DETAILED ACTION
This communication is in response to Application No. 18/440,629 filed on February 13th, 2024
in which claims 1-19 are presented for examination.
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 .
Information Disclosure Statement
The information disclosure statement submitted on 06/05/2025 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement was considered by the examiner.
Specification
The contents of the specification are sufficient for examination purposes.
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.
Claims 1-19 are rejected under 35 U.S.C. 112(b) as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor regards as the invention.
Regarding Claim 1, the recites “wherein the initial block is executed on a first state and a block of the one or more additional blocks is executed in a second state wherein the initial block is executed on the first state and block of the one or more additional blocks is executed on a second state” (Claim 1, ln. 7-10), which renders the claim indefinite. Specifically, “wherein the initial block is executed on a first state” appears to be redundantly repeated by “wherein the initial block is executed on the first state”, apart from the notable change from “a first state” to “the first state”. Additionally, “and a block of the one or more additional blocks is executed in a second state” appears to be redundantly repeated by “and block of the one or more additional blocks is executed on a second state”, apart from the notable change from “and a block” to “and block”. Therefore, the scope of the claim is indefinite because it is not claim whether these repeated recitations should be interpreted as a drafting error or as additional limitations that further restrict the scope of the claim. For example, the change from “a first state” to “the first state” could be interpreted as requiring a second execution of “the initial block” on the “first state”, and the change from “and a block” to “and block” could be interpreted as specify a specific block from the “one or more additional blocks”. Whereas the repeated recitations could alternatively be interpreted as a drafting error where the change from “the initial block” on the “first state” was meant to preserve proper antecedent basis and the omission of the “a” in “and block” was further done in error. As a result, the claim is rejected. The claim should be amended to resolve this indefiniteness, such as by removing the redundant recitations of claimed limitations.
Additionally, the claim recites “the initial state” (Claim 1, ln. 14). There is insufficient antecedent basis for this limitation in the claim. Therefore, the scope of the claim is indefinite because it is not clear what “initial state” that “the initial state” is referencing. As a result, the claim is rejected. The claim should be amended to resolve this indefiniteness.
Regarding Claims 2-6, the claims are rejected because they are dependent on an indefinite claim.
Regarding Claim 7, the claim recites the term “predetermined” (Claim 7, ln. 3), which is a relative term that renders the claim indefinite. The term “predetermined” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. Therefore, the scope of the claim is indefinite because it is not clear whether a given “threshold” should be considered a “predetermined threshold” (Claim 7, ln. 3). As a result, the claim is rejected. The claim should be amended to resolve this indefiniteness. Additionally, the claim is rejected because it is dependent on an indefinite claim.
Regarding Claim 8, the claim recites the term “gradually” (Claim 8, ln. 2), which is a relative term that renders the claim indefinite. The term “gradually” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. Therefore, the scope of the claim is indefinite because it is not clear whether a given “increasing [of] the size of each block” should be considered as “gradually” (Claim 8, ln. 2). As a result, the claim is rejected. The claim should be amended to resolve this indefiniteness. Additionally, the claim is rejected because it is dependent on an indefinite claim.
Regarding Claims 9-11, the claims are rejected because they are dependent on an indefinite claim.
Regarding Claim 12, the claim recites “wherein the initial block is executed on a first state and a block of the one or more additional blocks is executed in a second state wherein the initial block is executed on the first state and block of the one or more additional blocks is executed on a second state” (Claim 12, ln. 9-12) and “the initial state” (Claim 12, ln. 16), which are indefinite for substantially the same reasoning as articulated in regard to the rejection of Claim 1 above. As such, the claim is similarly rejected and should be amended in a similar manner.
Regarding Claim 13-15, the claims are rejected because they are dependent on an indefinite claim.
Regarding Claim 16, the claim recites “a predetermined threshold” (Claim 16, ln. 3), which is indefinite for substantially the same reasoning as articulated in regard to the rejection of Claim 7 above. As such, the claim is similarly rejected and should be amended in a similar manner. Additionally, the claim is rejected because it is dependent on an indefinite claim.
Regarding Claim 17, the claim recites “increasing the size of each block gradually” (Claim 17, ln. 2), which is indefinite for substantially the same reasoning as articulated in regard to the rejection of Claim 8 above. As such, the claim is similarly rejected and should be amended in a similar manner. Additionally, the claim is rejected because it is dependent on an indefinite claim.
Regarding Claims 18-19, the claims are rejected because they are dependent on an indefinite 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.
Claims 1-5, 9, 12-15, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Cervera-Lierta et al. (hereinafter Cervera-Lierta) (“Meta-Variational Quantum Eigensolver: Learning Energy Profiles of Parameterized Hamiltonians for Quantum Simulation”) in view of Leyton-Ortega et al. (hereinafter Leyton-Ortega) (“Robust implementation of generative modeling with parametrized quantum circuits”).
Regarding Claim 1, Cervera-Lierta teaches a method comprising (Pg. 2, Col. 2, Para. 3-4, “we introduce the meta-VQE algorithm . . . we apply this algorithm to a state-of-the-art application: the simulation of transmon qubits in a quantum computer . . . The structure of the meta-VQE algorithm is shown
diagrammatically in Fig. 1”, where “appl[ication]” of “the meta-VQE algorithm” is the method; see also Pg. 3, Col. 1, Fig. 1, “Schematic illustration of the meta-VQE algorithm”):
receiving an input quantum circuit (Pg. 1, Abstract, “The meta-VQE algorithm introduces both a gain in efficiency for parameterized Hamiltonians in terms of the number of optimizations and a good starting point for the quantum circuit parameters for individual optimizations” and Pg. 3, Col. 1, Para. 1, “The circuit is initialized in the |0⟩⊗n state, or in the Hartree-Fock state”, where an input quantum circuit, “the quantum circuit”, is received through initialization, “The circuit is initialized”);
dividing the input quantum circuit into a set of blocks, wherein the set of blocks includes an initial block and one or more additional blocks, wherein the one or more additional blocks are consecutive and contiguous to the initial block, wherein the set of blocks comprises a series of gates (Pg. 3, Col. 1, Para. 1, “The circuit is initialized . . . The first part of the circuit is the encoding unitary S containing parameterized gates, the arguments of which include the Hamiltonian parameters of one of the training points ⃗ λi and the variational parameters ⃗. The second unitary of the circuit, U, also contains parameterized gates that depend on variables ⃗, but not the Hamiltonian parameters”; Pg. 9, Col. 1, Para. 3, “the meta-VQE circuit is divided into two parts, the encoding and processing unitaries”; and Pg. 3, Col. 1, Fig. 1, where the input quantum circuit, “The circuit”, is divided into a set of blocks, “is divided into two parts”, wherein the set of blocks includes an initial block, “the encoding unitary”, and one or more additional blocks, “processing unitar[y]”, wherein, as depicted in Fig. 1, the additional block, “Processing”, is consecutive and contiguous to the initial block, “encoding”, because they are next to each other without interruption and share a common boarder, and wherein the set of blocks comprises a series of gates, “containing parameterized gates . . . contains parameterized gates”; see generally Pg. 3, Col. 2, Para. 5, “we construct the quantum circuit with minimal blocks, called layers, that are concatenated and progressively introduce more variational parameters and generate entanglement” and Pg. 5, Fig. 3, “Circuit ansatz corresponding to four qubits and two encoding and processing layers”, where the “layers” themselves could be considered the set of “blocks”, such that the second layer of the “two encoding” layers is the initial block and the first layer of the “processing layers” is the additional block);
executing . . . the set of blocks, wherein the initial block is executed on a first state and a block of the one or more additional blocks is executed in a second state wherein the initial block is executed on the first state and block of the one or more additional blocks is executed on a second state, wherein the . . . [method] executes each of the additional blocks in consecutive states until an end of the input quantum circuit (Pg. 2, Col. 2, Para. 2, “As these algorithms propose, we design a parameterized quantum circuit to be trained with a set of values from a given model, in our case, a physical Hamiltonian. We treat the encoding part of the meta-VQE as a quantum neural network that learns the encoding of the Hamiltonian. The processing part guides the encoded state towards the ground state. We observe two advantages of this algorithm: (i) it can be used to first explore the ground state energies of the Hamiltonian parameter space with only a few training points and then use the result as an initial state for a precise VQE, and (ii) the encoding in a VQA proves valuable and helps these algorithms to find the ground state”; Pg. 3, Col. 1, Para. 1, “The first part of the circuit is the encoding unitary S containing parameterized gates, the arguments of which include the Hamiltonian parameters of one of the training points ⃗ λi and the variational parameters ⃗. The second unitary of the circuit, U, also contains parameterized gates that depend on variables ⃗, but not the Hamiltonian parameters. The final state of the circuit can be written as
PNG
media_image1.png
48
472
media_image1.png
Greyscale
”; and Pg. 3, Col. 1, Fig. 1, where the set of blocks, “The first part of the circuit is the encoding unitary . . . The second unitary of the circuit, U”, are executed to generate “the ground state”, wherein the initial block, “the encoding unitary S”, which is depicted in Fig 1. as the “Encoding” unitary, is executed on a first state, as shown in Fig. 1, and a block of the one or more additional blocks, “The second unitary of the circuit, U”, which is depicted in Fig 1. as the “Processing” unitary, is executed in a second state, “the encoded state”, wherein, as depicted in Fig. 1 and the above reproduced equation, the method executes each of the additional blocks in consecutive states until an end of the input quantum circuit, “The processing part guides the encoded state towards the ground state”);
training . . . a parametrized quantum circuit such that the parametrized quantum circuit is operable to reproduce the second state (Pg. 2, Col. 2, Para. 2, “we design a parameterized quantum circuit to be trained with a set of values from a given model, in our case, a physical Hamiltonian. We treat the encoding part of the meta-VQE as a quantum neural network that learns the encoding of the Hamiltonian. The processing part guides the encoded state towards the ground state”; Pg. 3, Col. 2, Para. 2, “Once the meta-VQE circuit is trained, we can proceed to test its performance”; and Pg. 3, Col. 1, Fig. 1, where a parametrized quantum circuit is trained, “a parameterized quantum circuit to be trained”, such that, during “test[ing of] its performance”, the parametrized quantum circuit is operable to reproduce the second state, “the encoded state”),
wherein the second state is a quantum state that encodes results of all the quantum operations from the initial state until the operations of the block of the one or more additional block at the second state (Pg. 2, Col. 2, Para. 2, “As these algorithms propose, we design a parameterized quantum circuit to be trained with a set of values from a given model, in our case, a physical Hamiltonian. We treat the encoding part of the meta-VQE as a quantum neural network that learns the encoding of the Hamiltonian. The processing part guides the encoded state towards the ground state. We observe two advantages of this algorithm: (i) it can be used to first explore the ground state energies of the Hamiltonian parameter space with only a few training points and then use the result as an initial state for a precise VQE, and (ii) the encoding in a VQA proves valuable and helps these algorithms to find the ground state”; Pg. 3, Col. 1, Para. 1, “The first part of the circuit is the encoding unitary S containing parameterized gates, the arguments of which include the Hamiltonian parameters of one of the training points ⃗ λi and the variational parameters ⃗. The second unitary of the circuit, U, also contains parameterized gates that depend on variables ⃗, but not the Hamiltonian parameters. The final state of the circuit can be written as
PNG
media_image1.png
48
472
media_image1.png
Greyscale
”; and Pg. 3, Col. 1, Fig. 1, wherein the second state, “the encoded state”, is a quantum state, see Pg. 1, Col. 2, Para. 1, “The second block guides the algorithm around a particular region of the space of quantum states”, that encodes results of all the quantum operations from the initial state, “We treat the encoding part of the meta-VQE as a quantum neural network that learns the encoding of the Hamiltonian”, until the operations of the block of the one or more additional block at the second state, “The processing part guides the encoded state towards the ground state”); and
outputting . . . an expected output of the input quantum circuit (Pg. 3, Col. 2, Para. 2, “We compute the expected value of the Hamiltonian with other values of ⃗ λ by running the meta VQE circuit with the trained parameters ⃗opt and ⃗opt and the corresponding ⃗ λ values for the test point” and Pg. 10, Col. 1, Para. 2, “With a meta-VQE, one can scan over the same set of parameters with a single minimization, obtain the energy profile, and identify the areas of interest, e.g., those with minimum (for equilibrium structures) or maximum (for transition state searches) energy”, where an expected output, “compute the expected value of the Hamiltonian”, of the input quantum circuit, “by running the meta VQE circuit” is output to “obtain the energy profile, and identify the areas of interest, e.g., those with minimum (for equilibrium structures) or maximum (for transition state searches) energy”; see also Pg. 3, Col. 1, Fig. 1, “By computing the expected value of the Hamiltonian for multiple values of ⃗ λ, we design a cost function Loss to be minimized”, where an expected output of the input quantum circuit, “the expected value of the Hamiltonian for multiple values”, is output to “design a cost function Loss to be minimized”).
Cervera-Lierta does not explicitly disclose . . . by the Quantum Processing Unit . . . (where the resources used to operationalize the executing, training, and outputting are not specifically described as comprising a Quantum Processing Unit, see Pg. 2, Col. 2, Para. 3, “we apply this algorithm to a state-of-the-art application: the simulation of transmon qubits in a quantum computer”; redundant recitations of Quantum Processing Unit omitted).
Alternatively, it could be argued that the processing units needed to execute the method (see Pg. 2, Col. 2, Para. 3-4, “we introduce the meta-VQE algorithm . . . we apply this algorithm to a state-of-the-art application: the simulation of transmon qubits in a quantum computer . . . The structure of the meta-VQE algorithm is shown diagrammatically in Fig. 1”; see also Pg. 3, Col. 1, Fig. 1, “Schematic illustration of the meta-VQE algorithm”) are within the broadest reasonable interpretation of quantum processing units because they simulate qubits in a quantum computer and represent quantum objectives (see Pg. 2, Col. 2, Para. 3, “we apply this algorithm to a state-of-the-art application: the simulation of transmon qubits in a quantum computer”; see also Pg. 2, Col. 2, Para. 3, “Results are obtained via implementation of the meta-VQE protocol using TEQUILA [30] and choosing QULACS [31] as a quantum backend. TEQUILA represents quantum objectives as generalized functions of abstract expectation values, allowing arbitrary transformations and arithmetic operations on those data types to be done in a blackboard fashion”).
Regardless, Leyton-Ortega teaches . . . [executing, training, and outputting] by the Quantum Processing Unit . . . (Pg. 6, Col. 2, Para. 1, “In this work, we implement a hybrid quantum-classical algorithm for unsupervised machine learning tasks introduced in (Benedetti et al. 2019) at Rigetti’s superconducting quantum computer (QPU)”; Pg. 1, Abstract, “each of the training curves requires the execution of thousands of quantum circuits in the quantum computer”; and Pg. 2, Col. 1, Fig. 1, where the “execution of thousands of quantum circuits” is by the Quantum Processing Unit, “QPU”, and as a result, the “training” and outputting, which require the “execution” are also, at least in part, by the Quantum Processing Unit; see also Pg. 1, Abstract, “Here, we leverage on Rigetti’s Quantum Cloud Services (QCS™) to overcome this implementation barrier, and we study the on-hardware performance of the data-driven quantum circuit learning (DDQCL) for three different state-of-the art classical solvers, and on two-different circuit ansatze associated to different entangling connectivity graphs for the same task”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the processing units, which simulate quantum computers and represent quantum objectives to executing a set of blocks of an input quantum circuit, train a parametrized quantum circuit, and output an expected output of the input quantum circuit of Cervera-Lierta with the executing, training, and outputting by the Quantum Processing Unit of Leyton-Ortega in order to utilize quantum hardware to enhance the method, which will allow for efficient and robust operations (Leyton-Ortega, Pg. 2, Col. 2, Para. 2, “To achieve this systematic and robust study, we leverage on Rigetti’s Quantum Cloud Services (QCS™) and its enhanced framework for handling the requirements of HQC algorithms efficiently. Some unique features of this platform include pre-compilation of the PQC programming cycles, active reset allowing for faster repetition rates, and low-latency due to co-location of the quantum and classical hosts”) that will produce more dependable results than could be achieved with classical hardware alone (Leyton-Ortega, Pg. 4, Col. 2, Para. 1, “this discrepancy in results between in silico simulations and the on-hardware ones is not to the topology itself, but most likely it is due to the quality of the qubits used for each experiment, with better quality for the subset used for the star experiments” with Cervera-Lierta, Pg. 10, Col. 1, Para. 1, “All simulations shown in this work are exact, i.e., they do not contain any noise model. We expect the meta-VQE to suffer similar effects of noise to other VQAs, effects that can be partially compensated by the use of error-mitigation techniques”), which can be implemented with high efficacy readout correction procedures (Leyton-Ortega, Pg. 9, Col. 1, Para. 2, “To test the efficacy of this procedure, in Fig. 5, we present the training with the bare and the readout corrected training. The latter shows significant improvement with a minimal score of KL ∼ 0.13 from ∼ 1.0 obtained using bare readouts. Therefore, we adopted the readout correction for all the experiments reported in the main text”), and can be achieved through cloud services that overcome the resource constraints that serve as barriers to implementation of quantum computers (Leyton-Ortega, Pg. 1, Abstract, “Since each of the training curves requires the execution of thousands of quantum circuits in the quantum computer, such a robust study remained a steep challenge for most hybrid platforms available today. Here, we leverage on Rigetti’s Quantum Cloud Services (QCS™) to overcome this implementation barrier”).
Regarding Claim 2, Cervera-Lierta in view of Leyton-Ortega teach the method of claim 1, wherein the additional block receives inputs only from the initial block (Cervera-Lierta, Pg. 3, Col. 1, Fig. 1 and Cervera-Lierta, Pg. 3, Col. 1, Para. 1, “The first part of the circuit is the encoding unitary S containing parameterized gates, the arguments of which include the Hamiltonian parameters of one of the training points ⃗ λi and the variational parameters ⃗. The second unitary of the circuit, U, also contains parameterized gates that depend on variables ⃗, but not the Hamiltonian parameters. The final state of the circuit can be written as
PNG
media_image1.png
48
472
media_image1.png
Greyscale
”, wherein the additional block, “The second unitary of the circuit, U”, which is depicted in Fig 1. as the “Processing” unitary, receives only inputs, “variables ⃗”, from the initial block, “the encoding unitary S”, which is depicted in Fig 1. as the “Encoding” unitary, and not other inputs such as “Hamiltonian parameters”, as shown in Fig. 1, where the “Processing” unitary only receives input from the “Encoding” unitary).
Regarding Claim 3, Cervera-Lierta in view of Leyton-Ortega teach the method of claim 1, wherein each block of the set of blocks receives inputs of an approximate quantum state from a set of preceding blocks (Cervera-Lierta, Pg. 3, Col. 1, Fig. 1, “the Hamiltonian parameters ⃗ λ are encoded together with the variational parameters ⃗ and ⃗.”; Cervera-Lierta, Pg. 2, Col. 2, Para. 2, “As these algorithms propose, we design a parameterized quantum circuit to be trained with a set of values from a given model, in our case, a physical Hamiltonian. We treat the encoding part of the meta-VQE as a quantum neural network that learns the encoding of the Hamiltonian. The processing part guides the encoded state towards the ground state. We observe two advantages of this algorithm: (i) it can be used to first explore the ground state energies of the Hamiltonian parameter space with only a few training points and then use the result as an initial state for a precise VQE, and (ii) the encoding in a VQA proves valuable and helps these algorithms to find the ground state”; and Cervera-Lierta, Pg. 3, Col. 1, Para. 1, “The first part of the circuit is the encoding unitary S containing parameterized gates, the arguments of which include the Hamiltonian parameters of one of the training points ⃗ λi and the variational parameters ⃗. The second unitary of the circuit, U, also contains parameterized gates that depend on variables ⃗, but not the Hamiltonian parameters. The final state of the circuit can be written as
PNG
media_image1.png
48
472
media_image1.png
Greyscale
”, wherein each block of the set of blocks receives inputs from a set of preceding blocks because the initial block, “the encoding unitary S”, receives an input state, as shown in Fig. 1, which must be from an operative unit that is within the broadest reasonable interpretation of a preceding block, and the additional block, “The second unitary of the circuit, U”, receives an input state, “the encoded state”, from the preceding initial block, “The processing part guides the encoded state towards the ground state”; see also Cervera-Lierta, Pg. 1, Col. 2, Para. 1, “The second block guides the algorithm around a particular region of the space of quantum states”; Cervera-Lierta, Pg. 6-7, Col. 2-1, Para. 4, “the objective is to approximate ground states of qubit Hamiltonians, generated in this example by transforming second-quantized fermionic operators using various transforms”; and Cervera-Lierta, Pg. 2, Col. 1, Para. 4, “the meta-VQE is trained with a small set of Hamiltonian parameters by constructing a cost function that is a combination of all expected values”, where each state is an approximate quantum state because they are generated to “approximate ground states of qubit Hamiltonians” that are “of quantum states”, which are approximated through a “meta-VQE [that] is trained”).
Regarding Claim 4, Cervera-Lierta in view of Leyton-Ortega teach the method of claim 3, wherein the approximate quantum state is approximated using any one of a parameterized quantum circuit, a Tensor Network, a Neural Network Quantum State, or other data structures that can be decomposed into a quantum circuit (Cervera-Lierta, Pg. 2, Col. 2, Para. 2, “As these algorithms propose, we design a parameterized quantum circuit to be trained with a set of values from a given model, in our case, a physical Hamiltonian. We treat the encoding part of the meta-VQE as a quantum neural network that learns the encoding of the Hamiltonian. The processing part guides the encoded state towards the ground state” and Cervera-Lierta, Pg. 3, Col. 1, Para. 1, “The first part of the circuit is the encoding unitary S containing parameterized gates, the arguments of which include the Hamiltonian parameters of one of the training points ⃗ λi and the variational parameters ⃗. The second unitary of the circuit, U, also contains parameterized gates that depend on variables ⃗, but not the Hamiltonian parameters. The final state of the circuit can be written as
PNG
media_image1.png
48
472
media_image1.png
Greyscale
”, wherein the approximate quantum state, “the encoded state”, is approximated using, at the least, a parameterized quantum circuit, “we design a parameterized quantum circuit to be trained with a set of values from a given model, in our case, a physical Hamiltonian”, wherein “The first part of the circuit is the encoding unitary S containing parameterized gates”).
Regarding Claim 5, Cervera-Lierta in view of Leyton-Ortega teach the method of claim 1, wherein dividing the input quantum circuit into a set of blocks comprises maximizing a size for each block in the set of blocks (Cervera-Lierta, Pg. 3, Col. 1, Para. 1, “The circuit is initialized . . . The first part of the circuit is the encoding unitary S containing parameterized gates, the arguments of which include the Hamiltonian parameters of one of the training points ⃗ λi and the variational parameters ⃗. The second unitary of the circuit, U, also contains parameterized gates that depend on variables ⃗, but not the Hamiltonian parameters”; Cervera-Lierta, Pg. 9, Col. 1, Para. 3, “the meta-VQE circuit is divided into two parts, the encoding and processing unitaries”; and Cervera-Lierta, Pg. 3, Col. 1, Fig. 1, where the input quantum circuit, “The circuit”, is divided into a set of blocks, “is divided into two parts”, such that, as depicted in figure 3, see Cervera-Lierta, Pg. 5, Fig. 3, the length of each block, “Encoding” and “Processing” is maximized in the vertical direction to be the entire vertical length of the “circuit”)
such that the set of blocks are executable on the Quantum Processing Unit (Cervera-Lierta, Pg. 2, Col. 2, Para. 2, “As these algorithms propose, we design a parameterized quantum circuit to be trained with a set of values from a given model, in our case, a physical Hamiltonian. We treat the encoding part of the meta-VQE as a quantum neural network that learns the encoding of the Hamiltonian. The processing part guides the encoded state towards the ground state. We observe two advantages of this algorithm: (i) it can be used to first explore the ground state energies of the Hamiltonian parameter space with only a few training points and then use the result as an initial state for a precise VQE, and (ii) the encoding in a VQA proves valuable and helps these algorithms to find the ground state”; Cervera-Lierta, Pg. 3, Col. 1, Para. 1, “The first part of the circuit is the encoding unitary S containing parameterized gates, the arguments of which include the Hamiltonian parameters of one of the training points ⃗ λi and the variational parameters ⃗. The second unitary of the circuit, U, also contains parameterized gates that depend on variables ⃗, but not the Hamiltonian parameters. The final state of the circuit can be written as
PNG
media_image1.png
48
472
media_image1.png
Greyscale
”; and Cervera-Lierta, Pg. 3, Col. 1, Fig. 1, where the set of blocks, “The first part of the circuit is the encoding unitary . . . The second unitary of the circuit, U”, are executed to generate “the ground state”, which, in view of Leyton-Ortega, is by the quantum processing unit, see Leyton-Ortega, Pg. 6, Col. 2, Para. 1, “In this work, we implement a hybrid quantum-classical algorithm for unsupervised machine learning tasks introduced in (Benedetti et al. 2019) at Rigetti’s superconducting quantum computer (QPU)”; Leyton-Ortega, Pg. 1, Abstract, “each of the training curves requires the execution of thousands of quantum circuits in the quantum computer”; and Leyton-Ortega, Pg. 2, Col. 1, Fig. 1)
having a noise property, wherein the noise property is any one of a single qubit error rate, a gate error rate, a crosstalk error, a readout error or a device dependent parameter (Leyton-Ortega, Pg. 2-3, Col. 2-2, Section, “DDQCL execution on Rigetti’s QPU” and Leyton-Ortega, Pg. 7-8, Col. 2-2, Para. 5-1, “DDQCL is based on the probability distribution of the N qubit quantum statev . . . Unfortuantely, no device is 100% free of errors in this transformation from quantum state to bitstring output. Next, we describe a simple model to cope with this classical readout channel. The simple error readout correction implemented here consist of the measurement characterization of all the 2N . . . In a noiseless scenario that measurement should yield a bitstring x = 0010 in the classical register for every single shot . . . However, in the experiment the readouts correspond to different elements . . . due to assignment errors”, where the “DDQCL execution on Rigetti’s QPU” has a noise property, “no device is 100% free of errors”, which is a readout error, “error readout” when compared to “a noiseless scenario”, which is “due to assignment errors”; see also Leyton-Ortega, Pg. 5, Col. 2, Para. 2, “We considered ansatze that prepare an extensive set of 4-qubit QCBMs with real amplitudes, which under noiseless condition, have an arbitrary excellent performance in the learning of the 2 × 2 BAS dataset considered here. In contrast, these same ansatze and training settings have suboptimal performance which is limited by the different noise sources affecting the quantum processor output. Although we cannot disentangle the various noise sources, these application-inspired benchmarks offer valuable insights into how much device noise affects the performance of the machine in the generative model”).
The reasons for obviousness were discussed in regard to the rejection of Claim 1 above and remain applicable here.
Regarding Claim 9, Cervera-Lierta in view of Leyton-Ortega teach the method of claim 1, wherein executing, by the Quantum Processing Unit, the set of blocks does not require classical simulation for all the qubits of the input circuit (Cervera-Lierta, Pg. 2, Col. 2, Para. 2, “As these algorithms propose, we design a parameterized quantum circuit to be trained with a set of values from a given model, in our case, a physical Hamiltonian. We treat the encoding part of the meta-VQE as a quantum neural network that learns the encoding of the Hamiltonian. The processing part guides the encoded state towards the ground state. We observe two advantages of this algorithm: (i) it can be used to first explore the ground state energies of the Hamiltonian parameter space with only a few training points and then use the result as an initial state for a precise VQE, and (ii) the encoding in a VQA proves valuable and helps these algorithms to find the ground state”, where the set of blocks, “The first part of the circuit is the encoding unitary . . . The second unitary of the circuit, U”, are executed to generate “the ground state”, which in view of Leyton-Ortega, is by the Quantum Processing Unit, Leyton-Ortega, Pg. 6, Col. 2, Para. 1, “In this work, we implement a hybrid quantum-classical algorithm for unsupervised machine learning tasks introduced in (Benedetti et al. 2019) at Rigetti’s superconducting quantum computer (QPU)”; Leyton-Ortega, Pg. 1, Abstract, “each of the training curves requires the execution of thousands of quantum circuits in the quantum computer”; and Leyton-Ortega, Pg. 2, Col. 1, Fig. 1, such that it does not require classical simulation for all the qubits of the input circuit, see Leyton-Ortega, Pg. 3, Col. 1, Para. 2, “For the experimental implementation of DDQCL, we consider a 4-qubit subunit connected in star and line entangling topology, according to the connectivity constraints in the 16-qubit Aspen quantum computer” and Leyton-Ortega, Pg. 1, Abstract, “we leverage on Rigetti’s Quantum Cloud Services (QCS™) to overcome this implementation barrier, and we study the on-hardware performance of the data-driven quantum circuit learning (DDQCL) for three different state-of-the art classical solvers”; see generally Leyton-Ortega, Pg. 2, Col. 1, Para. 3, “Experimental implementations of QCBMs and of DDQCL where the entire learning process performed on hardware have been recently demonstrated in both, super conducting qubits”).
The reasons for obviousness were discussed in regard to the rejection of Claim 1 above and remain applicable here.
Regarding Claim 12, Cervera-Lierta in view of Leyton-Ortega teach a system comprising: a cloud computing environment configured to: . . . (Leyton-Ortega, Pg. 1, Abstract, “Here, we leverage on Rigetti’s Quantum Cloud Services (QCS™) to overcome this implementation barrier, and we study the on-hardware performance of the data-driven quantum circuit learning (DDQCL) for three different state-of-the art classical solvers, and on two-different circuit ansatze associated to different entangling connectivity graphs for the same task”, where a system comprises a cloud computing environment, “we leverage on Rigetti’s Quantum Cloud Services (QCS™) . . . [to] study the on-hardware performance”, to perform operations, “performance of the data-driven quantum circuit learning”; see also Leyton-Ortega, Pg. 2, Col. 2, Para. 2, “To achieve this systematic and robust study, we leverage on Rigetti’s Quantum Cloud Services (QCS™) and its enhanced framework for handling the requirements of HQC algorithms efficiently”, where the “on Rigetti’s Quantum Cloud Services (QCS™)” is used in the system of Cervera-Lierta, see Cervera-Lierta, Pg. 2, Col. 2, Para. 3-4, “we introduce the meta-VQE algorithm . . . we apply this algorithm to a state-of-the-art application: the simulation of transmon qubits in a quantum computer . . . The structure of the meta-VQE algorithm” and Cervera-Lierta, Pg. 2, Col. 2, Para. 3, “Results are obtained via implementation of the meta-VQE protocol using TEQUILA [30] and choosing QULACS [31] as a quantum backend. TEQUILA represents quantum objectives as generalized functions of abstract expectation values, allowing arbitrary transformations and arithmetic operations on those data types to be done in a blackboard fashion”).
The reasons for obviousness were discussed in regard to the rejection of Claim 1 above and remain applicable here.
The remaining limitations are substantially the same as limitations of Claim 1, therefore it is
rejected under the same rationale.
Regarding Claim 13, the additional elements of the dependent claim are substantially the same as limitations of Claim 2, therefore it is rejected under the same rationale.
Regarding Claim 14, the additional elements of the dependent claim are substantially the same as limitations of Claim 3, therefore it is rejected under the same rationale.
Regarding Claim 15, the additional elements of the dependent claim are substantially the same as limitations of Claim 5, therefore it is rejected under the same rationale.
Regarding Claim 18, the additional elements of the dependent claim are substantially the same as limitations of Claim 9, therefore it is rejected under the same rationale.
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Cervera-Lierta in view of Leyton-Ortega and Wilson et al. (hereinafter Wilson) (“Just-in-time Quantum Circuit Transpilation Reduces Noise”).
Regarding Claim 6, Cervera-Lierta in view of Leyton-Ortega teach the method of claim 5, wherein the noise property is . . . [a readout error] (Leyton-Ortega, Pg. 2-3, Col. 2-2, Section, “DDQCL execution on Rigetti’s QPU” and Leyton-Ortega, Pg. 7-8, Col. 2-2, Para. 5-1, “DDQCL is based on the probability distribution of the N qubit quantum statev . . . Unfortuantely, no device is 100% free of errors in this transformation from quantum state to bitstring output. Next, we describe a simple model to cope with this classical readout channel. The simple error readout correction implemented here consist of the measurement characterization of all the 2N . . . In a noiseless scenario that measurement should yield a bitstring x = 0010 in the classical register for every single shot . . . However, in the experiment the readouts correspond to different elements . . . due to assignment errors”, where the “DDQCL execution on Rigetti’s QPU” has a noise property, “no device is 100% free of errors”, which is a readout error, “error readout” when compared to “a noiseless scenario”, which is “due to assignment errors”; see also Leyton-Ortega, Pg. 5, Col. 2, Para. 2, “We considered ansatze that prepare an extensive set of 4-qubit QCBMs with real amplitudes, which under noiseless condition, have an arbitrary excellent performance in the learning of the 2 × 2 BAS dataset considered here. In contrast, these same ansatze and training settings have suboptimal performance which is limited by the different noise sources affecting the quantum processor output. Although we cannot disentangle the various noise sources, these application-inspired benchmarks offer valuable insights into how much device noise affects the performance of the machine in the generative model”).
The reasons for obviousness were discussed in regard to the rejection of Claim 1 above and remain applicable here.
Cervera-Lierta in view of Leyton-Ortega do not explicitly disclose . . . between 0 percent and 25 percent.
However, Wilson teaches . . . [a readout error] between 0 percent and 25 percent (Pg. 3, Col. 1, Para. 2, “Readout errors affect any circuit, and their probability is relatively high on today’s NISQ computers. Readout errors are reported to be in the order of 10−2 for IBM Q devices. We even observed that sometimes they can be as high as 10%”; see also Pg. 3, Col. 2, Para. 3, “The readout error is this calculated as the percent of incorrect results returned from the respective circuits”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the noise property that is a readout error of Cervera-Lierta in view of Leyton-Ortega with the readout error between 0 percent and 25 percent of Wilson in order to increase the applicability and utility of the readout error correction procedure (Leyton-Ortega, Pg. 9, Col. 1, Para. 2, “To test the efficacy of this procedure, in Fig. 5, we present the training with the bare and the readout corrected training. The latter shows significant improvement with a minimal score of KL ∼ 0.13 from ∼ 1.0 obtained using bare readouts. Therefore, we adopted the readout correction for all the experiments reported in the main text”) by improving accuracy (Wilson, Pg. 1, Col. 1, Abstract, “The framework assess two major sources of noise, namely readout errors (measurement errors) and two-qubit gate/connection errors. Experiments indicate that the accuracy of circuit results improves by 3-304% on average and up to 400% with on-the-fly circuit mappings based on error measurements just prior to application execution”) of quantum circuits with readout error percentages that fall within expected noise levels for IBM devices (Wilson, Pg. 3, Col. 1, Para. 2, “Readout errors affect any circuit, and their probability is relatively high on today’s NISQ computers. Readout errors are reported to be in the order of 10−2 for IBM Q devices. We even observed that sometimes they can be as high as 10%” and Wilson, Pg. 1, Col. 1, Abstract, “Running quantum programs is fraught with challenges on on today’s noisy intermediate scale quantum (NISQ) devices”).
Claims 7 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Cervera-Lierta
in view of Leyton-Ortega and Sadeghi et al. (hereinafter Sadeghi) (“DC: Depth Control on Quantum Classical Circuit”).
Regarding Claim 7, Cervera-Lierta in view of Leyton-Ortega teach the method of claim 1, wherein dividing the input quantum circuit into a set of blocks comprises selecting a size for each block in the set of blocks . . . (Cervera-Lierta, Pg. 3, Col. 1, Para. 1, “The circuit is initialized . . . The first part of the circuit is the encoding unitary S containing parameterized gates, the arguments of which include the Hamiltonian parameters of one of the training points ⃗ λi and the variational parameters ⃗. The second unitary of the circuit, U, also contains parameterized gates that depend on variables ⃗, but not the Hamiltonian parameters”; Cervera-Lierta, Pg. 9, Col. 1, Para. 3, “the meta-VQE circuit is divided into two parts, the encoding and processing unitaries”; and Cervera-Lierta, Pg. 3, Col. 1, Fig. 1, where the input quantum circuit, “The circuit”, is divided into a set of blocks, “is divided into two parts”, which comprises selecting a size, “two . . . layers”, for each block in the set of blocks, “encoding and processing”, see Cervera-Lierta, Pg. 4, Col. 2, Para. 3, “Figure 3 shows the circuit ansatz described above for n =4qubits and two encoding and processing layers” and Cervera-Lierta, Pg. 5, Fig. 3)
Cervera-Lierta in view of Leyton-Ortega do not explicitly disclose . . . such that a variance of a fidelity is larger than a predetermined threshold, wherein the variance of the fidelity is obtained by varying one or many gate parameters in the block.
However, Sadeghi teaches . . . [selecting block size of a quantum circuit] (Pg. 1, Col. 1, Abstract, “This study presents a revolutionary Depth Control (DC) method ology that involves slicing and lowering the depth of conventional circuits. This strategy aims to improve the reliability and decrease the mapping costs associated with quantum hardware. DC pro vides reliable outcomes for circuits of indefinite size on any Noisy Intermediate-Scale Quantum (NISQ) system” and Pg. 7, Col. 1, Para. 2, “Our main objective is to determine the block size that enables us achieve our goal of limiting error propagation for an infinitely-large circuit”, where block size of a quantum circuit is selected, “determine the block size” “for circuits of indefinite size on any Noisy Intermediate-Scale Quantum (NISQ) system”)
such that a variance of a fidelity is larger than a predetermined threshold, wherein the variance of the fidelity is obtained by varying one or many gate parameters in the block (Pg. 7, Col. 2, Algo. 1 and Pg. 8, Col. 2, Para.2-4, “we propose Dynamic-Depth-Control (DDC), which is essentially a ’divide-and-conquer’ approach, designed to establish the appropriate block sizes . . . the DDC approach allows for ’varying’ block sizes, based on the gate operations within each block . . . In each iteration, we divide the circuit in half and evaluate the block’s fidelity. If the fidelity satisfies the threshold, we proceed to use the same counting standard specified in SDC V-B, initializing the remainder of the circuit with the output of the previous iteration. If the fidelity is below the threshold, we continue to halve the block until the ’desired fidelity level’ is achieved”, where “the fidelity” at a given “iteration” can reasonably be described as a variance of a fidelity because it is a “fidelity” that has changed based on the “divid[ing]” operation, “In each iteration, we divide the circuit in half and evaluate the block’s fidelity”, such that the selecting block size is based on the variance of a fidelity is larger than a predetermined threshold, “If the fidelity satisfies the threshold, we proceed . . . If the fidelity is below the threshold, we continue to halve the block until the ’desired fidelity level’ is achieved”, wherein the variance of the fidelity is obtained by varying one or many gate parameters in the block, “the DDC approach allows for ’varying’ block sizes, based on the gate operations within each block” and see algorithm 1 where varying one or more gate parameters, “Circuit[i]←−Circuit.break( n i+1) ; // grab the first n i+1 gates in the circuit”, is used to “calculate the fidelity of circuit”; see also Pg. 8, Col. 2, Para. 3, “For the sake of consistency with the SDC method, we use ’gate counts’ in our algorithm”, where “gate counts” is a gate parameter; see generally Pg. 7-8, Col. 1-2, Para. 3-1, section “Static Depth Control (SDC)”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the dividing of the input quantum circuit into a set of blocks, which comprises selecting a size for each block in the set of blocks of Cervera-Lierta in view of Leyton-Ortega with the selecting block size of a quantum circuit such that a variance of a fidelity is larger than a predetermined threshold, wherein the variance of the fidelity is obtained by varying one or many gate parameters in the block of Sadeghi in order to determine block sizes that satisfy a desired fidelity level (Sadeghi, Pg. 8, Col. 2, Para. 4, “In each iteration, we divide the circuit in half and evaluate the block’s fidelity. If the fidelity satisfies the threshold, we proceed to use the same counting standard specified in SDC V-B, initializing the remainder of the circuit with the output of the previous iteration. If the fidelity is below the threshold, we continue to halve the block until the ’desired fidelity level’ is achieved”), using an efficient algorithm that reduces the number of jobs for higher thresholds (Sadeghi, Pg. 10, Col. 1, Para. 3, “In the DDC method on the other hand, the number of jobs is o(klogn) =o(logn) (where n is circuit depth and k is a constant related to system reliability and the large order remains as O(n)), if the system is sufficiently reliable and the intended fidelity margin is large”; Sadeghi, Pg. 8, Col. 2, Para. 5, “A higher threshold leads to a lower error rate and a larger number of jobs. Conversely, a lower threshold would result in fewer jobs at the expense of accepting a higher error risk”; and Sadeghi, Pg. 11, Col. 2, Para. 2, “Our results on single input state also show that Dynamic Depth Control (DDC) achieves a 54.7% PST, while Static Depth Control (SDC) reported a higher PST of 72.1%. DDC results indicate that it can produce reliable outputs Avg(E −max(F) =44.3%) when the threshold is adequately high and can concurrently reduce the number of jobs by approximately 3x”).
Regarding Claim 16, the additional elements of the dependent claim are substantially the same as limitations of Claim 7, therefore it is rejected under the same rationale.
Claims 8 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Cervera-Lierta
in view of Leyton-Ortega, Sadeghi, and Hartmanns et al. (hereinafter Hartmanns) (“Optimistic Value Iteration”).
Regarding Claim 8, Cervera-Lierta in view of Leyton-Ortega and Sadeghi teach the method of claim 7, wherein selecting a size for each block in the set of blocks comprises . . . (Cervera-Lierta, Pg. 3, Col. 1, Para. 1, “The circuit is initialized . . . The first part of the circuit is the encoding unitary S containing parameterized gates, the arguments of which include the Hamiltonian parameters of one of the training points ⃗ λi and the variational parameters ⃗. The second unitary of the circuit, U, also contains parameterized gates that depend on variables ⃗, but not the Hamiltonian parameters”; Cervera-Lierta, Pg. 9, Col. 1, Para. 3, “the meta-VQE circuit is divided into two parts, the encoding and processing unitaries”; and Cervera-Lierta, Pg. 3, Col. 1, Fig. 1, where the input quantum circuit, “The circuit”, is divided into a set of blocks, “is divided into two parts”, which comprises selecting a size, “two . . . layers”, for each block in the set of blocks, “encoding and processing”, see Cervera-Lierta, Pg. 4, Col. 2, Para. 3, “Figure 3 shows the circuit ansatz described above for n =4qubits and two encoding and processing layers” and Cervera-Lierta, Pg. 5, Fig. 3)
[altering] the size of each block gradually in iterations until the size of each block reaches the selected block size during the execution of each block (Sadeghi, Pg. 7, Col. 2, Algo. 1 and Sadeghi, Pg. 8, Col. 2, Para.2-4, “we propose Dynamic-Depth-Control (DDC), which is essentially a ’divide-and-conquer’ approach, designed to establish the appropriate block sizes . . . the DDC approach allows for ’varying’ block sizes, based on the gate operations within each block . . . In each iteration, we divide the circuit in half and evaluate the block’s fidelity. If the fidelity satisfies the threshold, we proceed to use the same counting standard specified in SDC V-B, initializing the remainder of the circuit with the output of the previous iteration. If the fidelity is below the threshold, we continue to halve the block until the ’desired fidelity level’ is achieved”, where the size of each block is gradually altered in iterations, “In each iteration, we divide the circuit in half”, until the size reaches selected “desired fidelity level”, which corresponds with the size of each block during the execution of each block, see Sadeghi, Pg.7, Col. 1, Para. 2, “Our main objective is to determine the block size that enables us achieve our goal of limiting error propagation for an infinitely-large circuit” and Cervera-Lierta, Pg. 2, Col. 2, Para. 2, “As these algorithms propose, we design a parameterized quantum circuit to be trained with a set of values from a given model, in our case, a physical Hamiltonian. We treat the encoding part of the meta-VQE as a quantum neural network that learns the encoding of the Hamiltonian. The processing part guides the encoded state towards the ground state. We observe two advantages of this algorithm: (i) it can be used to first explore the ground state energies of the Hamiltonian parameter space with only a few training points and then use the result as an initial state for a precise VQE, and (ii) the encoding in a VQA proves valuable and helps these algorithms to find the ground state”; Cervera-Lierta, Pg. 3, Col. 1, Para. 1, “The first part of the circuit is the encoding unitary S containing parameterized gates, the arguments of which include the Hamiltonian parameters of one of the training points ⃗ λi and the variational parameters ⃗. The second unitary of the circuit, U, also contains parameterized gates that depend on variables ⃗, but not the Hamiltonian parameters. The final state of the circuit can be written as
PNG
media_image1.png
48
472
media_image1.png
Greyscale
”; and Cervera-Lierta, Pg. 3, Col. 1, Fig. 1, where the set of blocks, “The first part of the circuit is the encoding unitary . . . The second unitary of the circuit, U”, are executed to generate “the ground state”).
The reasons for obviousness were discussed in regard to the rejection of Claim 7 above and remain applicable here.
Cervera-Lierta in view of Leyton-Ortega and Sadeghi do not explicitly disclose . . . increasing . . . .
However, Hartmanns teaches . . . increasing [a value gradually in iterations until a specific value is reached] (Pg. 1-2, Para. 3-1, “The standard algorithm to compute optimal (maximum or minimum) probabilities or reward values on MDP is value iteration (VI) . . . [it] locally improves the values iteratively until a “convergence” criterion is met, and then reports the final value for the initial state as the overall result. The initial values are chosen to be an underapproximation of the true values (e.g. 0 for all states in case of probabilities or non-negative rewards). The final values are then an improved underapproximation of the true values”, where a value, “The initial values are chosen to be an underapproximation of the true values (e.g. 0”, is increased gradually in iterations until a specific value is reached, “locally improves the values iteratively until a “convergence” criterion is met” and “The final values are then an improved underapproximation of the true values”; see also Pg. 5-8, section “3 Value Iteration” and Pg. 6, Algo. 1).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the selecting a size for each block in the set of blocks that comprises altering the size of each block gradually in iterations until the size of each block reaches the selected block size during the execution of each block of Cervera-Lierta in view of Leyton-Ortega and Sadeghi with the increasing a value gradually in iterations until a specific value is reached of Hartmanns to improve efficiency by avoiding the need to determine an upper bound starting block size, which may turn out to be either incorrect or an overly conservative estimate (compare Hartmanns, Pg. 16, Para. 1, “The plots in Fig. 5 paint a very clear picture of OVI being significantly faster for expected rewards than II (which suffers from the need to precompute initial upper bounds that then turn out to be rather conservative)” and Hartmanns, Pg. 1-2, Para. 3-1, “The standard algorithm to compute optimal (maximum or minimum) probabilities or reward values on MDP is value iteration (VI) . . . [it] locally improves the values iteratively until a “convergence” criterion is met, and then reports the final value for the initial state as the overall result. The initial values are chosen to be an underapproximation of the true values (e.g. 0 for all states in case of probabilities or non-negative rewards). The final values are then an improved underapproximation of the true values”, where an initial size of “0” can be applied uniformly across applications and is guaranteed to be less than the ideal block size with Sadeghi, Pg. 8, Col. 2, Para.2-4, “we propose Dynamic-Depth-Control (DDC), which is essentially a ’divide-and-conquer’ approach, designed to establish the appropriate block sizes . . . the DDC approach allows for ’varying’ block sizes, based on the gate operations within each block . . . In each iteration, we divide the circuit in half and evaluate the block’s fidelity. If the fidelity satisfies the threshold, we proceed to use the same counting standard specified in SDC V-B, initializing the remainder of the circuit with the output of the previous iteration. If the fidelity is below the threshold, we continue to halve the block until the ’desired fidelity level’ is achieved”, where the initial “block size” needs to be calculated and may be smaller than ideal).
Regarding Claim 17, the additional elements of the dependent claim are substantially the same as limitations of Claim 8, therefore it is rejected under the same rationale.
Claims 10-11 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Cervera-Lierta
in view of Leyton-Ortega and Holmes et al. (hereinafter Holmes) (“Connecting Ansatz Expressibility to Gradient Magnitudes and Barren Plateaus”).
Regarding Claim 10, Cervera-Lierta in view of Leyton-Ortega teach the method of claim 1 further comprising . . . [operations] (Cervera-Lierta, Pg. 2, Col. 2, Para. 3-4, “we introduce the meta-VQE algorithm . . . we apply this algorithm to a state-of-the-art application: the simulation of transmon qubits in a quantum computer . . . The structure of the meta-VQE algorithm is shown diagrammatically in Fig. 1”; see also Cervera-Lierta, Pg. 3, Col. 1, Fig. 1, “Schematic illustration of the meta-VQE algorithm”).
Cervera-Lierta in view of Leyton-Ortega do not explicitly disclose . . . detecting an onset of a Barren Plateau, wherein the detection of the onset includes identifying any one of a lack of expressibility, noise impact, quantum entanglement impact, or an overparameterization.
However, Holmes teaches . . . detecting an onset of a Barren Plateau, wherein the detection of the onset includes identifying any one of a lack of expressibility, noise impact, quantum entanglement impact, or an overparameterization (Pg. 4, Col. 2, Para. 2-3, “There is a growing awareness of the so-called barren plateau phenomenon . . . For a given ansatz U(θ), a cost C is said to exhibit a barren plateau if its gradients vanish exponentially with the number of qubits n. This is typically relaxed to a probabilistic definition, where the gradient vanishes exponentially with high probability. This would follow from Chebyshev’s inequality, given in Eq. (9), if the variance in the partial derivative vanishes exponentially, i.e., if Var[∂kC] ∈ O(2−pn) for any integer p >0. For costs that exhibit barren plateaus, exponentially precise measurements may be required to determine the minimization direction and hence the cost is effectively untrainable for large problem sizes”, where an onset of a Barren Plateau, “exhibit barren plateaus” as shown based on “a cost C [that] is said to exhibit a barren plateau if its gradients vanish exponentially with the number of qubits n”, is detected mathematically, “This is typically relaxed to a probabilistic definition . . . follow[ing] from Chebyshev’s inequality, given in Eq. (9)”, which includes identifying expressibility, see Pg. 9, Col. 2, Para. 3, “Our bounds demonstrate that increasing the expressibility of anansatz can result in smaller cost gradients” and Pg. 1, Abstract, “Our resulting bounds indicate that highly expressive ansatze exhibit flatter cost landscapes and therefore will be harder to train”, which is a lack of expressibility for anything less than “maximally expressive ansatze”, compare Pg. 4, Col. 2, Para. 5, “maximally expressive ansatze exhibit barren plateaus” with Pg. 5, Col. 1, Para. 2, “highly inexpressive ansatze need not always experience large cost gradients and in fact they may exhibit vanishing gradient”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the quantum circuit operations of the method operationalized by quantum processing units of a cloud environment of Cervera-Lierta in view of Leyton-Ortega with the detecting an onset of a Barren Plateau, wherein the detection of the onset includes identifying a lack of expressibility of Holmes in order to conduct quantum experiments using the cloud environment (Leyton-Ortega, Pg. 2, Col. 2, Para. 2, “In this work, we present the first experimentally robust comparison of each of the aspects of HQC algorithms . . . To achieve this systematic and robust study, we leverage on Rigetti’s Quantum Cloud Services (QCS™) and its enhanced frame work for handling the requirements of HQC algorithms efficiently”) to find an optimal ansatz solution that is both expressive and trainable (Holmes, Pg. 1, Col. 1, Para. 2, “To successfully find an optimal solution, the ansatz should ideally be both expressive and trainable. Specifically, the ansatz must be sufficiently expressive such that it contains a circuit that well-approximates the optimal solution. Concurrently, the cost landscape must be sufficiently featured to be able to train the parameters to find this optimal solution”), where increasing expressibility in response to a lack of expressibility will allow measurement to determine a minimization direction and upper bonds for trainability (Holmes, Pg. 4, Col. 2, Para. 2, “For costs that exhibit barren plateaus, exponentially precise measurements may be required to determine the minimization direction”; Holmes, Pg. 2, Col. 1, Para. 3-4, “Our main results can be summarized in Fig. 1 . . . highly expressive ansatze, which are generically used for many problems, as they can access a much larger space [Fig. 1(c)], are shown to lead to small gradients and hence can have trainability issues. Since our analytic bounds are upper bounds, they leave open the questions of how reducing the expressibility of an ansatz changes the cost landscape and hence how reducing the expressibility can be used to avoid the barren plateau phenomenon”; and Holmes, Pg. 2, Col. 2, Fig. 1).
Regarding Claim 11, Cervera-Lierta in view of Leyton-Ortega and Holmes teach the method of claim 10, wherein in response to identifying the lack of expressibility, a cloud computing environment or a quantum edge environment is further configured to adjust an Ansatz to increase the expressibility (Leyton-Ortega, Pg. 2, Col. 2, Para. 2, “In this work, we present the first experimentally robust comparison of each of the aspects of HQC algorithms . . . To achieve this systematic and robust study, we leverage on Rigetti’s Quantum Cloud Services (QCS™) and its enhanced frame work for handling the requirements of HQC algorithms efficiently”, where a cloud computing environment, “Rigetti’s Quantum Cloud Services (QCS™) and its enhanced frame work”, is further configured to perform operations, “we leverage . . . for handling the requirements of HQC algorithms efficiently”, which, in view of Holmes, comprises adjusting an Ansatz to increase expressibility, see Holmes, Pg. 4, Col. 2, Para. 2, “For costs that exhibit barren plateaus, exponentially precise measurements may be required to determine the minimization direction” and Holmes, Pg. 1, Col. 1, Para. 2, “To successfully find an optimal solution, the ansatz should ideally be both expressive and trainable. Specifically, the ansatz must be sufficiently expressive such that it contains a circuit that well-approximates the optimal solution”, in response to identifying the lack of expressibility, see Holmes, Pg. 2, Col. 1, Para. 3-4, “Our main results . . . [include] analytic bounds that are upper bounds” and Holmes, Pg. 4, Col. 2, Para. 2-3, “There is a growing awareness of the so-called barren plateau phenomenon . . . For a given ansatz U(θ), a cost C is said to exhibit a barren plateau if its gradients vanish exponentially with the number of qubits n. This is typically relaxed to a probabilistic definition, where the gradient vanishes exponentially with high probability.
This would follow from Chebyshev’s inequality, given in Eq. (9), if the variance in the partial derivative vanishes exponentially, i.e., if Var[∂kC] ∈ O(2−pn) for any integer p >0. For costs that exhibit barren plateaus, exponentially precise measurements may be required to determine the minimization direction and hence the cost is effectively untrainable for large problem sizes”, where an onset of a Barren Plateau, “exhibit barren plateaus” as shown based on “a cost C [that] is said to exhibit a barren plateau if its gradients vanish exponentially with the number of qubits n”, is detected mathematically, “This is typically relaxed to a probabilistic definition . . . follow[ing] from Chebyshev’s inequality, given in Eq. (9)”, which includes identifying expressibility, see Holmes, Pg. 9, Col. 2, Para. 3, “Our bounds demonstrate that increasing the expressibility of anansatz can result in smaller cost gradients” and Holmes, Pg. 1, Abstract, “Our resulting bounds indicate that highly expressive ansatze exhibit flatter cost landscapes and therefore will be harder to train”, which is a lack of expressibility for anything less than “maximally expressive ansatze”, compare Holmes, Pg. 4, Col. 2, Para. 5, “maximally expressive ansatze exhibit barren plateaus” with Holmes, Pg. 5, Col. 1, Para. 2, “highly inexpressive ansatze need not always experience large cost gradients and in fact they may exhibit vanishing gradient”).
The reasons for obviousness were discussed in regard to the rejection of Claim 10 above and remain applicable here.
Regarding Claim 19, the additional elements of the dependent claim are substantially the same as limitations of Claim 10, therefore it is rejected under the same rationale.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Gentinetta et al. (“Overhead-constrained circuit knitting for variational quantum dynamics”) discloses alternative operations for dividing quantum circuits (Pg. 5, Fig. 1).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MATTHEW BRYCE GOLAN whose telephone number is (571)272-5159. The examiner can normally be reached Monday through Friday, 8:00 AM to 5:00 PM ET.
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, Alexey Shmatov can be reached at (571) 270-3428. 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.
/MATTHEW BRYCE GOLAN/Examiner, Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123
