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 .
This action is in response to the amendment filed 3/2/2026. Claims 1, 3, 10, 14, 16-28 have been amended. Claims 2, 15 have been canceled. Claims 1, 3-14, 16-28 are pending and have been considered below.
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, 2, 3, 6, 7, 8, 9, 14, 15, 16,17, 18, 19, 20, 21, 22, 23, 28 is/are rejected under 35 U.S.C. 103 as being unpatentable over Johnson et al. (US 2020/0057957 A1) in view of Herbster et al.(US 2020/0005154 A1) and further in view of Sun et al. (US 2023/0023812 A1).
Claim 1. Johnson discloses a method, performed on a hybrid quantum-classical computer system for computing initializing parameters for a parametrized quantum circuit (PQC), the hybrid quantum-classical computer system comprising a classical computer and a quantum computer,
the classical computer including a processor, a non-transitory computer-readable medium, and computer instructions stored in the non-transitory computer-readable medium, a classical computer (P 0005);
the quantum computer including a quantum component, a quantum computer (P 0140),
having a plurality of qubits, the quantum computer includes a plurality of qubits (P 0140),
which receives a sequence of instructions to evolve a quantum state based on a sequence of quantum gates, one or more circuit parameters are updated such that a parametrized quantum circuit outputs an updated quantum state that better approximates a ground state of [problem] Hamiltonian (P 0010) the present invention extends the capabilities of quantum computers for solving problems (P 0041) a plurality of linear combinations of the individual output values are utilized to estimate the optimal value being targeted to exploit algebraic structure in the problem instance (e.g., the quantum Hamiltonian) to process this data and to output a lower estimate of the ground state energy (P 0043) utilize quantum states that approximate the ground state of the Hamiltonian of a target quantum state (P 0046) a quantum circuit on the quantum computer includes any number of gates (P 0141) a quantum gate is applied to each qubit (P 0161) the classical computer provides control signals to the quantum computer which applies gate operations specified in the signals to the qubits (P 0166) The quantum circuit receives and is used to solve a quantum problem represented by a Hamiltonian represented by a linear combinations [sequence] of the individual output values utilized to estimate the optimal value being targeted;
wherein the computer instructions, when executed by the processor, perform the method, comprising: defining a PQC problem having a parametrized circuit ansatz and an objective, a computational problem to be solved and an initial Hamiltonian are proved to the quantum computer to prepare a well-known initial state (P 0135) for an initial or ansatz state (P 0155),
based on a system size comprising N qubits and a quantum circuit with K parameters, the problem Hamiltonian is defined and mapped to a sum of Pauli product terms on N qubits (P 0051);
… mapping the PQC problem to a set of initial parameters … containing information about the parameter k, the quantum circuit, and the objective C, an additional classical optimization routine is used to suggest new state preparation parameters based on the energy expectation value estimate (Note: depending on the classical optimization routine, multiple loss function evaluations may be executed before new circuit parameters are suggested.) (P 0069) the parameterized quantum circuit is programmed via one or more circuit parameters (P 0114);
using the … initial parameters as a starting point, … minimize a …loss function having a loss as output, a classical optimization routine is used to suggest new state preparation parameters based on the energy expectation value estimate (Note: depending on the classical optimization routine, multiple loss function evaluations may be executed before new circuit parameters are suggested.) (P 0056) an additional classical optimization routine is used to suggest new state preparation parameters (P 0068); and
backpropagating the loss …, multiple loss function evaluations are executed before new circuit parameters are suggested (P 0068) if the circuit parameters are updated by an amount below a threshold, then the circuit parameters have converged, or else the circuit parameters are updated by an amount that is above the threshold, then the circuit parameters have not converged, the process repeats to obtain a better approximation of the ground state and the corresponding ground-state energy, the updated quantum state is generated and the process is repeated until it is determined that the circuit parameters have converged (P 0115) to produce a solution to the original computational problem (P 0136).
Johnson does not disclose using an encoder of an encoder-decoder, mapping the PQC problem to a set of initial parameters, wherein each parameter k of the K parameters is represented as an encoding vector, each encoding vector containing information about the parameter k, the quantum circuit, and an objective, as disclosed in the claims. While Johnson does not disclose an encoding vector, the circuit parameters in Johnson are analogous to encoding vectors. Applicant’s specification discloses “Each of the K parameters of the quantum circuit U is first represented as an encoding vector hk. This encoding contains information about the parameter itself, the overall circuit, and optionally the objective.” on Page 3, 3rd full paragraph. Johnson discloses a parametrized quantum circuit, programmable via one or more circuit parameters I used to generate the quantum state of the quantum circuit. The circuit parameters are updated so that the parametrized quantum circuit outputs an updated quantum state that better approximates the ground state of the Hamiltonian (P 0114). That is, Johnson describes circuit parameters in such a way that they satisfy the claimed encoding vector, but are not described as a vector. In the same field of invention, Herbster discloses a classical compression of the image data (e.g. an auto-encoder trained to compress the image data) may be used to compress the data to a size that can be loaded on to the quantum computer (P 0007) the training of the auto-encoder may comprise providing an auto-encoder comprising an encoder part having an input layer, and a decoder part having an output layer, and one or more hidden layers connecting the encoder part to the decoder part. A classical optimization algorithm may be used to select the new circuit parameters to minimize a cost function that quantifies a distance between the quantum state and a target state (e.g., a ground state) (P 0014) the encoder generates a feature vector (P 0015). Therefore, considering the teachings of Johnson and Herbster, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine using an encoder of an encoder-decoder, mapping the PQC problem to a set of initial parameters, wherein each parameter k of the K parameters is represented as an encoding vector, each encoding vector containing information about the parameter k, the quantum circuit, and an objective with the teachings of Johnson with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004).
Johnson does not disclose using a trainable decoding element of the encoder-decoder, decoding each encoding vector by a neural network having a plurality of weights and wherein a single value is output for each encoding vector; generating a vector of initial parameters having a dimension equal to K, each initial parameter equal to the single value that is output for each encoding vector, as disclosed in the claims. However, Johnson discloses one or more circuit parameters are updated such that a parametrized quantum circuit outputs an updated quantum state that better approximates a ground state of [problem] Hamiltonian (P 0010) the present invention extends the capabilities of quantum computers for solving problems (P 0041) a plurality of linear combinations of the individual output values are utilized to estimate the optimal value being targeted to exploit algebraic structure in the problem instance (e.g., the quantum Hamiltonian) to process this data and to output a lower estimate of the ground state energy (P 0043) utilize quantum states that approximate the ground state of the Hamiltonian of a target quantum state (P 0046) the circuit parameters (analogous to the encoding vector) are updated and if a threshold is met, converged (P 0115) and a representation of the quantum state converges (P 0117) the final state of the qubits serving as the initial state of the next iteration (P 0167) and Herbster discloses an artificial neural network is trained including (P 0012) training an encoder and decoder (P 0014) the training method starts with an initial training vector (P 0044) by adjusting the biases and weights (P 0068) to initialize a neural network (P 0107). As noted above, a quantum circuit receives and is used to solve a quantum problem Hamiltonian represented as linear combinations of the individual output values and the Hamiltonian is used to set the quantum circuit parameters. Therefore, considering the teachings of Johnson and Herbster, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine using a trainable decoding element of the encoder-decoder, decoding each encoding vector by a neural network having a plurality of weights and wherein a single value is output for each encoding vector; generating a vector of initial parameters having a dimension equal to K, each initial parameter equal to the single value that is output for each encoding vector with the teachings of Johnson and Herbster with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004).
Johnson does not disclose using the vector of initial parameters as a starting point, determining the initializing parameters by performing a gradient descent (GD) optimization for S steps to minimize a meta-loss function having a loss as output, as disclosed in the claims. However, in the same field of invention, Sun discloses a method of selecting a meta-loss function (P 0009) determining a solution for a given parameter, and “optimizing” for some minimum or maximum of a misfit function, which achieves some measurement of similarity or difference between the elements (vectors) (P 0033) which can fall into a local minimum for gradient-based optimization methods (P 0036) the function taking a input vector (P 0038) such that the minimum value is reduced (P 0046) the meta-loss function is derived by gradient-descent (P 0048) to minimize the difference between the observed and simulated data (P 0049) the optimization includes regularization applied to the model (i.e., applying a total variation minimization of the model) using a trained neural network to take in the model and output its regularization measure given by a scalar as part of the optimization using meta learning for data fitting or model fitting for meta-loss (P 0077). Therefore, considering the teachings of Johnson, Herbster and Sun, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine using the vector of initial parameters as a starting point, determining the initializing parameters by performing a gradient descent (GD) optimization for S steps to minimize a meta-loss function having a loss as output with the teachings of Johnson and Herbster with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004).
Johnson does not disclose backpropagating the loss to update the plurality of weights, as disclosed in the claims. However, Herbster discloses the training method starts with an initial training vector (P 0044) by adjusting the biases and weights (P 0068) quantum pre-training is arranged to calculate an appropriate loss of function which measures a difference between the updated feature vector generated by the RBM and the original feature vector input to the input layer of the RBM by the encoder, and is arranged to use the value of that loss function in adjusting the biases and weights applied to the nodes of the RBM in such a way as to minimize the value of the loss function thereby optimizing the accuracy of the updated feature vector (P 0107). Therefore, considering the teachings of Johnson, Herbster and Sun, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine backpropagating the loss to update the plurality of weights with the teachings of Johnson, Herbster and Sun with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004).
Claim 2. Canceled.
Claim 3. Johnson, Herbster and Sun disclose the method of claim 1, and Johnson discloses a plurality of linear combinations of the individual Pauli product expectation value estimate output values are used, not merely the fixed linear combination of those values (P 0043) the problem Hamiltonian is mapped to a sum of Pauli product terms to measure the expectation value of each Pauli term (P 0048) in the variational quantum eigensolver algorithm, the updated quantum state is prepared based on the updated circuit parameters (P 0049) to suggest new state preparation parameters based on the energy expectation value estimate (Note: depending on the classical optimization routine, multiple loss function evaluations may be executed before new circuit parameters are suggested) (P 0051) the optimization procedure uses Hamiltonians decomposed into a linear combination of Pauli strings (P0080). It is clear that the circuit parameters are mapped to a set of Pauli operators, similar to the claimed encoding vector, but the circuit parameters remain fixed, only the linear combination of sets of Pauli operators are varied, not the number of circuit parameters themselves. Herbster has been combined with Johnson to put the circuit parameters in the form of a vector. Therefore, considering the teachings of Johnson, Herbster and Sun, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine wherein each encoding vector is a fixed size and uniquely represents its respective parameter k with the teachings of Johnson, Herbster and Sun with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004).
Claim 6. Johnson, Herbster and Sun disclose the method of claim 1, and Johnson discloses applying the method to quantum circuits of varying sizes, appending quantum gates to a variational quantum circuit to artificially extend the depth of the circuit (P 0040).
Claim 7. Johnson, Herbster and Sun disclose the method of claim 1, and Herbster discloses the encoder part of this network will learn the data representation through the features extracted by convolutional layers (P 0086) and Sun discloses meta-learning includes ML algorithms that try to learn from observations on how other neural networks perform and then establish a system that learns from this experience (learning to learn) (P 0032)and is flexible in solving learning problems and tries to improve the performance of existing learning algorithms or to learn (extract) the learning algorithms itself (P 0037) based on the ML-misfit function framework it is possible to learn a misfit function using a machine, which can incorporate the feature embedded in the dataset and as a result provide desired features for the misfit function (P 0068). Therefore, considering the teachings of Johnson, Herbster and Sun, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine wherein the method is applied to learning arbitrarily-sized quantum circuits with the teachings of Johnson, Herbster and Sun with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004).
Claim 8. Johnson, Herbster and Sun disclose the method of claim 1, and Johnson discloses a parametrized quantum circuit, programmable via one or more circuit parameters I used to generate the quantum state of the quantum circuit. The circuit parameters are updated so that the parametrized quantum circuit outputs an updated quantum state that better approximates the ground state of the Hamiltonian (P 0114) multiple loss function evaluations are executed before new circuit parameters are suggested (P 0068) if the circuit parameters are updated by an amount below a threshold, then the circuit parameters have converged, or else the circuit parameters are updated by an amount that is above the threshold, then the circuit parameters have not converged, the process repeats to obtain a better approximation of the ground state and the corresponding ground-state energy, the updated quantum state is generated and the process is repeated until it is determined that the circuit parameters have converged (P 0115) to produce a solution to the original computational problem (P 0136). That is, Johnson describes circuit parameters in such a way that they satisfy the claimed encoding vector, but are not described as a vector. In the same field of invention, Herbster discloses a classical compression of the image data (e.g. an auto-encoder trained to compress the image data) may be used to compress the data to a size that can be loaded on to the quantum computer (P 0007) the training of the auto-encoder may comprise providing an auto-encoder comprising an encoder part having an input layer, and a decoder part having an output layer, and one or more hidden layers connecting the encoder part to the decoder part. A classical optimization algorithm may be used to select the new circuit parameters to minimize a cost function that quantifies a distance between the quantum state and a target state (e.g., a ground state) (P 0014) the encoder generates a feature vector (P 0015). Therefore, considering the teachings of Johnson, Herbster and Sun, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine wherein encoding vectors are fully defined by the PQC problem with the teachings of Johnson, Herbster and Sun with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004).
Claim 9. Johnson, Herbster and Sun disclose the method of claim 1, and Johnson discloses whereby the method computes the initializing parameters for QAOA applied to max-cut problems, quantum approximate optimization algorithm [?] generates approximate solutions to optimization problems, in particular, focusing on the graph-theoretic problem of MAXCUT (P 0110).
Claim(s) 14, 16, 20, 21, 22, 23 is/are directed to hybrid quantum-classical computer system claim(s) similar to the method claim(s) of Claim(s) 1, 3, 6, 7, 8, 9 and is/are rejected with the same rationale.
Claim 17. Johnson, Herbster and Sun disclose the system of claim 14, and Johnson discloses wherein the optimization method accelerates initial parameters generation, the optimization procedure accelerates the optimization of the algorithm (P 0050).
Claim 18. Johnson, Herbster and Sun disclose the system of claim 14, and Sun discloses a method of selecting a meta-loss function (P 0009) determining a solution for a given parameter, and “optimizing” for some minimum or maximum of a misfit function, which achieves some measurement of similarity or difference between the elements (vectors) (P 0033) which can fall into a local minimum for gradient-based optimization methods (P 0036) the function taking a input vector (P 0038) such that the minimum value is reduced (P 0046) the meta-loss function is derived by gradient-descent (P 0048) to minimize the difference between the observed and simulated data (P 0049) the optimization includes regularization applied to the model (i.e., applying a total variation minimization of the model) using a trained neural network to take in the model and output its regularization measure given by a scalar as part of the optimization using meta learning for data fitting or model fitting for meta-loss (P 0077). Therefore, considering the teachings of Johnson, Herbster and Sun, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine The system of claim 1, wherein learning a set of initial parameters is efficiently refined by gradient descent GD with the teachings of Johnson and Herbster with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004).
Claim 19. Johnson, Herbster and Sun disclose the system of claim 14, and Sun discloses meta-learning is used including machine learning algorithms (P 0032). Therefore, considering the teachings of Johnson, Herbster and Sun, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine wherein the method uses meta-learning with the teachings of Johnson, Herbster and Sun with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004).
Claim 28 is directed to a method for solving a parameterized quantum circuit (PQC) problem and includes all the limitations in Claim 1 rejected in view of the same prior art with the same rationale.
Claim(s) 4, 5 is/are rejected under 35 U.S.C. 103 as being unpatentable over Johnson et al. (US 2020/0057957 A1) in view of Herbster et al.(US 2020/0005154 A1) and Sun et al. (US 2023/0023812 A1) and further in view of Turbin et al. (US 2011/0276526 A1).
Claim 4. Johnson, Herbster and Sun disclose the method of claim 1, but Johnson does not disclose wherein the generation of the vector of initial parameters is accelerated relative to random generation of initial parameters, as disclosed in the claims. Johnson discloses the optimization procedure accelerates the optimization of the algorithm (P 0050) the expectation value may be calculated deterministically or the expectation value may be determined stochastically (e.g., to simulate the randomness inherent to measurements performed on the quantum computer) (P 0116) the outcome of each individual measurement is random, but they are related in such a way that the computation always succeeds (P 0138). The claim language does not clearly explain “relative to random generation of initial parameters”. Johnson discloses that the optimization procedure accelerates the optimization of the algorithm, and Johnson further discloses that algorithm is performed in such a way that although the outcome of each measurement is relative, the computation is always successful, meaning that the circuit parameters must be chosen to affect this outcome. Sun discloses a meta-loss function is defined with randomly generated data (P 0052) and a first model is generated from a second model by randomly changing one or more parameters of the first model (P 0057). Herbster has been combined with Johnson to put the circuit parameters in the form of a vector. In the same field of invention, Turbin discloses the number of the training iterations is reduced dramatically if an initial set of weights and biases are reused from a known detector comparing to the number of iterations used for a random starting point in a weights space where the parameters are randomly initialized (P 0192). Therefore, considering the teachings of Johnson, Herbster, Sun and Turbin, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine wherein the generation of the vector of initial parameters is accelerated relative to random generation of initial parameters with the teachings of Johnson, Herbster and Sun with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004) and to dramatically reduce iterations for training (Turbin: P 0192).
Claim 5. Johnson, Herbster and Sun disclose the method of claim 1, but Johnson does not disclose wherein the initialization method uses machine learning to provide a flexible initializer for arbitrarily-sized parametrized quantum circuits, as disclosed in the claim. Sun discloses a misfit function is determined using machine learning that adapts better to data (P 0024). In the same field of invention, Turbin discloses the calculation of a single layer of "n" neurons and "m" in the previous layer can be written in a matrix form, where W is a weight matrix of the size n x m elements and consists of n rows, X is an m-dimensional vector of the output values from the previous layer and b is an n-dimensional vector of the layer bias, y is the n-dimensional output vector of the layer, the function f is meant to be calculated for each individual element of the vector (P 0208) the input vector X is the response of the detector, for multi-electrode anode type position sensitive detectors, the response is a vector having a number of elements equal to the number of electrodes of the anode, input vector elements may be raw detected values of measured from the electrodes or the result of heuristic based computation (P 0210) in delay-line detectors, the response is a vector having a number of elements equal to the number of delay lines (P 0211). Therefore, considering the teachings of Johnson, Herbster, Sun and Turbin, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine wherein the initialization method uses machine learning to provide a flexible initializer for arbitrarily-sized parametrized quantum circuits with the teachings of Johnson, Herbster and Sun with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004) and to dramatically reduce iterations for training (Turbin: P 0192).
Claim(s) 10, 11, 13, 24, 25, 27 is/are rejected under 35 U.S.C. 103 as being unpatentable over Johnson et al. (US 2020/0057957 A1) in view of Herbster et al.(US 2020/0005154 A1) and Sun et al. (US 2023/0023812 A1) and further in view of Dallaire-Demers et al. (US 2020/0394549 A1).
Claim 10. Johnson, Herbster and Sun disclose the method of claim 1, but Johnson does not disclose whereby the method computes the initializing parameters for optimizing a 1D Fermi-Hubbard model (1D) FHM, as disclosed in the claims. However, in the same field of invention, Dallaire-Demers discloses using a one-dimensional Fermi-Hubbard model (P 0113). Therefore, considering the teachings of Johnson, Herbster, Sun and Dallaire-Demers, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine whereby the method computes the initializing parameters for optimizing the 1D Fermi-Hubbard model (1D) FHM with the teachings of Johnson, Herbster and Sun with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004) and to dramatically reduce iterations for training (Turbin: P 0192) and a one-dimensional Fermi-Hubbard model can be solved exactly (Dallaire-Demers: P 0113).
Claim 11. Johnson, Herbster and Sun disclose the method of claim 1, but Johnson does not disclose wherein performing the GD optimization comprises minimizing local minima, as disclosed in the claims. However, Sun discloses a method of selecting a meta-loss function (P 0009) determining a solution for a given parameter, and “optimizing” for some minimum or maximum of a misfit function, which achieves some measurement of similarity or difference between the elements (vectors) (P 0033) which can fall into a local minimum for gradient-based optimization methods (P 0036) the function taking a input vector (P 0038) such that the minimum value is reduced (P 0046). In the same field of invention, Dallaire-Demers discloses free energy is minimized to eliminate the imaginary time evolution getting stuck in a local minimum (P 0113). Therefore, considering the teachings of Johnson, Herbster, Sun and Dallaire-Demers, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine wherein performing the GD optimization comprises minimizing local minima with the teachings of Johnson, Herbster and Sun with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004) and to dramatically reduce iterations for training (Turbin: P 0192).
Claim 13. Johnson, Herbster and Sun disclose the method of claim 1, but Johnson does not disclose wherein the parametrized circuit ansatz comprises a Low Depth Circuit Ansatz (LDCA), as disclosed in the claims. Sun discloses a method of selecting a meta-loss function (P 0009) determining a solution for a given parameter, and “optimizing” for some minimum or maximum of a misfit function, which achieves some measurement of similarity or difference between the elements (vectors) (P 0033) which can fall into a local minimum for gradient-based optimization methods (P 0036) the function taking a input vector (P 0038) such that the minimum value is reduced (P 0046). However, in the same field of invention, Dallaire-Demers discloses a low-depth circuit ansatz (LDCA), consisting of matchgates circuit plus additional nearest-neighbor phase coupling is described (P 0039). Therefore, considering the teachings of Johnson, Herbster, Sun and Dallaire-Demers, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine wherein the parametrized circuit ansatz comprises a Low Depth Circuit Ansatz (LDCA) with the teachings of Johnson, Herbster and Sun with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004) and to dramatically reduce iterations for training (Turbin: P 0192) and a one-dimensional Fermi-Hubbard model can be solved exactly (Dallaire-Demers: P 0113).
Claim(s) 24, 25, 27 is/are directed to hybrid quantum-classical computer system claim(s) similar to the method claim(s) of Claim(s) 10, 11, 13 and is/are rejected with the same rationale.
Claim(s) 12, 26 is/are rejected under 35 U.S.C. 103 as being unpatentable over Johnson et al. (US 2020/0057957 A1) in view of Herbster et al.(US 2020/0005154 A1) and Sun et al. (US 2023/0023812 A1) and further in view of Bocharov et al. (US 2021/0256416 A1).
Claim 12. Johnson, Herbster and Sun disclose the method of claim 1, but Johnson does not disclose wherein performing the GD optimization comprises minimizing barren plateaus, as disclosed in the claims. However, in the same field of invention, Bocharov discloses the invention presents advantages over the other techniques (P 0096) to mitigate barren plateaus (P 0098). Therefore, considering the teachings of Johnson, Herbster, Sun and Bocharov, one having ordinary skill in the art before the effective filing date of the invention would have been motivated to combine wherein the parametrized circuit ansatz comprises a Low Depth Circuit Ansatz (LDCA) with the teachings of Johnson, Herbster and Sun with the motivation to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004) and to dramatically reduce iterations for training (Turbin: P 0192).
Claim(s) 26 is/are directed to hybrid quantum-classical computer system claim(s) similar to the method claim(s) of Claim(s) 12 and is/are rejected with the same rationale.
Response to Arguments
Applicant's arguments filed 3/2/2026 have been fully considered but they are not persuasive.
The applicant argues:
At no point does Johnson disclose or suggest a parameter is anything other than a value applied to a gate.
But unlike Johnson's scalar value, which is determined from an iterative optimization routine that updates the circuit parameters based on state measurements, the scalar value in the claimed invention is determined from a neural network's decoding of descriptive metadata representing the parameter, the circuit, and the objective. This critical distinction enables implementations of the claimed invention to incorporate the context of the circuit and the objective of the circuit's output into the parameter generation process.
The cited paragraph [0114] of Johnson merely explains that "a classical optimization algorithm may be used to select the new circuit parameters to minimize a cost function," and suggests nothing about a higher-dimensional encoding-in any form whether vector or otherwise-of information about the quantum circuit and an objective. Indeed, paragraph [0114] states that the "parametrized quantum circuit [is] programmable via one or more circuit parameters," further demonstrating that Johnson's circuit parameters are the actual values used to tune a gate and nothing more. See Johnson, [0130] ("tuning parameters may correspond to the angles of individual optical elements.").
The explanation in paragraph [0114] that "a parametrized quantum circuit, programmable via one or more circuit parameters, to generate the quantum state [wherein] the circuit parameters are updated so that the parametrized quantum circuit outputs an updated quantum state that better approximates the ground state of the Hamiltonian" suggests nothing about the circuit parameters encoding information about the quantum circuit and an objective as required by claim 1. To support a prima facie case of obviousness, the Examiner must provide an articulated reasoning with some rational underpinnings as to how Johnson's circuit parameters are analogous to the encoding vectors of claim 1. See KSR Int'l Co. V. Teleflex Inc., 550 U.S. 398, 418 (2007). No such reasoning has been articulated.
The examiner respectfully disagrees. John discloses one or more circuit parameters are updated such that a parametrized quantum circuit outputs an updated quantum state that better approximates a ground state of [problem] Hamiltonian (P 0010) the present invention extends the capabilities of quantum computers for solving problems (P 0041) a plurality of linear combinations of the individual output values are utilized to estimate the optimal value being targeted to exploit algebraic structure in the problem instance (e.g., the quantum Hamiltonian) to process this data and to output a lower estimate of the ground state energy (P 0043) utilize quantum states that approximate the ground state of the Hamiltonian of a target quantum state (P 0046) a quantum circuit on the quantum computer includes any number of gates (P 0141) a quantum gate is applied to each qubit (P 0161) the classical computer provides control signals to the quantum computer which applies gate operations specified in the signals to the qubits (P 0166) The quantum circuit receives and is used to solve a quantum problem represented by a Hamiltonian represented by a linear combinations [sequence] of the individual output values utilized to estimate the optimal value being targeted. Herbster discloses an artificial neural network is trained including (P 0012) training an encoder and decoder (P 0014) the training method starts with an initial training vector (P 0044) by adjusting the biases and weights (P 0068) to initialize a neural network (P 0107). As noted above, a quantum circuit receives and is used to solve a quantum problem Hamiltonian represented as linear combinations of the individual output values and the Hamiltonian is used to set the quantum circuit parameters. While Johnson does not explicitly disclose an encoding vector, Johnson describes circuit parameters in such a way that they satisfy the claimed encoding vector. Herbster was combined with Johnson for the explicit disclosure that an encoder generates feature, which, when applied to Johnson, allows the linear combination of output values of the problem Hamiltonian in Johnson, which are applied to the circuit parameters for solving the problem, to be represented as an encoding vector.
The applicant argues:
The Office Action relies on Herbster to disclose the decoding element of Step 1C that is missing from Johnson, which recites "using a trainable decoding element of the encoder-decoder, decoding each encoding vector by a neural network wherein a single value is output for each encoding vector." While Herbster discloses a decoder, it the decoder does not output a single value for a respective input, such as the encoding vector in claim 1.
The examiner respectfully disagrees. As noted above the problem Hamiltonian represented by a linear combination of the individual output values are input to a quantum circuit comprising circuit parameters so as to solve problem represented by the Hamiltonian. Furthermore, the claim discloses each parameter k of the K parameters is represented as an encoding vector, each encoding vector containing information about the parameter k, the quantum circuit, and an objective, and each deciding vector is decoded by a neural network having a plurality of weights and a single value is output for each decoding vector, and the a vector of initial parameters is generated having a dimension equal to K, each initial parameter equal to the single value that is output for each encoding vector. That is, it is each encoding vector contains information about the parameter k, the quantum circuit, and an objective. Therefore it is unclear whether the single value of the encoding vector represents the magnitude of a vector with dimensions representing parameter k, the quantum circuit and an objective, or the single value applies to each of single values for parameter k, the quantum circuit and an objective. Furthermore, Johnson discloses the circuit parameters are updated and if a threshold is me, the circuit parameters are converged. This disclosure of Johnson appears to disclose the embodiment where the single value represents a single value of the encoding vector, i.e. the magnitude of the vector with dimensions parameter k, the quantum circuit and an objective.
The applicant argues:
The Office Action improperly relies on Sun to disclose the meta-loss optimization of Step 1E of claim 1. Sun is non-analogous art that a person of ordinary skill in the art (POSITA) of quantum computing would not have considered. Further, the Examiner has failed to provide a rational motivation to combine it with the other cited references.
In response to applicant's argument that Sun is nonanalogous art, it has been held that a prior art reference must either be in the field of the inventor’s endeavor or, if not, then be reasonably pertinent to the particular problem with which the inventor was concerned, in order to be relied upon as a basis for rejection of the claimed invention. See In re Oetiker, 977 F.2d 1443, 24 USPQ2d 1443 (Fed. Cir. 1992). In this case, Sun was not used to reject any limitations directed to quantum states or quantum systems. The claims are directed to decoding each encoding vector by a neural network. Nothing in the claims limits the use of the claimed neural network strictly to quantum systems. Neural networks operate on data input to the network and the parameters disclosed in the claims are not of such nature that could not be processed by a neural network. That is nothing in the claims or Applicant’s specification describes a neural network that can only process quantum systems.
In response to applicant’s argument that there is no teaching, suggestion, or motivation to combine the references, the examiner recognizes that obviousness may be established by combining or modifying the teachings of the prior art to produce the claimed invention where there is some teaching, suggestion, or motivation to do so found either in the references themselves or in the knowledge generally available to one of ordinary skill in the art. See In re Fine, 837 F.2d 1071, 5 USPQ2d 1596 (Fed. Cir. 1988), In re Jones, 958 F.2d 347, 21 USPQ2d 1941 (Fed. Cir. 1992), and KSR International Co. v. Teleflex, Inc., 550 U.S. 398, 82 USPQ2d 1385 (2007). In this case, the examiner cited th motivation to combine Sun with Johnson and Herbster directly from Herbster, i.e. to provide a more robust system for processing large, multidimensional data requiring high computational load (Herbster: P 0002-0004). Neural networks are designed specifically for more efficient and accurate processing of complex data. Clearly, one would have been motivated to use a neural network to process the complex data disclosed in both Johnson and Herbster.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication should be directed to JOHN M HEFFINGTON at telephone number (571)270-1696.
Examiner interviews are available via a variety of formats. See MPEP § 713.01. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JOHN M HEFFINGTON whose telephone number is (571)270-1696. The examiner can normally be reached on Monday through Friday from 9:30 am to 5:30 pm Eastern.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Cesar B Paula, can be reached at telephone number 571-272-4128. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from Patent Center. Status information for published applications may be obtained from Patent Center. Status information for unpublished applications is available through Patent Center to authorized users only. Should you have questions about access to the USPTO patent electronic filing system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
Examiner interviews are available via a variety of formats. See MPEP § 713.01. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) Form at https://www.uspto.gov/InterviewPractice.
/J.M.H/Examiner, Art Unit 2145 5/27/2026
Prosecution Insights