input data. In one or more embodiments, quantum component 304 is implemented by a quantum processor such as quantum processor 142 of quantum processing system 140. [Paragraph 33], An Artificial Neural Network (ANN)—also referred to simply as a neural network—is a computing system made up of a number of simple, highly interconnected processing elements (nodes), which process information by their dynamic state response to external inputs. [Paragraph 74], With reference to FIG. 4, this figure depicts a block diagram of an example process 400 for training a hybrid classical-quantum neural network in accordance with an illustrative embodiment. Process 400 includes a training operation 402 configured to receive an input including a user-defined structure 404 for the neural network, and a dataset 406 including training data for training the neural network. In particular embodiments, the user-defined structure for the neural network includes a number of layers, a neuron population in each layer, and activation functions for the neural network. In one or more embodiments, one or more portions of training operation 402 are implemented using an application within a classical processing system such as application 105 of FIG. 1.)
As per claim 5, rejection of claim 2 is incorporated:
Gunnels teaches wherein the domain input constraints and domain output constraints ensure that user specification of available data elements is respected and that target end goals of the computational workflow achieve user specified target data elements. ([Paragraph 73], FIG. 3B illustrates identifies a subnetwork component for quantum feature enhancement by determining that a component including a first classical neural node 302A and a second classical neural node 302B of neural network 300 is suitable for replacement by a quantum component such as a quantum kernel component. FIG. 3C illustrates replacement of first classical neural node 302A and second classical neural node 302B with a quantum component 304 such as a quantum kernel component to produce a hybrid classical-quantum neural network for classification of input data. In one or more embodiments, quantum component 304 is implemented by a quantum processor such as quantum processor 142 of quantum processing system 140. [Paragraph 33], An Artificial Neural Network (ANN)—also referred to simply as a neural network—is a computing system made up of a number of simple, highly interconnected processing elements (nodes), which process information by their dynamic state response to external inputs. [Paragraph 74], With reference to FIG. 4, this figure depicts a block diagram of an example process 400 for training a hybrid classical-quantum neural network in accordance with an illustrative embodiment. Process 400 includes a training operation 402 configured to receive an input including a user-defined structure 404 for the neural network, and a dataset 406 including training data for training the neural network. In particular embodiments, the user-defined structure for the neural network includes a number of layers, a neuron population in each layer, and activation functions for the neural network. In one or more embodiments, one or more portions of training operation 402 are implemented using an application within a classical processing system such as application 105 of FIG. 1.)
Pinho also teaches ([Paragraph 18], In general, example embodiments of the invention may bring together both classical and quantum computing in a framework that is operable to decide the amount of resources for, and to optimize the execution of, hybrid quantum-classic computation problems. Example embodiments may leverage any available on-premises classical hardware for both classical computation and quantum simulation, and leverage any available quantum hardware for quantum computations to satisfy performance constraints and SLAs (service level agreements). Some embodiments employ an orchestration pipeline that receives a computation workflow in the form of a graph, and then alters the graph in such a way that each node of the graph contains the resource allocation configuration that is needed to execute the workflow while also satisfying SLA (service level agreement) constraints. Embodiments may leverage experimental observations about how quantum algorithms consume resources in simulation engines, and may use those experimental observations to build an estimation module that can estimate SLA metrics directly from parameters of the quantum computation.)
As per claim 6, rejection of claim 2 is incorporated:
Pinho teaches wherein the domain input constraints and domain output constraints assign corresponding data element variables at the start of optimization. ([Paragraph 26], Despite the potential benefits claimed by quantum computing researchers, this new computing paradigm still largely depends and will most likely continue to depend on classical computing infrastructure for the purposes of control and communication between classical and quantum devices. As an example of this, and with reference to the example QML algorithm 100 of FIG. 1, in state-of-the-art QML (quantum machine learning), a parameterized quantum circuit 102 is optimized via a classic optimization algorithm 104 to obtain a, locally optimal, solution for a certain problem, as illustrated in FIG. 1. This type of arrangement is known as a VQA (variational quantum algorithm).)
Kuttimalai also teaches ([Paragraph 16], In some embodiments, said one or more objective functions comprise one or more of: a linear objective function, a quadratic objective function, a non-linear objective function, a non-linear convex objective function, a non-linear non-convex objective function, a look-up table objective function, or a black-box objective function. In some embodiments, said at least zero constraints comprises one or more of: a linear equality constraint, a linear inequality constraint, a quadratic constraint, a non-linear parametric constraint, a look-up table constraint, or a black-box constraint.)
Gunnels also teaches ([Paragraph 13], Another embodiment further includes receiving one or more user defined parameters, wherein the neural network is constructed based upon the one or more user defined parameters. Thus, the embodiment provides for the capability of a user to tailor the structure of the neural network according to requirements of a particular application. [Paragraph 38], In an embodiment, feature space extension of a neural network is achieved by modifying network components characterized to be indifferent to input of different classes. In the embodiment, various measures relying upon analysis of information flow and/or sensitivity are used for characterization/identification of classical subnetwork components of the neural network whose feature space can benefit from a quantum feature space enhancement. Once a subnetwork component has been identified, the input and output of the of the subnetwork are “rewired” to a quantum kernelized feature space component implemented by a quantum processor. In the embodiment, the quantum neural network involves parameters that are learnable by the full classical-quantum neural network structure. In particular embodiments, the parameters may include variational settings that may determine the formation of the quantum feature space or other manipulations applied to data.)
As per claim 7, rejection of claim 2 is incorporated:
Kuttimalai teaches wherein the resource constraints ensure classical and quantum computing resources running on machines are within equipped resources on the machines. ([Paragraph 47], A feasible configuration may be a configuration which satisfies each of the set of constraints. In some cases, a feasible configuration may be a configuration x which satisfies problem constraints in the set C(x). [Paragraph 49], In some cases, a search space is the set of possible configurations D. In some applications, the search space may be further adjusted by directly incorporating all, at least some, or at least one of the problem constraints such that it may include feasible configurations subject to all, at least some, or at least one of the problem constraints.)
As per claim 8, rejection of claim 2 is incorporated:
Pinho teaches wherein costs comprise at least one of: monetary cost; amount of time; power expended; or error incurred. ([Paragraph 37], The case with estimating SLA metrics for a true quantum device is somewhat different that just described. Although SLA metrics in quantum devices are also affected by the number of qubits and depth of algorithm, the concept of resource allocation above does not directly apply. Other factors, such as queuing time, data loading times and required number of measurements of outputs affect execution times and costs. Still, data such as that in FIG. 4 can also be collected to model a function m=f.sub.QC(Q, D|QC.sub.dev), which measures an SLA metric as a function of the execution of a quantum circuit on a given quantum device QC.sub.dev.)
Kuttimalai also teaches ([Paragraph 82], Some examples of minimizing an objective function include the total time and distance traveled in routing applications, the total time to process jobs in scheduling applications, the number of resources used in resource planning applications, the operational cost in healthcare and service applications, etc. [Paragraph 11], In some embodiments, said at least one constraints handling procedure comprises at least one of: incorporating static and adaptive penalty terms into said one or more objective function, rejection sampling or rejection of infeasible configurations, extending the search space to include Lagrange multipliers, or embedding techniques, optionally, wherein said embedding techniques comprise one or more of constraint programming, linear programming, and mixed-integer programming in said search space to restrict said search space. In some embodiments, said one or more algorithms employs said at least one constraints handling procedure.)
As per claims 12-18, these are system claims corresponding to the method claims 2-8. Therefore, rejected based on similar rationale.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DONG U KIM whose telephone number is (571)270-1313. The examiner can normally be reached 9:00am - 5:00pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Bradley Teets can be reached at 5712723338. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/DONG U KIM/Primary Examiner, Art Unit 2197
Prosecution Insights