DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 01/14/2026 has been entered.
Response to Arguments
Applicant's arguments filed 12/15/2025 have been fully considered and they are partially persuasive.
Regarding applicant’s remarks directed to the rejection of claims under 35 USC § 103, the arguments are directed to newly amended limitations that were not previously examined by the examiner. Therefore, applicants arguments are rendered moot. The examiner refers to the rejection under 35 USC § 103 in the current office action for more details.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 16-22 and 36-37 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.
Claims 16-18 and 21-22 recites the limitation "the at least one Bayesian network model arrangement." There is insufficient antecedent basis for this limitation in the claims.
Claims 19-10 and 36-37 are further rejected on virtue of their dependencies to the rejected parent claims.
Claim 22 recites the limitation “the representation of the real physical system.” There is insufficient antecedent basis for this limitation in the claim.
Claim 37 is further rejected on virtue of its dependency to claim 22.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 15-16, 18-20, 22-24, 26-28, 30-32, 34 and 37-38 are rejected under 35 U.S.C. 103 as being unpatentable over Benedetti, Marcello, et al. "Parameterized quantum circuits as machine learning models." arXiv preprint arXiv:1906.07682 (2019) (“Benedetti”) in view of Jordan, Michael I., et al. "An introduction to variational methods for graphical models." Machine learning 37.2 (1999): 183-233. (“Jordan”)
In regards to claim 15,
Benedetti teaches A control system for controlling or monitoring a real physical system using a computing system separate from the real physical system, wherein the computing system comprises a hybrid combination of a classical computer and a quantum computer, wherein the control system is configured to receive input data at the classical computer from the real physical system, wherein the classical computer and the quantum computer are configured to exchange data therebetween,
(Benedetti, Section I., “The general hybrid approach [A control system] is illustrated in Fig. 1 and is made of three main components: the human, the classical computer, and the quantum computer [using a computing system separate from the real physical system, wherein the computing system comprises a hybrid combination of a classical computer and a quantum computer]. The human interprets the problem information and selects an initial model to represent it [monitoring a real physical system; ie problem information from the real physical system (human-provided)]. The data is pre-processed on a classical computer to determine a set of parameters for the PQC [wherein the control system is configured to receive input data at the classical computer from the real physical system]. The quantum hardware prepares a quantum state as prescribed by a PQC and performs measurements. Measurement outcomes are post-processed by the classical computer to generate a forecast. To improve the forecast, the classical computer implements a learning algorithm that updates the model’s parameters. The overall algorithm is run in a closed loop between the classical and quantum hardware [wherein the classical computer and the quantum computer are configured to exchange data therebetween]. The human supervises the process and uses forecasts towards the goal.” See annotated fig. 1
PNG
media_image1.png
504
854
media_image1.png
Greyscale
)
Further, Benedetti discloses successful applications of the hybrid approach such as searching the ground state of the electronic Hamiltonian of molecules
(Benedetti, Section I., “The hybrid approach turned out to be successful in attacking scaled-down problems in chemistry, combinatorial optimization and machine learning. For example, the variational quantum eigensolver (VQE) [5] has been used for searching the ground state of the electronic Hamiltonian of molecules [6, 7].”)
Benedetti teaches and to use a variational inference arrangement executed on the hybrid combination to process the input data to generate corresponding output data from the classical computer for use in controlling or monitoring operation of the real physical system,
(Benedetti, Section I., “The focus of this Review is on hybrid approaches for machine learning [a variational inference arrangement ie hybrid approaches for machine learning executed on the hybrid combination]. In this field, quantum circuits are seen as components of a model for some data-driven task. Learning describes the process of iteratively updating the model’s parameters towards the goal [to process the input data to generate corresponding output data from the classical computer for use in controlling or monitoring operation of the real physical system].”)
Benedetti teaches and wherein the computing system generates an approximate posterior distribution of unobserved discrete variables based on a parameterized family of probability distributions, using a Born machine implemented using the quantum computer
PNG
media_image2.png
361
388
media_image2.png
Greyscale
Examiner interprets the limitation in light of figure 2 and the specification of the instant application ([0028], “The probabilistic models include a classical model that comprises a prior over unobserved discrete variables and a likelihood over observed variables. There is used a quantum model that approximates a posterior distribution of the unobserved variables given observed data.”)
(Benedetti, Section III B., “We now discuss generative modeling, an unsupervised learning task where the goal is to model an unknown probability distribution and generate synthetic data accordingly…Concretely, the task is to learn a model distribution qθ [based on a parameterized family of probability distributions; wherein θ is the parameter] that is close to a target distribution p. The closeness is defined in terms of a divergence D on the statistical manifold, and learning consists of minimizing this divergence; that is,
PNG
media_image3.png
27
292
media_image3.png
Greyscale
Since the target probability distribution is unknown, it is approximated [generates an approximate posterior distribution of unobserved discrete variables] using a dataset D = {v(i)}N i=1 which we have access to and which is distributed according to the target distribution. As an example, v(i) could be natural images extracted from the Internet…
The resulting generative model, known as the quantum circuit Born machine (QCBM) [26, 32], implements the probability distribution [using a Born machine implemented using the quantum computer]
PNG
media_image4.png
36
321
media_image4.png
Greyscale
”)
However, Benedetti does not explicitly teach wherein the variational inference arrangement uses at least one Bayesian network model;
Rather, Benedetti discloses one can exploit the probabilistic nature of quantum measurements to define a variety of machine learning models (such as a Bayesian network model) [wherein the variational inference arrangement uses at least one… model]
(Benedetti, Section II., “As we will see, one can exploit the probabilistic nature of quantum measurements to define a variety of machine learning models, and PQCs offer a concrete way to implement adjustable unitary operators U.”)
Jordan teaches wherein the variational inference arrangement uses at least one Bayesian network model,
(Jordan, Section 3.5, “In many problem domains it is natural to make additional structural assumptions about the state space and the transition probabilities that are not available within the simple HMM framework. A number of structured variations on HMMs have been considered in recent years (see Smyth et al., 1997); generically these variations can be viewed as “dynamic belief networks” (Dean & Kanazawa, 1989; Kanazawa, Koller, & Russell, 1995). Here we consider a particularly simple variation on the HMM theme known as the “factorial hidden Markov model” (Ghahramani & Jordan, 1997; Williams & Hinton, 1991). The graphical model for a factorial HMM (FHMM) is shown in figure 9 [uses at least one Bayesian network model]. The system is composed of a set of M chains indexed by m. Let the state node for the mth chain at time i be represented by X(m) i and let the transition matrix for the mth chain be represented by A(m) . We can view the effective state space for the FHMM as the Cartesian product of the state spaces associated with the individual chains.”
PNG
media_image5.png
362
604
media_image5.png
Greyscale
)
Benedetti is considered to be analogous to the claimed invention because they are in the same field of hybrid classical-quantum machine learning systems. Jordan is considered to be analogous to the claimed invention because they are in the same field of approximate inferencing. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Benedetti to incorporate the teachings of Jordan in order to provide a natural model capable of representing a large effective state space with a smaller number of parameters (Jordan, Section 3.5, “The FHMM is a natural model for systems in which the hidden state is realized via the joint configuration of an uncoupled set of dynamical systems. Moreover, an FHMM is able to represent a large effective state space with a much smaller number of parameters than a single unstructured Cartesian product HMM. For example, if we have 5 chains and in each chain the nodes have 10 states, the effective state space is of size 100,000, while the transition probabilities are represented compactly with only 500 parameters. A single unstructured HMM would require 1010 parameters for the transition matrix in this case. The fact that the output is a function of the states of all of the chains implies that the states become stochastically coupled when the outputs are observed.”)
In regards to claim 16,
Benedetti and Jordan teach The control system of claim 15,
Benedetti teaches wherein the at least one Bayesian network model arrangement is trained to represent the real physical system based on prior data and posterior data obtained from the real physical system, and wherein the Born machine is configured to generate the approximate posterior distribution of the unobserved variables by converging the parameterized family of probability distributions to the approximate posterior distribution.
(Benedetti, Section III B., “We now discuss generative modeling, an unsupervised learning task where the goal is to model an unknown probability distribution and generate synthetic data accordingly…Concretely, the task is to learn a model distribution qθ that is close to a target distribution p [wherein the at least one Bayesian network model arrangement is trained to represent the real physical system based on prior data ie initial belief and posterior data ie updated beliefs obtained from the real physical system (see fig. 1); wherein learning is training the model]. The closeness is defined in terms of a divergence D on the statistical manifold, and learning consists of minimizing this divergence; that is,
PNG
media_image3.png
27
292
media_image3.png
Greyscale
Since the target probability distribution is unknown, it is approximated [wherein the Born machine is configured to generate the approximate posterior distribution of the unobserved variables by converging the parameterized family of probability distributions to the approximate posterior distribution] using a dataset D = {v(i)}N i=1 which we have access to and which is distributed according to the target distribution.”)
In regards to claim 18,
Benedetti and Jordan teach The control system of claim 15,
Jordan teaches wherein the at least one Bayesian network model arrangement comprises a nested series of models,
(Jordan, Section 3.5, “In many problem domains it is natural to make additional structural assumptions about the state space and the transition probabilities that are not available within the simple HMM framework. A number of structured variations on HMMs have been considered in recent years (see Smyth et al., 1997); generically these variations can be viewed as “dynamic belief networks” (Dean & Kanazawa, 1989; Kanazawa, Koller, & Russell, 1995). Here we consider a particularly simple variation on the HMM theme known as the “factorial hidden Markov model” (Ghahramani & Jordan, 1997; Williams & Hinton, 1991). The graphical model for a factorial HMM (FHMM) is shown in figure 9 [wherein the at least one Bayesian network model arrangement comprises a nested series of models]. The system is composed of a set of M chains indexed by m. Let the state node for the mth chain at time i be represented by X(m) i and let the transition matrix for the mth chain be represented by A(m) . We can view the effective state space for the FHMM as the Cartesian product of the state spaces associated with the individual chains.”)
However, Jordan does not explicitly teach wherein at least one of the models of nested series models is implemented using the quantum computer.
Benedetti teaches wherein at least one of the models of nested series models is implemented using the quantum computer
(Benedetti, Section I., “The general hybrid approach is illustrated in Fig. 1 and is made of three main components: the human, the classical computer, and the quantum computer.”)
In regards to claim 19,
Benedetti and Jordan teach The control system of claim 18,
Jordan teaches wherein the models of the nested series of models are mutually different and at least two models of the nested series of models are specialized such that each performs a different function associated with the real physical system.
(Jordan, Section 3.5, “The graphical model for a factorial HMM (FHMM) is shown in figure 9. The system is composed of a set of M chains indexed by m [models of the nested series of models are mutually different and at least two models of the nested series of models are specialized such that each performs a different function associated with the real physical system; ie chain m is different from m+1 wherein Examiner interprets each chain to be performed on a subset (state node) of the input data (wherein the input data is provided by problem information of Benedetti and thus ‘associated’ with the real physical system)]. Let the state node for the mth chain at time i be represented by X(m) i and let the transition matrix for the mth chain be represented by A(m) .
PNG
media_image6.png
379
526
media_image6.png
Greyscale
”)
In regards to claim 20,
Benedetti and Jordan teach The control system of claim 18,
Jordan teaches wherein the nested series of models comprises a nested series of hidden Markov models.
(Jordan, Section 3.5, “In many problem domains it is natural to make additional structural assumptions about the state space and the transition probabilities that are not available within the simple HMM framework. A number of structured variations on HMMs have been considered in recent years (see Smyth et al., 1997); generically these variations can be viewed as “dynamic belief networks” (Dean & Kanazawa, 1989; Kanazawa, Koller, & Russell, 1995). Here we consider a particularly simple variation on the HMM theme known as the “factorial hidden Markov model” (Ghahramani & Jordan, 1997; Williams & Hinton, 1991). The graphical model for a factorial HMM (FHMM) is shown in figure 9 [wherein the nested series of models comprises a nested series of hidden Markov models]. The system is composed of a set of M chains indexed by m. Let the state node for the mth chain at time i be represented by X(m) i and let the transition matrix for the mth chain be represented by A(m) . We can view the effective state space for the FHMM as the Cartesian product of the state spaces associated with the individual chains.”)
In regards to claim 22,
Benedetti and Jordan teach The control system of claim 15,
Benedetti teaches wherein the control system is configured to infer an operating condition of the real physical system
(Benedetti, Fig. 1, “The role of the human is to set up the model using prior information, assess the learning process, and exploit the forecasts [to infer an operating condition ie a forecast of the real physical system].”)
Benedetti teaches from an error signal used to compensate for deviations in operation of the real physical system relative to a learnt representation of the real physical system,
(Benedetti, Section II C., “Just like classical models, PQC models are trained to perform data-driven tasks. The task of learning an arbitrary function from data is mathematically expressed as the minimization of a loss function L(θ) [from an error signal used to compensate for deviations in operation of the real physical system relative to a learnt representation of the real physical system; wherein loss operates on minimizing error], also known as the objective function, with respect to the parameter vector θ.”)
Benedetti teaches wherein the learnt representation of the real physical system is provided by the at least one Bayesian network model arrangement that is at least partially implemented using the quantum computer, and wherein the at least one Bayesian network model arrangement is trained to provide the representation of the real physical system.
(Benedetti, Section III B., “We now discuss generative modeling, an unsupervised learning task where the goal is to model an unknown probability distribution and generate synthetic data accordingly…Concretely, the task is to learn a model distribution qθ that is close to a target distribution p [wherein the learnt representation ie the learned model distribution closest to the target distribution of the real physical system is provided by the at least one Bayesian network model arrangement that is at least partially implemented using the quantum computer, and wherein the at least one Bayesian network model arrangement is trained to provide the representation of the real physical system; wherein the system is as taught in claim 1]. The closeness is defined in terms of a divergence D on the statistical manifold, and learning consists of minimizing this divergence; that is,
PNG
media_image3.png
27
292
media_image3.png
Greyscale
Since the target probability distribution is unknown, it is approximated using a dataset D = {v(i)}N i=1 which we have access to and which is distributed according to the target distribution.”)
In regards to claim 37,
Benedetti and Jordan teach The control system of claim 22,
Benedetti teaches wherein the control system is configured to generate the error signal by determining a difference between the corresponding output data, generated using the variational inference arrangement, and a system output data generated by the real physical system.
(Benedetti, Section I., “Measurement outcomes are post-processed by the classical computer to generate a forecast. To improve the forecast, the classical computer implements a learning algorithm that updates the model’s parameters.”)
(Benedetti, Section III A., “Let us first consider supervised learning tasks, e.g., classification and regression, on classical data. Given a dataset D = {(x(i),y(i))}N i=1, of N samples, the goal is to learn a model function f : X → Y that maps each x ∈ X to its corresponding target y ∈ Y [a system output data ie target generated by the real physical system]. A standard approach is to minimize a suitable regularized loss function, that is,
PNG
media_image7.png
53
393
media_image7.png
Greyscale
where θ is the set of parameters defining the model function, L quantifies the error of a forecast [wherein the control system is configured to generate the error signal by determining a difference between the corresponding output data, generated using the variational inference arrangement ie the hybrid approach, and a system output data generated by the real physical system], and R is a regularization function penalizing undesired values for the parameters.”)
Claims 23, 31-32, and 34 are rejected on the same grounds under 35 U.S.C. 103 as claim 15.
Examiner’s note: The recitation of “probabilistic graphical model arrangement” in claims 32 and 34 is not a term of art and the specification does not provide an explicit definition; thus, it is interpreted to be analogous to the Bayesian network model of claim 15.
Claim 24 is rejected on the same grounds under 35 U.S.C. 103 as claim 16.
Claim 26 is rejected on the same grounds under 35 U.S.C. 103 as claim 18.
Claim 27 is rejected on the same grounds under 35 U.S.C. 103 as claim 19.
Claim 28 is rejected on the same grounds under 35 U.S.C. 103 as claim 20.
Claim 30 is rejected on the same grounds under 35 U.S.C. 103 as claim 22.
Claim 38 is rejected on the same grounds under 35 U.S.C. 103 as claim 37.
Claim(s) 17 and 25 are rejected under 35 U.S.C. 103 as being unpatentable over Benedetti in view of Jordan in further view of Ranganath, Rajesh, et al. "Operator variational inference." Advances in Neural Information Processing Systems 29 (2016). (“Ranganath”)
In regards to claim 17,
Benedetti and Jordan teach The control system of claim 16,
Ranganath teaches wherein the at least one Bayesian network model arrangement is configured to use operator variational inference to generate the approximate posterior distribution by reducing a variation distance between an initial approximate posterior distribution and a true posterior distribution in a repeated manner.
(Ranganath, Section 2.4, “We developed the operator variational objective. It is a class of tractable objectives, each of which can be optimized to yield an approximation to the posterior [wherein the at least one Bayesian network model arrangement is configured to use operator variational inference to generate the approximate posterior distribution]. An operator variational objective is built from an operator, function class, and distance function to zero [by reducing a variation distance]. We now use this construction to design a new type of variational objective…
Section 3 Operator Variational Inference
We described operator variational objectives, a broad class of objectives for variational inference. We now examine how it can be optimized. We develop a black box algorithm [32, 23] based on Monte Carlo estimation and stochastic optimization. Our algorithm applies to a general class of models and any operator objective. Minimizing the operator objective involves two optimizations: minimizing the objective with respect to the approximating family Q and maximizing the objective with respect to the function class F (which is part of the objective) [between an initial approximate posterior distribution and a true posterior distribution]. We index the family Q with variational parameters λ and require that it satisfies properties typically assumed by black box methods [23]: the variational distribution q(z; λ) has a known and tractable density; we can sample from q(z; λ); and we can tractably compute the score function ∇λ log q(z; λ). We index the function class F with parameters θ, and require that fθ(·) is differentiable. In the experiments, we use neural networks, which are flexible enough to approximate a general family of test functions [10].”
PNG
media_image8.png
154
541
media_image8.png
Greyscale
[in a repeated manner; ie while not converged])
Ranganath is considered to be analogous to the claimed invention because they are in the same field of variational inferencing. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Benedetti and Jordan to incorporate the teachings of Ranganath in order to provide variational operators that allows inferences to massively scale and permit rich approximating families as to expand the class of variational families and fidelity of the resulting approximation (Ranganath, Section 1, “We develop operator variational inference (opvi), a black box algorithm that optimizes any operator objective. In the context of opvi, we show that the Langevin-Stein objective enjoys two good properties. First, it is amenable to data subsampling, which allows inference to scale to massive data. Second, it permits rich approximating families, called variational programs, which do not require analytically tractable densities. This greatly expands the class of variational families and the fidelity of the resulting approximation. (We note that the traditional kl is not amenable to using variational programs.)”)
Claim 25 is rejected on the same grounds under 35 U.S.C. 103 as claim 17.
Claim(s) 21, 29, 33 and 35-36 are rejected under 35 U.S.C. 103 as being unpatentable over Benedetti in view of Jordan in further view of Mohamed, Shakir, and Balaji Lakshminarayanan. "Learning in implicit generative models." arXiv preprint arXiv:1610.03483 (2016). (“Mohamed”)
In regards to claim 21,
Benedetti and Jordan teache The control system of claim 16,
Jordan teaches wherein the at least one Bayesian network model arrangement of the variational inference arrangement is taught by minimizing an objective function at least one of: (i) a Kullback-Leibler (KL) divergence of a true posterior distribution relative to the approximate posterior distribution of the unobserved variables and using a classifier that estimates a probability ratio between the approximate posterior distribution and a prior of the unobserved variables; and (ii) a kernelized Stein discrepancy (KSD) between the true posterior distribution and the approximate posterior distribution.
(Jordan, Section 6, “More formally, let P(S)represent the joint distribution on the graphical model of interest, where as before S represents all of the nodes of the graph and H and E are disjoint subsets of S representing the hidden nodes and the evidence nodes, respectively. We wish to approximate the conditional probability P(H | E). We introduce an approximating family of conditional probability distributions, Q(H | E, λ), where λ are variational parameters. The graph representing Q is not generally the same as the graph representing P; generally it is a sub-graph. From the family of approximating distributions Q, we choose a particular distribution by minimizing the Kullback-Leibler (KL) divergence [minimizing an objective function at least one of: (i) a Kullback-Leibler (KL) divergence], D(Q k P), with respect to the variational parameters:
PNG
media_image9.png
26
575
media_image9.png
Greyscale
where for any probability distributions Q(S) and P(S) the KL divergence is defined as follows:
PNG
media_image10.png
42
568
media_image10.png
Greyscale
[of a true posterior distribution relative to the approximate posterior distribution of the unobserved variables]
The minimizing values of the variational parameters, λ∗, define a particular distribution, Q(H | E, λ∗), that we treat as the best approximation of P(H | E) in the family Q(H | E, λ).”)
However, Jordan does not explicitly teach and using a classifier that estimates a probability ratio between the approximate posterior distribution and a prior of the unobserved variables
Mohamed teaches and using a classifier that estimates a probability ratio between the approximate posterior distribution and a prior of the unobserved variables
(Mohamed, Section 2.2, “The density ratio can be computed by building a classifier to distinguish observed data from that generated by the model [and using a classifier that estimates a probability ratio]. This is the most popular approach for density ratio estimation and the first port of call for learning in implicit models….
Section 2.3 Divergence Minimisation
The f-divergences contain the KL divergence as a special case and are equipped with an exploitable variational formulation:
PNG
media_image11.png
87
289
media_image11.png
Greyscale
[between the approximate posterior distribution and a prior of the unobserved variables; wherein the approximate posterior distribution and a prior of the unobserved variables is provided by Jordan]
where f is a convex function with derivative f 0 and Fenchel conjugate f † ; this divergence class instantiates many familiar divergences, such as the KL and Jensen-Shannon divergence. The variational formulation introduces the functions t(x) whose optimum is related to the density ratio since t ∗ (x) = f 0 (r(x)).”)
Mohamed is considered to be analogous to the claimed invention because they are in the same field of learning implicit models (such as a Born machine). Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Benedetti and Jordan to incorporate the teachings of Mohamed in order to develop more general and flexible implicit models. (Mohamed, Section 1, “We are interested in developing more general and flexible implicit generative models where the function G is a nonlinear function with d > m, specified by deep networks.”)
Claim 29, 33 and 35 are rejected on the same grounds under 35 U.S.C. 103 as claim 21.
In regards to claim 36,
Benedetti and Jordan and Ranganath teach The control system of claim 17,
Jordan teaches wherein objective function of the operator variational inference comprises Kullback-Leibler (KL) divergence or Stein discrepancy.
(Jordan, Section 6, “More formally, let P(S)represent the joint distribution on the graphical model of interest, where as before S represents all of the nodes of the graph and H and E are disjoint subsets of S representing the hidden nodes and the evidence nodes, respectively. We wish to approximate the conditional probability P(H | E). We introduce an approximating family of conditional probability distributions, Q(H | E, λ), where λ are variational parameters. The graph representing Q is not generally the same as the graph representing P; generally it is a sub-graph. From the family of approximating distributions Q, we choose a particular distribution by minimizing the Kullback-Leibler (KL) divergence [wherein objective function of the operator variational inference comprises Kullback-Leibler (KL) divergence], D(Q k P), with respect to the variational parameters:…”)
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Benedetti, Marcello, et al. "A generative modeling approach for benchmarking and training shallow quantum circuits." arXiv preprint arXiv:1801.07686 (2019).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JASMINE THAI whose telephone number is (703)756-5904. The examiner can normally be reached M-F 8-4.
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, Michael Huntley can be reached at (303) 297-4307. 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.
/J.T.T./Examiner, Art Unit 2129
/MICHAEL J HUNTLEY/Supervisory Patent Examiner, Art Unit 2129
Prosecution Insights