DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application is being examined under the pre-AIA first to invent provisions.
Response to Arguments
112 Rejection Arguments
Applicant asserts:
Applicant argues, on page 9-10, that “The element "obtaining training samples efficiently" in original claims 1 and 15 has been amended to "automatically generating training samples". Support for this amendment is found in the specification, which describes that "training samples can be generated automatically and inexpensively." A person of ordinary skill in the art would understand this to be a specific mechanism for obtaining samples efficiently.”
Examiner response:
Examiner respectfully disagrees. The term “efficiently” 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. The examiner notes that it is not clear how a system or person generates training samples more efficiently. What aspect of generating the training samples makes it more efficient? Is there less training samples to be processed; is the training data capturing more accurately; etc.
Examiner notes that Applicant gives examples to what is considered physical characteristics of a family of quantum devices. However, these examples are not actively claimed. Therefore, Examiner uses BRI to include any information relating to physical characteristics of quantum devices.
Applicant asserts:
Applicant argues, on page 9-10, that "The phrase "rather than a narrow task aimed at a specific computational end" has been removed from claims 1 and 15. The distinction is now clarified by the amended language specifying that the "general computational task" is one for which training samples are "automatically generated," which contrasts with a specific end-task that would typically require a specialized, non-automatically generated dataset.”
Examiner response:
Examiner respectfully disagrees. The term “general computational task” 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. The examiner notes that it is not clear how a computational task is defined as a general. What is considered a general task?
102 Arguments
Applicant asserts:
Applicant argues, on page 11-12, The rejection of claims 1, 5-8, 15, and 19-22 under 35 U.S.C. § 102 as anticipated by Andrea is respectfully traversed. Particularly, “Andrea fails to do so Specifically, amended independent claims 1 and 15 recite pretraining a model "to embed information about physical characteristics of a family of quantum devices."”
Examiner response:
Examiner respectfully disagrees. When examining the claims, the Examiner uses BRI. Applicant gives examples to what is considered physical characteristics of a family of quantum devices. However, these examples are not actively claimed. Therefore, under BRI, any information relating to physical properties of quantum devices will disclose the claimed invention.
Applicant asserts:
Applicant argues, on page 11-12, The rejection of claims 1, 5-8, 15, and 19-22 under 35 U.S.C. § 102 as anticipated by Andrea is respectfully traversed. Particularly, “amended claims 1 and 15 recite "automatically generating training samples" for the pre-training task.”
Examiner response:
Applicant’s arguments with respect to claim(s) 1, 5-8, 15, and 19-22 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Claim Rejections - 35 USC § 112b
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 1-28 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
The term “generating training samples efficiently” and “general computational task” in claim 1 and 15 is a relative term which renders the claim indefinite. The term “generating training samples efficiently” and “general computational task” 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. The examiner notes that it is not clear how a system or person generates training samples more efficiently; how a computational task is defined as a general.
Dependent claims 2-14 and 16-28 are rejected for taking on the limitations of the claim it is dependent on.
Regarding claim 4 and 18, the phrase "including" renders the claim indefinite because it is unclear whether the limitation(s) following the phrase are part of the claimed invention. See MPEP § 2173.05(d).
Double Patenting
The nonstatutory double patenting rejection is based on a judicially created doctrine grounded in public policy (a policy reflected in the statute) so as to prevent the unjustified or improper timewise extension of the “right to exclude” granted by a patent and to prevent possible harassment by multiple assignees. A nonstatutory double patenting rejection is appropriate where the conflicting claims are not identical, but at least one examined application claim is not patentably distinct from the reference claim(s) because the examined application claim is either anticipated by, or would have been obvious over, the reference claim(s). See, e.g., In re Berg, 140 F.3d 1428, 46 USPQ2d 1226 (Fed. Cir. 1998); In re Goodman, 11 F.3d 1046, 29 USPQ2d 2010 (Fed. Cir. 1993); In re Longi, 759 F.2d 887, 225 USPQ 645 (Fed. Cir. 1985); In re Van Ornum, 686 F.2d 937, 214 USPQ 761 (CCPA 1982); In re Vogel, 422 F.2d 438, 164 USPQ 619 (CCPA 1970); In re Thorington, 418 F.2d 528, 163 USPQ 644 (CCPA 1969).
A timely filed terminal disclaimer in compliance with 37 CFR 1.321(c) or 1.321(d) may be used to overcome an actual or provisional rejection based on nonstatutory double patenting provided the reference application or patent either is shown to be commonly owned with the examined application, or claims an invention made as a result of activities undertaken within the scope of a joint research agreement. See MPEP § 717.02 for applications subject to examination under the first inventor to file provisions of the AIA as explained in MPEP § 2159. See MPEP § 2146 et seq. for applications not subject to examination under the first inventor to file provisions of the AIA . A terminal disclaimer must be signed in compliance with 37 CFR 1.321(b).
The filing of a terminal disclaimer by itself is not a complete reply to a nonstatutory double patenting (NSDP) rejection. A complete reply requires that the terminal disclaimer be accompanied by a reply requesting reconsideration of the prior Office action. Even where the NSDP rejection is provisional the reply must be complete. See MPEP § 804, subsection I.B.1. For a reply to a non-final Office action, see 37 CFR 1.111(a). For a reply to final Office action, see 37 CFR 1.113(c). A request for reconsideration while not provided for in 37 CFR 1.113(c) may be filed after final for consideration. See MPEP §§ 706.07(e) and 714.13.
The USPTO Internet website contains terminal disclaimer forms which may be used. Please visit www.uspto.gov/patent/patents-forms. The actual filing date of the application in which the form is filed determines what form (e.g., PTO/SB/25, PTO/SB/26, PTO/AIA /25, or PTO/AIA /26) should be used. A web-based eTerminal Disclaimer may be filled out completely online using web-screens. An eTerminal Disclaimer that meets all requirements is auto-processed and approved immediately upon submission. For more information about eTerminal Disclaimers, refer to www.uspto.gov/patents/apply/applying-online/eterminal-disclaimer.
Claim 1-28 provisionally rejected under 35 U.S.C. 101 as claiming the same invention as that of claim 1-28 of copending Application No. 17/952,935 (reference application). Although the claims at issue are not identical, they are not patentably distinct from each other because The co-pending claim discloses all the limitation of the claims 1-28 of in instant application. The only different between 1 of the co-pending application and 1 of the instant application is that 1 of the instant application discloses in claim 14 of the co-pending application which cites “The method of claim 1, wherein the pretraining task involves generating unitary matrices, circuits, or microwave pulses.” However, this discloses in claim 1 of the co-pending application which cites “the training samples for training samples for the pretraining task involve unitary matrices, circuits, or microwave pulses for a family of quantum devices.”. Thus claim 1 and 4 of instant application is anticipated by claims 1 of the co-pending application. Additionally, the co-pending application discloses the remaining claim 2-28 of the instant application. See the chart below for claim mapping between the two applications.
17/935,284
17/952,935
1. A computer-implemented method comprising:
1. A computer-implement method comprising:
automatically generating training samples efficiently for a general computational task and a family of quantum devices; and
14. (Original) The method of claim 1, wherein the pretraining task involves generating unitary matrices, circuits, or microwave pulses.
obtaining training samples by automatically generating said samples for a pretraining task involving unitary matrices, circuits, or microwave pulses for a family of quantum devices;
pretraining a quantum foundation model to embed information about physical characteristics of a family of quantum devices and a general computational task
pretraining a quantum foundation model with training samples to embed information about the family of quantum devices and pretraining task;
fine-tuning the model with a specialized dataset to perform a specific computational task on a specific quantum computational device; and
fine-tuning the model with a specialized dataset to perform a specific computational task on a specific quantum computational device; and
using the fine-tuned model to generate inputs configured to reduce a metric selected from the group consisting of gate execution times, cross-talk, leakage, gate errors on the quantum device, and number of gates needed to execute an algorithm on the quantum device.
using the fine-tuned model to generate inputs to the quantum device, wherein the generated inputs reduce gate execution time, reduce gate errors, reduce cross-talk, reduce leakage, or reduce the number of gates needed to execute an algorithm on the quantum device.
2. (Currently Amended) The method of claim 1, wherein the quantum foundation model further comprises:
pretraining with a process that uses a generative adversarial model to evaluate the quality of states, wherein the evaluation is based on a generator output produced by a generator network.
2. (Currently Amended) The method of claim 1, wherein the quantum foundation model further comprises:
Pretraining with a process that uses a generative adversarial model, the generative adversarial model comprising a generator and a discriminator, to evaluate the quality of states.
The co-pending applicant cites that the generative adversarial model comprising a generator and a discriminator which is not cited in the co-pending claim 2. However, all generative adversarial networks (GANs) consist of a generator, which creates data samples and discriminator that evaluates them. Thus, it is inherent to GANs.
3. (Currently Amended) The method of claim 2, further comprising: pretraining the foundation model with a structure that uses classical simulators to transform the generator output of the generative adversarial model into a quantum state.
3. (Original) The method of claim 2, further comprising: pretraining the foundation model with a structure that uses classical simulators to transform the generator output into a quantum state.
4. (Currently Amended) The method of claim 3, further comprising: simulating quantum systems classically with different noise parameters, including
4. (Original) The method of claim 3, further comprising: simulating quantum systems classically with different noise parameters, including but not limited to gate errors and thermal relaxation times.
5. (Currently Amended) The method of claim 1, wherein an output of the quantum foundation model is used without fine-tuning.
5. (Currently Amended) The method of claim 1, wherein an output generated by the quantum foundation model after pretraining is used without fine-tuning.
6. (Original) The method of claim 1, wherein the quantum foundation model is not fine-tuned and its output is directly input into a specialized model.
6. (Original) The method of claim 1, wherein the quantum foundation model is not fine-tuned and its output is directly input into a specialized model.
7. (Original) The method of claim 1, wherein the generated quantum device inputs are a sequence of gates, a sequence of microwave pulses, or a sequence of unitary operations.
7. (Original) The method of claim 1, wherein the generated quantum device inputs are a sequence of gates, a sequence of microwave pulses, or a sequence of unitary operations.
8. (Original) The method of claim 1, wherein the quantum foundation model has a neural network architecture.
8. (Original) The method of claim 1, wherein the quantum foundation model has a neural network architecture.
9. (Original) The method of claim 1, wherein the quantum foundation model uses a transformer model architecture.
9. (Original) The method of claim 1, wherein the quantum foundation model uses a transformer model architecture.
10. (Original) The method of claim 1, wherein the pretraining sample is prepared by generating random unitary matrices or random gate sequences.
10. (Original) The method of claim 1, wherein the pretraining sample is prepared by generating random unitary matrices or random gate sequences.
11. (Original) The method of claim 1, wherein the pretraining sample is prepared by generating random microwave pulses and simulating them classically.
11. (Original) The method of claim 1, wherein the pretraining sample is prepared by generating random microwave pulses and simulating them classically.
12. (Original) The method of claim 1, wherein the target family of quantum devices are superconducting circuits, ion traps, quantum annealers, or Boson samplers.
12. (Original) The method of claim 1, wherein the target family of quantum devices are superconducting circuits, ion traps, quantum annealers, or Boson samplers.
13. (Original) The method of claim 1, wherein the target family of quantum devices are universal quantum computers.
13. (Original) The method of claim 1, wherein the target family of quantum devices are universal quantum computers.
14. (Original) The method of claim 1, wherein the pretraining task involves generating unitary matrices, circuits, or microwave pulses.
14. (Original) The method of claim 1, wherein the pretraining task involves generating unitary matrices, circuits, or microwave pulses.
15. A system comprising:
15. A system comprising:
a processor; and
a classical processor; and
a non-transitory computer-readable medium storing instructions that, when executed by the processor, cause the system to perform operations comprising:
a non-transitory computer-readable medium storing instructions that, when executed by the classical processor, cause the system to perform operations comprising:
automatically generating training samples efficiently for a general computational task and a family of quantum devices: and
Obtaining training samples by automatically generating said samples for a pretraining task involving unitary matrices, circuits, or microwave pulses for a family of quantum devices;
pretraining a quantum foundation model to embed information about physical characteristics of a family of quantum devices and a general computational task
pretraining a quantum foundation model with the training samples to embed information about the family of quantum devices and the pretraining task;
fine-tuning the model with a specialized dataset to perform a specific computational task on a specific computational device; and
fine-tuning the model with a specialized dataset to perform a specific computational task on a specific computational device; and
using the fine-tuned model to generate inputs configured to reduce a metric selected from the group consisting of gate execution times, cross-talk, leakage, gate errors, and number of gates needed to execute an algorithm on the quantum device.
using the fine-tuned model to generate higher quality inputs to the quantum device, wherein the generated inputs reduce gate execution time, reduce gate errors, reduce cross-talk, reduce leakage, or reduce the number of gates needed to execute an algorithm on the quantum device.
16. (Currently Amended) The system of claim 15, wherein the quantum foundation model further comprises: pretraining with a process that uses a generative adversarial model to evaluate the quality of states, wherein the evaluation is based on a generator output produced by a generator network.
16. The system of claim 15, wherein the quantum foundation model further comprises:
Pretraining with a process that uses a generative adversarial model, the generative adversarial model comprising a generator and a discriminator, to evaluate the quality of states.
17. (Currently Amended) The system of claim 16, further comprising: pretraining the foundation model with a structure that uses classical simulators to transform the generator output of the generative adversarial model- into a quantum state.
17. (Original) The system of claim 16, further comprising: pretraining the foundation model with a structure that uses classical simulators to transform the generator output into a quantum state.
18. (Currently Amended) The system of claim 17, further comprising: simulating quantum systems classically with different noise parameters, including gate errors and thermal relaxation times.
18. (Original) The system of claim 17, further comprising: simulating quantum systems classically with different noise parameters, including but not limited to gate errors and thermal relaxation times.
19. (Currently Amended) The system of claim 15, wherein an output of the quantum foundation model is used without fine-tuning.
19. The system of claim 15, wherein an output generated by the quantum foundation model after pretraining is used without fine-tuning.
20. (Original) The system of claim 15, wherein the quantum foundation model is not fine-tuned and its output is directly input into a specialized model.
20. (Original) The system of claim 15, wherein the quantum foundation model is not fine-tuned and its output is directly input into a specialized model.
21. (Original) The system of claim 15, wherein the generated quantum device inputs are a sequence of gates, a sequence of microwave pulses, or a sequence of unitary operations.
21. (Original) The system of claim 15, wherein the generated quantum device inputs are a sequence of gates, a sequence of microwave pulses, or a sequence of unitary operations.
22. (Original) The system of claim 15, wherein the quantum foundation model has a neural network architecture.
22. (Original) The system of claim 15, wherein the quantum foundation model has a neural network architecture.
23. (Original) The system of claim 15, wherein the quantum foundation model uses a transformer model architecture.
23. (Original) The system of claim 15, wherein the quantum foundation model uses a transformer model architecture.
24. (Original) The system of claim 15, wherein the pretraining sample is prepared by generating random unitary matrices or random gate sequences.
24. (Original) The system of claim 15, wherein the pretraining sample is prepared by generating random unitary matrices or random gate sequences.
25. (Original) The system of claim 15, wherein the pretraining sample is prepared by generating random microwave pulses and simulating them classically.
25. (Original) The system of claim 15, wherein the pretraining sample is prepared by generating random microwave pulses and simulating them classically.
26. (Original) The system of claim 15, wherein the target family of quantum devices are superconducting circuits, ion traps, quantum annealers, or Boson samplers.
26. (Original) The system of claim 15, wherein the target family of quantum devices are superconducting circuits, ion traps, quantum annealers, or Boson samplers.
27. (Original) The system of claim 15, wherein the target family of quantum devices are universal quantum computers.
27. (Original) The system of claim 15, wherein the target family of quantum devices are universal quantum computers.
28. (Original) The system of claim 15, wherein the pretraining task involves generating unitary matrices, circuits, or microwave pulses.
28. (Original) The system of claim 15, wherein the pretraining task involves generating unitary matrices, circuits, or microwave pulses.
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, 5-8, 15, and 19-22 are rejected under 35 U.S.C. 103 as being unpatentable over Andrea Mari et al; “Transfer learning in hybrid classical-quantum neural networks” (hereinafter “Andrea”) in view of Zachary Albert Mayer et al; US 20210287089 A1 (hereinafter “Mayer”) in further view of Steven Thomas Flammia; US 12488170 B1 (hereinafter “Steven”).
Regarding claim 1, Andrea teaches A computer-implemented method comprising: [automatically generating training samples efficiently for a general computational task and] a family of quantum devices; and (Andrea Page 11 Paragraph 5; " From our theoretical and experimental analysis, we can conclude that transfer learning is a promising approach which can be particularly convenient in the context of near-term quantum devices.” Examiner notes that a family of quantum devices (near-term quantum devices) is present)
pretraining a quantum foundation model to embed information about physical characteristics of a family of quantum devices and a general computational task (Andrea Page 4 Paragraph 1; "1. Take a network A that has been pre-trained on a dataset DA and for a given task TA." Andrea Page 5 Paragraph 6; "In this case a quantum network A is pre-trained for a generic task and dataset" Andrea Page 2 Paragraph 7; " This fact is due to the fundamental unitary nature of quantum mechanics and, as discussed at the end of this section, should be taken into account when designing quantum networks" Examiner notes that quantum network A/quantum foundation model is pretrained to embed information about a physical characteristics of a family of quantum devices/fundamental unitary nature of quantum mechanics being taken account of and a general computational task/given task TA, rather than a narrow task aimed at a specific computational end)
fine-tuning the model with a specialized dataset to perform a specific computational task on a specific quantum computational device; and (Andrea Page 4 Paragraph 1; "2. Remove some of the final layers. In this way, the resulting truncated network A0 can be used as a feature extractor. 3. Connect a new trainable network B at the end of the pre-trained network A0. 4. Keep the weights of A0 constant, and train the final block B with a new dataset DB and for a new task of interest TB." Andrea Page 5 Paragraph 6; "Successively, some of the final quantum layers are removed, and replaced by a trainable quantum network B which will be optimized for a specific problem" Examiner notes that steps 2-4 is fine tuning the model with a specialized dataset/dataset DB to perform a specific computational task/new task of interest TB on a specific quantum computational device/network B)
using the fine-tuned model to generate inputs [configured to reduce a metric selected from the group consisting of gate execution times, cross-talk, leakage, gate errors on the quantum device, and number of gates needed to execute an algorithm on the quantum device.] (Andrea Page 6 Paragraph 1; “K is an entangling unitary operation made of three controlled NOT gates” Andrea Page 10 Paragraph 2; "the model can generate any single-mode unitary operation" Examiner notes that model/fine-tuned model generates inputs/single-mode unitary operation configured to )
Andrea does not teach A computer-implemented method comprising: automatically generating training samples efficiently for a general computational task [and a family of quantum devices]; and
However, Mayer does teach A computer-implemented method comprising: automatically generating training samples efficiently for a general computational task [and a family of quantum devices]; and (Mayer Paragraph 0082; “Raw training data 210 (e.g., tabular data having rows and columns) is provided to a pre-processing module 212. The pre-processing module 212 can perform one or more data processing operations on the raw training data 210 to generate a processed training data” Examiner notes raw training data is automatically processed to generating training samples efficiently (processed training data) for a general computational task (for training the neural network))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea and Mayer. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. One of ordinary skill would have motivation to combine Andrea and Mayer to make a more efficient and accurate neural network model “Such pre-processing techniques can reduce computation times required for training and making predictions, and can result in more efficient and accurate neural network models.” (Mayer Paragraph 0007).
Andrea in view of Mayer does not teach inputs configured to reduce a metric selected from the group consisting of gate execution times, cross-talk, leakage, gate errors on the quantum device, and number of gates needed to execute an algorithm on the quantum device.
However, Flammia does teach inputs configured to reduce a metric selected from the group consisting of gate execution times, cross-talk, leakage, gate errors on the quantum device, and number of gates needed to execute an algorithm on the quantum device. (Flammia Column 16 Line 6; “recommend improvements to a submitted quantum circuit to reduce logical errors resulting from gate errors” Examiner notes that inputs (recommend improvements to a submitted quantum circuit) is configured to reduce gate errors on the quantum device (reduce logical errors resulting from gate errors))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, and Flammia. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. One of ordinary skill would have motivation to combine Andrea, Mayer, and Flammia to make more fault tolerant quantum computation “Estimating errors in quantum computers enables progress towards fault tolerant quantum computation.” (Flammia Column 3 Line 1).
Regarding claim 5, Andrea teaches The method of claim 1, wherein an output of the quantum foundation model is used without fine-tuning. (Andrea Page 11 Paragraph 1; "We compare the results obtained with and without the pre-trained layer A0 (i.e., with and without transfer learning), for a fixed total depth of 3 or 4 layers. It is clear that the QQ transfer learning approach offers a strong advantage in terms of training efficiency." Examiner notes the output of the quantum foundation model is used without fine-tuning the model/without the pre-trained layer)
Regarding claim 6, Andera teaches The method of claim 1, wherein the quantum foundation model is not fine-tuned and its output is directly input into a specialized model. (Andrea Page 11 Paragraph 1; "In Fig. 9 we plot the loss function (cross entropy) of our quantum variational classifier with respect to the number of training iterations. We compare the results obtained with and without the pre-trained layer A0 (i.e., with and without transfer learning), for a fixed total depth of 3 or 4 layers." Examiner notes the loss function is the specialized model)
Regarding claim 7, Andrea teaches The method of claim 1, wherein the generated quantum device inputs are a sequence of gates, a sequence of microwave pulses, or a sequence of unitary operations. (Andrea Page 10 Paragraph 2; "the model can generate any single-mode unitary operation")
Regarding claim 8, Andrea teaches The method of claim 1, wherein the quantum foundation model has a neural network architecture. (Andrea Page 2 Paragraph 8; "One of the possible quantum generalizations of feedforward neural networks can be given in terms of variational quantum circuits")
Regarding claim 15, Andrea teaches [automatically generating training samples efficiently for a general computational task and] a family of quantum devices; and (Andrea Page 11 Paragraph 5; " From our theoretical and experimental analysis, we can conclude that transfer learning is a promising approach which can be particularly convenient in the context of near-term quantum devices.” Examiner notes that a family of quantum devices (near-term quantum devices) is present)
pretraining a quantum foundation model to embed information about physical characteristics of a family of quantum devices and a general computational task (Andrea Page 4 Paragraph 1; "1. Take a network A that has been pre-trained on a dataset DA and for a given task TA." Andrea Page 5 Paragraph 6; "In this case a quantum network A is pre-trained for a generic task and dataset" Andrea Page 2 Paragraph 7; " This fact is due to the fundamental unitary nature of quantum mechanics and, as discussed at the end of this section, should be taken into account when designing quantum networks" Examiner notes that quantum network A/quantum foundation model is pretrained to embed information about a physical characteristics of a family of quantum devices/fundamental unitary nature of quantum mechanics being taken account of and a general computational task/given task TA, rather than a narrow task aimed at a specific computational end)
fine-tuning the model with a specialized dataset to perform a specific computational task on a specific quantum computational device; (Andrea Page 4 Paragraph 1; "2. Remove some of the final layers. In this way, the resulting truncated network A0 can be used as a feature extractor. 3. Connect a new trainable network B at the end of the pre-trained network A0. 4. Keep the weights of A0 constant, and train the final block B with a new dataset DB and for a new task of interest TB." Andrea Page 5 Paragraph 6; "Successively, some of the final quantum layers are removed, and replaced by a trainable quantum network B which will be optimized for a specific problem" Examiner notes that steps 2-4 is fine tuning the model with a specialized dataset/dataset DB to perform a specific computational task/new task of interest TB on a specific quantum computational device/network B)
using the fine-tuned model to generate inputs [configured to reduce a metric selected from the group consisting of gate execution times, cross-talk, leakage, gate errors on the quantum device, and number of gates needed to execute an algorithm on the quantum device.] (Andrea Page 6 Paragraph 1; “K is an entangling unitary operation made of three controlled NOT gates” Andrea Page 10 Paragraph 2; "the model can generate any single-mode unitary operation" Examiner notes that model/fine-tuned model generates inputs/single-mode unitary operation configured to )
Andrea does not teach A system comprising: a processor; and a non-transitory computer-readable medium storing instructions that, when executed by the processor, cause the system to perform operations comprising:
automatically generating training samples efficiently for a general computational task [and a family of quantum devices]; and
However, Mayer does teach A system comprising: a processor; and a non-transitory computer-readable medium storing instructions that, when executed by the processor, cause the system to perform operations comprising: (Mayer Paragraph 0014; “the subject matter described in this specification can be embodied in a non-transitory computer-readable medium having instructions stored thereon that, when executed by one or more computer processors, cause the one or more computer processors to perform operations”)
automatically generating training samples efficiently for a general computational task [and a family of quantum devices]; and (Mayer Paragraph 0082; “Raw training data 210 (e.g., tabular data having rows and columns) is provided to a pre-processing module 212. The pre-processing module 212 can perform one or more data processing operations on the raw training data 210 to generate a processed training data” Examiner notes raw training data is automatically processed to generating training samples efficiently (processed training data) for a general computational task (for training the neural network))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea and Mayer. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. One of ordinary skill would have motivation to combine Andrea and Mayer to make a more efficient and accurate neural network model “Such pre-processing techniques can reduce computation times required for training and making predictions, and can result in more efficient and accurate neural network models.” (Mayer Paragraph 0007).
Andrea in view of Mayer does not teach inputs configured to reduce a metric selected from the group consisting of gate execution times, cross-talk, leakage, gate errors on the quantum device, and number of gates needed to execute an algorithm on the quantum device.
However, Flammia does teach inputs configured to reduce a metric selected from the group consisting of gate execution times, cross-talk, leakage, gate errors on the quantum device, and number of gates needed to execute an algorithm on the quantum device. (Flammia Column 16 Line 6; “recommend improvements to a submitted quantum circuit to reduce logical errors resulting from gate errors” Examiner notes that inputs (recommend improvements to a submitted quantum circuit) is configured to reduce gate errors on the quantum device (reduce logical errors resulting from gate errors))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, and Flammia. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. One of ordinary skill would have motivation to combine Andrea, Mayer, and Flammia to make more fault tolerant quantum computation “Estimating errors in quantum computers enables progress towards fault tolerant quantum computation.” (Flammia Column 3 Line 1).
Regarding claim 19, Andrea teaches The system of claim 15, wherein an output of the quantum foundation model is used without fine-tuning. (Andrea Page 11 Paragraph 1; "We compare the results obtained with and without the pre-trained layer A0 (i.e., with and without transfer learning), for a fixed total depth of 3 or 4 layers. It is clear that the QQ transfer learning approach offers a strong advantage in terms of training efficiency." Examiner notes the output of the quantum foundation model is used without fine-tuning the model/without the pre-trained layer)
Regarding claim 20, Andera teaches The system of claim 15, wherein the quantum foundation model is not fine-tuned and its output is directly input into a specialized model. (Andrea Page 11 Paragraph 1; "In Fig. 9 we plot the loss function (cross entropy) of our quantum variational classifier with respect to the number of training iterations. We compare the results obtained with and without the pre-trained layer A0 (i.e., with and without transfer learning), for a fixed total depth of 3 or 4 layers." Examiner notes the loss function is the specialized model)
Regarding claim 21, Andrea teaches The system of claim 15, wherein the generated quantum device inputs are a sequence of gates, a sequence of microwave pulses, or a sequence of unitary operations. (Andrea Page 10 Paragraph 2; "the model can generate any single-mode unitary operation")
Regarding claim 22, Andrea teaches The system of claim 15, wherein the quantum foundation model has a neural network architecture. (Andrea Page 2 Paragraph 8; "One of the possible quantum generalizations of feedforward neural networks can be given in terms of variational quantum circuits")
Claim(s) 2 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Andrea Mari et al; “Transfer learning in hybrid classical-quantum neural networks” (hereinafter “Andrea”) in view of Zachary Albert Mayer et al; US 20210287089 A1 (hereinafter “Mayer”) in further view of Steven Thomas Flammia; US 12488170 B1 (hereinafter “Steven”) in view further of Jason Brownlee; “A Gentle Introduction to Generative Adversarial Networks (GANs)” (hereinafter “Jason”)
Regarding claim 2, Andrea does not teach The method of claim 1, wherein the quantum foundation model further comprises: pretraining with a process that uses a generative adversarial model to evaluate the quality of states, wherein the evaluation is based on a generator output produced by a generator network.
However, Jason does teach The method of claim 1, wherein the quantum foundation model further comprises: pretraining with a process that uses a generative adversarial model to evaluate the quality of states, wherein the evaluation is based on a generator output produced by a generator network. (Jason Section "GANs as a Two Player Game" Paragraph 2; "The generator generates a batch of samples, and these, along with real examples from the domain, are provided to the discriminator and classified as real or fake. The discriminator is then updated to get better at discriminating real and fake samples in the next round, and importantly, the generator is updated based on how well, or not, the generated samples fooled the discriminator." Examiner notes that the discriminator part of the generative adversarial model is evaluating the quality of states/samples; evaluation is based on a generator output (generator generates a batch of samples) produced by a generator network (GAN))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Jason. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Jason teaches using GANs to train a generative model. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Jason to pretrain the quantum foundation model using a GAN for quality “GANs are a clever way of training a generative model by framing the problem as a supervised learning problem with two sub-models: the generator model that we train to generate new examples, and the discriminator model that tries to classify examples as either real (from the domain) or fake (generated).” (Jason Section Introduction Paragraph 3).
Regarding claim 16, Andrea does not teach The system of claim 15, wherein the quantum foundation model further comprises: pretraining with a process that uses a generative adversarial model to evaluate the quality of states, wherein the evaluation is based on a generator output produced by a generator network.
However, Jason does teach The system of claim 15, wherein the quantum foundation model further comprises: pretraining with a process that uses a generative adversarial model to evaluate the quality of states, wherein the evaluation is based on a generator output produced by a generator network. (Jason Section "GANs as a Two Player Game" Paragraph 2; "The generator generates a batch of samples, and these, along with real examples from the domain, are provided to the discriminator and classified as real or fake. The discriminator is then updated to get better at discriminating real and fake samples in the next round, and importantly, the generator is updated based on how well, or not, the generated samples fooled the discriminator." Examiner notes that the discriminator part of the generative adversarial model is evaluating the quality of states/samples; evaluation is based on a generator output (generator generates a batch of samples) produced by a generator network (GAN))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Jason. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Jason teaches using GANs to train a generative model. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Jason to pretrain the quantum foundation model using a GAN for quality “GANs are a clever way of training a generative model by framing the problem as a supervised learning problem with two sub-models: the generator model that we train to generate new examples, and the discriminator model that tries to classify examples as either real (from the domain) or fake (generated).” (Jason Section Introduction Paragraph 3).
Claim(s) 3, 4, 17, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Andrea Mari et al; “Transfer learning in hybrid classical-quantum neural networks” (hereinafter “Andrea”) in view of Zachary Albert Mayer et al; US 20210287089 A1 (hereinafter “Mayer”) in further view of Steven Thomas Flammia; US 12488170 B1 (hereinafter “Steven”) in view further of Jason Brownlee; “A Gentle Introduction to Generative Adversarial Networks (GANs)” (hereinafter “Jason”) in further view of Zuccarelli; Luigi et al; US 20200117764 A1 (hereinafter “Luigi”).
Regarding claim 3, Andrea does not teach The method of claim 2, further comprising: pretraining the foundation model with a structure that uses classical simulators to transform the generator output into a quantum state.
However, Luigi does teach The method of claim 2, further comprising: pretraining the foundation model with a structure that uses classical simulators to transform the generator output of the generative adversarial model into a quantum state. (Luigi Paragraph 0002; "quantum computers process data in units of quantum bits (qubits) that can be in superpositions of states. “Superposition” refers to the ability of each qubit to represent both a “1” and a “0” at the same time." Luigi Paragraph 0011; "quantum simulators are classical computers that attempt to mimic the functioning of a quantum computer." Examiner notes that classical computers mimic/transform the generator output from the generative adversarial model into a quantum state.)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, Jason, and Luigi. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Jason teaches using GANs to train a generative model. Luigi teaches simulating quantum computers with classical computers. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, Jason, and Luigi to simulate quantum computers for safely experimenting in quantum-like environments without consuming large resources “Because quantum simulators run on classical computer hardware, users may safely experiment with different input algorithms in a quantum-like environment without consuming valuable quantum computing resources.” (Luigi Paragraph 0011).
Regarding claim 4, Andrea does not teach The method of claim 3, further comprising: simulating quantum systems classically with different noise parameters, including gate errors and thermal relaxation times.
However, Luigi does teach The method of claim 3, further comprising: simulating quantum systems classically with different noise parameters, including gate errors and thermal relaxation times. (Luigi Paragraph 0016; "the developer may be guided to apply different types or magnitudes of errors such as gate errors, readout errors, multiqubit errors, or other types of metadata that may affect a quantum computer's results.")
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, Jason, and Luigi. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Jason teaches using GANs to train a generative model. Luigi teaches simulating quantum computers with classical computers. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, Jason, and Luigi to simulate quantum computers for safely experimenting in quantum-like environments without consuming large resources “Because quantum simulators run on classical computer hardware, users may safely experiment with different input algorithms in a quantum-like environment without consuming valuable quantum computing resources.” (Luigi Paragraph 0011).
Regarding claim 17, Andrea does not teach The system of claim 16, further comprising: pretraining the foundation model with a structure that uses classical simulators to transform the generator output of the generative adversarial model- into a quantum state.
However, Luigi does teach The system of claim 16, further comprising: pretraining the foundation model with a structure that uses classical simulators to transform the generator output of the generative adversarial model- into a quantum state. (Luigi Paragraph 0002; "quantum computers process data in units of quantum bits (qubits) that can be in superpositions of states. “Superposition” refers to the ability of each qubit to represent both a “1” and a “0” at the same time." Luigi Paragraph 0011; "quantum simulators are classical computers that attempt to mimic the functioning of a quantum computer." Examiner notes that classical computers mimic/transform the generator output from the generative adversarial model into a quantum state.)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, Jason, and Luigi. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Jason teaches using GANs to train a generative model. Luigi teaches simulating quantum computers with classical computers. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, Jason, and Luigi to simulate quantum computers for safely experimenting in quantum-like environments without consuming large resources “Because quantum simulators run on classical computer hardware, users may safely experiment with different input algorithms in a quantum-like environment without consuming valuable quantum computing resources.” (Luigi Paragraph 0011).
Regarding claim 18, Andrea does not teach The system of claim 17, further comprising: simulating quantum systems classically with different noise parameters, including gate errors and thermal relaxation times.
However, Luigi does teach The system of claim 17, further comprising: simulating quantum systems classically with different noise parameters, including to gate errors and thermal relaxation times. (Luigi Paragraph 0016; "the developer may be guided to apply different types or magnitudes of errors such as gate errors, readout errors, multiqubit errors, or other types of metadata that may affect a quantum computer's results.")
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, Jason, and Luigi. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Jason teaches using GANs to train a generative model. Luigi teaches simulating quantum computers with classical computers. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, Jason, and Luigi to simulate quantum computers for safely experimenting in quantum-like environments without consuming large resources “Because quantum simulators run on classical computer hardware, users may safely experiment with different input algorithms in a quantum-like environment without consuming valuable quantum computing resources.” (Luigi Paragraph 0011).
Claim(s) 9 and 23 are rejected under 35 U.S.C. 103 as being unpatentable over Andrea Mari et al; “Transfer learning in hybrid classical-quantum neural networks” (hereinafter “Andrea”) in view of Zachary Albert Mayer et al; US 20210287089 A1 (hereinafter “Mayer”) in further view of Steven Thomas Flammia; US 12488170 B1 (hereinafter “Steven”) in view further of Juan Carrasquilla; “Probabilistic Simulation of Quantum Circuits with the Transformer” (hereinafter “Juan”).
Regarding claim 9, Andrea does not teach The method of claim 1, wherein the quantum foundation model uses a transformer model architecture.
However, Juan does teach The method of claim 1, wherein the quantum foundation model uses a transformer model architecture. (Juan Page 5 Paragraph 1; "The Transformer architecture is constructed using the elements depicted in Fig. 3. The first and most important element is the self-attention mechanism. Self-attention takes an embedding of the measurement outcome a, and computes an auto-correlation matrix where the different measurement outcomes across the different qubits form the columns and rows. The embedding is simply a linear transformation on the original input a. The self-attention and its correlation matrix are useful to introduce correlations between qubits separated at any distance in the quantum system" Examiner notes the quantum system/quantum foundation model uses a transformer model architecture)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Juan. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Juan teaches using a transformer architecture for a quantum system. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Juan to better represent the correlations between qubits in quantum systems “In contrast to traditional sequence models based on recurrent neural networks, which tend to suppress correlations beyond a certain length ξ, the self-attention networks are suitable to model systems exhibiting power law correlations present in natural sequences as well as physical systems exhibiting (classical or quantum) critical behavior.” (Juan Page 5 Paragraph 1).
Regarding claim 23, Andrea does not teach The system of claim 15, wherein the quantum foundation model uses a transformer model architecture.
However, Juan does teach The system of claim 15, wherein the quantum foundation model uses a transformer model architecture. (Juan Page 5 Paragraph 1; "The Transformer architecture is constructed using the elements depicted in Fig. 3. The first and most important element is the self-attention mechanism. Self-attention takes an embedding of the measurement outcome a, and computes an auto-correlation matrix where the different measurement outcomes across the different qubits form the columns and rows. The embedding is simply a linear transformation on the original input a. The self-attention and its correlation matrix are useful to introduce correlations between qubits separated at any distance in the quantum system" Examiner notes the quantum system/quantum foundation model uses a transformer model architecture)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Juan. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Juan teaches using a transformer architecture for a quantum system. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Juan to better represent the correlations between qubits in quantum systems “In contrast to traditional sequence models based on recurrent neural networks, which tend to suppress correlations beyond a certain length ξ, the self-attention networks are suitable to model systems exhibiting power law correlations present in natural sequences as well as physical systems exhibiting (classical or quantum) critical behavior.” (Juan Page 5 Paragraph 1).
Claim(s) 10, 14, 24, and 28 are rejected under 35 U.S.C. 103 as being unpatentable over Andrea Mari et al; “Transfer learning in hybrid classical-quantum neural networks” (hereinafter “Andrea”) in view of Zachary Albert Mayer et al; US 20210287089 A1 (hereinafter “Mayer”) in further view of Steven Thomas Flammia; US 12488170 B1 (hereinafter “Steven”) in view further of LOW; Guang Hao; US 20240046137 A1 (hereinafter “Guang”).
Regarding claim 10, Andrea does not teach The method of claim 1, wherein the pretraining sample is prepared by generating random unitary matrices or random gate sequences.
However, Guang does teach The method of claim 1, wherein the pretraining sample is prepared by generating random unitary matrices or random gate sequences. (Guang Paragraph 0021; "The processor 22 of the classical computing device 20, as shown in FIG. 1, generates a Haar-random unitary matrix u with dimension n×n.")
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Guang. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Guang teaches generating random unitary matrices. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Guang to generate random unitary matrices for more efficient computation “According to this aspect, the Haar-random unitary matrix may have a dimension n×n, where n is a number of modes for which the fermion wavefunction is specified. The above features may have the technical effect of allowing the processor to efficiently compute the single-particle basis fermion rotation.” (Guang Paragraph 0082).
Regarding claim 14, Andrea does not teach The method of claim 1, wherein the pretraining task involves generating unitary matrices, circuits, or microwave pulses.
However, Guang does teach The method of claim 1, wherein the pretraining task involves generating unitary matrices, circuits, or microwave pulses. (Guang Paragraph 0021; "The processor 22 of the classical computing device 20, as shown in FIG. 1, generates a Haar-random unitary matrix u with dimension n×n.")
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Guang. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Guang teaches generating random unitary matrices. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Guang to generate random unitary matrices for more efficient computation “According to this aspect, the Haar-random unitary matrix may have a dimension n×n, where n is a number of modes for which the fermion wavefunction is specified. The above features may have the technical effect of allowing the processor to efficiently compute the single-particle basis fermion rotation.” (Guang Paragraph 0082).
Regarding claim 24, Andrea does not teach The system of claim 15, wherein the pretraining sample is prepared by generating random unitary matrices or random gate sequences.
However, Guang does teach The system of claim 15, wherein the pretraining sample is prepared by generating random unitary matrices or random gate sequences. (Guang Paragraph 0021; "The processor 22 of the classical computing device 20, as shown in FIG. 1, generates a Haar-random unitary matrix u with dimension n×n.")
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Guang. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Guang teaches generating random unitary matrices. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Guang to generate random unitary matrices for more efficient computation “According to this aspect, the Haar-random unitary matrix may have a dimension n×n, where n is a number of modes for which the fermion wavefunction is specified. The above features may have the technical effect of allowing the processor to efficiently compute the single-particle basis fermion rotation.” (Guang Paragraph 0082).
Regarding claim 28, Andrea does not teach The system of claim 15, wherein the pretraining task involves generating unitary matrices, circuits, or microwave pulses.
However, Guang does teach The system of claim 15, wherein the pretraining task involves generating unitary matrices, circuits, or microwave pulses. (Guang Paragraph 0021; "The processor 22 of the classical computing device 20, as shown in FIG. 1, generates a Haar-random unitary matrix u with dimension n×n.")
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Guang. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Guang teaches generating random unitary matrices. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Guang to generate random unitary matrices for more efficient computation “According to this aspect, the Haar-random unitary matrix may have a dimension n×n, where n is a number of modes for which the fermion wavefunction is specified. The above features may have the technical effect of allowing the processor to efficiently compute the single-particle basis fermion rotation.” (Guang Paragraph 0082).
Claim(s) 11 and 25 are rejected under 35 U.S.C. 103 as being unpatentable over Andrea Mari et al; “Transfer learning in hybrid classical-quantum neural networks” (hereinafter “Andrea”) in view of Zachary Albert Mayer et al; US 20210287089 A1 (hereinafter “Mayer”) in further view of Steven Thomas Flammia; US 12488170 B1 (hereinafter “Steven”) in view further of Wu; Xiaodong et al; “US 20090160698 A1” (hereinafter “Xiaodong”).
Regarding claim 11, Andrea does not teach The method of claim 1, wherein the pretraining sample is prepared by generating random microwave pulses and simulating them classically.
However, Xiaodong does teach The method of claim 1, wherein the pretraining sample is prepared by generating random microwave pulses and simulating them classically. (Xiaodong Paragraph 0008; "The microwave sensor comprises an oscillator configured to generate a microwave signal, a random pulse generator configured to set a random pulse period for each transmitted microwave signal" Examiner notes the microwave sensor generates random microwave pulses and simulates them classically)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Xiaodong. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Xiaodong teaches generating random microwave pulses classically. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Xiaodong to generate random microwave pulses classically for higher reliability in alarms “By using a random pulse period for each transmitted signal, the issue of overlapping pulse widths is minimized even when the transmitting microwave frequency is the same. The system therefore is more reliable as the occurrence of false alarms is considerably reduced.” (Xiaodong Paragraph 0031).
Regarding claim 25, Andrea does not teach The system of claim 15, wherein the pretraining sample is prepared by generating random microwave pulses and simulating them classically.
However, Xiaodong does teach The system of claim 15, wherein the pretraining sample is prepared by generating random microwave pulses and simulating them classically. (Xiaodong Paragraph 0008; "The microwave sensor comprises an oscillator configured to generate a microwave signal, a random pulse generator configured to set a random pulse period for each transmitted microwave signal" Examiner notes the microwave sensor generates random microwave pulses and simulates them classically)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Xiaodong. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Xiaodong teaches generating random microwave pulses classically. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Xiaodong to generate random microwave pulses classically for higher reliability in alarms “By using a random pulse period for each transmitted signal, the issue of overlapping pulse widths is minimized even when the transmitting microwave frequency is the same. The system therefore is more reliable as the occurrence of false alarms is considerably reduced.” (Xiaodong Paragraph 0031).
Claim(s) 12 and 26 are rejected under 35 U.S.C. 103 as being unpatentable over Andrea Mari et al; “Transfer learning in hybrid classical-quantum neural networks” (hereinafter “Andrea”) in view of Zachary Albert Mayer et al; US 20210287089 A1 (hereinafter “Mayer”) in further view of Steven Thomas Flammia; US 12488170 B1 (hereinafter “Steven”) in view further of Batruni; Roy; US 11663289 B1 (hereinafter “Roy”).
Regarding claim 12, Andrea does not teach The method of claim 1, wherein the target family of quantum devices are superconducting circuits, ion traps, quantum annealers, or Boson samplers.
However, Roy does teach The method of claim 1, wherein the target family of quantum devices are superconducting circuits, ion traps, quantum annealers, or Boson samplers. (Roy Column 5 Paragraph 6; "Qubit representation of data is typically used by quantum computer processors which include structures (e.g., ion traps)")
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Roy. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Roy teaches ion traps for qubit representations. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Roy to use ion traps for a reduction in memory and processing requirements, and data analysis speed “using qubit representation reduces memory and processing requirements, and improves data analysis speed.” (Roy column 5 Line 65).
Regarding claim 26, Andrea does not teach The system of claim 15, wherein the target family of quantum devices are superconducting circuits, ion traps, quantum annealers, or Boson samplers.
However, Roy does teach The system of claim 15, wherein the target family of quantum devices are superconducting circuits, ion traps, quantum annealers, or Boson samplers. (Roy Column 5 Paragraph 6; "Qubit representation of data is typically used by quantum computer processors which include structures (e.g., ion traps)")
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Roy. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Roy teaches ion traps for qubit representations. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Roy to use ion traps for a reduction in memory and processing requirements, and data analysis speed “using qubit representation reduces memory and processing requirements, and improves data analysis speed.” (Roy column 5 Line 65).
Claim(s) 13 and 27 are rejected under 35 U.S.C. 103 as being unpatentable over Andrea Mari et al; “Transfer learning in hybrid classical-quantum neural networks” (hereinafter “Andrea”) in view of Zachary Albert Mayer et al; US 20210287089 A1 (hereinafter “Mayer”) in further view of Steven Thomas Flammia; US 12488170 B1 (hereinafter “Steven”) in view further of Sebastian Horvat; “Universal quantum computation via quantum controlled classical operations” (Hereinafter “Sebastian”).
Regarding claim 13, Andrea does not teach The method of claim 1, wherein the target family of quantum devices are universal quantum computers.
However, Sebastian does teach The method of claim 1, wherein the target family of quantum devices are universal quantum computers. (Sebastian Page 1 Paragraph 3; "in our work we show that a parity computer, when supplemented with an appropriate quantum control, can perform universal quantum computation.")
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Sebastian. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Sebastian teaches universal quantum computers. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Sebastian to utilize universal quantum computers for approximating any other unitary operations "Universal sets of gates for quantum computation are discrete sets of elementary unitary transformations which can be used to approximate any other unitary operation." (Sebastian Page 1 Paragraph 1).
Regarding claim 27, Andrea does not teach The system of claim 15, wherein the target family of quantum devices are universal quantum computers.
However, Sebastian does teach The system of claim 15, wherein the target family of quantum devices are universal quantum computers. (Sebastian Page 1 Paragraph 3; "in our work we show that a parity computer, when supplemented with an appropriate quantum control, can perform universal quantum computation.")
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Andrea, Mayer, Flammia, and Sebastian. Andrea teaches using transfer learning with quantum networks. Mayer teaches a method for developing and using neural network models. Flammia teaches method of estimating noise for individual gates of a quantum circuit uses averaged circuit eigenvalue sampling to determine an overall eigenvalue for the circuit. Sebastian teaches universal quantum computers. One of ordinary skill would have motivation to combine Andrea, Mayer, Flammia, and Sebastian to utilize universal quantum computers for approximating any other unitary operations "Universal sets of gates for quantum computation are discrete sets of elementary unitary transformations which can be used to approximate any other unitary operation." (Sebastian Page 1 Paragraph 1).
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 or earlier communications from the examiner should be directed to DANIEL DUC TRAN whose telephone number is (571)272-6870. The examiner can normally be reached Mon-Fri 8:00-5:00 EST.
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, Viker Lamardo can be reached at (571) 270-5871. 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.
/D.D.T./Examiner, Art Unit 2147
/ERIC NILSSON/Primary Examiner, Art Unit 2151
Prosecution Insights