DETAILED ACTION
This action is in response to the application filed 06/18/2024. Claims 1-20 are pending and have been examined.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 06/18/2024, 04/03/2025, and 12/08/2025 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
Specification
The disclosure is objected to because of the following informalities:
Many of the formulas and symbols are fuzzy and difficult to parse.
Equations 2.5-2.11, 2.14-2.16, 2.19, 4.1-4.5, 4.9-4.12, 5.1-5.8
Some terms in page 13, paragraph 1; page 14, paragraph 2 & paragraph 3; page 17, paragraph 2; page 19, paragraph 1; page 20, paragraphs 2 & 4; page 23, paragraph 2
The pseudocode at the top of page 20
Appropriate correction is required.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-3 and 6 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Leiteritz (How to Avoid Trivial Solutions in Physics-Informed Neural Networks, published 12/10/2021, arXiv:2112.05620v1).
Regarding claim 1, Leiteritz discloses [a] method comprising:
training a physics-informed neural network using a plurality of training samples:
“In this work, we investigate the prediction performance of PINNs (physics-informed neural network[s]) with respect to the number of collocation points used to enforce the physics-based penalty terms” (Leiteritz, page 1, left column, Abstract)
“In this section we propose two ways to improve the standard collocation point sampling and loss function of PINNs. This enables the efficient use of PINNs in data-scarce simulation settings. We assume that a fixed number of “classical” training data samples, which typically only consist of initial and boundary information, is available to compute the MSE loss.” (Leiteritz, page 3, left column, paragraph 2)
“To consolidate the effects of regular grid sampling further, we did an extensive study by varying the number of collocation points
n
c
from 10 to 50 (plurality of training samples). For each number of collocation points we trained the network 30 times and calculated the resulting mean-squared error (MSE) to the analytical solution in 1000 equidistantly placed points on the time domain” (Leiteritz, page 5, left column, paragraph 2)
… wherein training the physics-informed neural network includes:
differentiating at least one partial differential equation characterizing a time-dependent behavior of a mechanical system: “To this end, we assume that our data is the product of a physical process that follows some dynamics (time-dependent) which can be described by a PDE (partial differential equation) in the general form of
PNG
media_image1.png
82
531
media_image1.png
Greyscale
Given that this information is available, PINNs exploit it by substituting the solution u with the prediction of the network
ϕ
(
x
;
θ
)
and evaluating the differential operator using automatic differentiation to form a new physical loss term
PNG
media_image2.png
46
316
media_image2.png
Greyscale
” (Leiteritz, page 2, right column, paragraph 5)
minimizing a loss function specifying an error of the physics-informed neural network with respect to the training samples by assigning a plurality of weights in a residual loss value:
“Following Raissi et al. (2019), we start with a fully connected feed-forward network to build a PINN (physics-informed neural network). Let
PNG
media_image3.png
41
261
media_image3.png
Greyscale
be a single layer of the network with inputs
x
∈
R
d
, learnable weights
w
∈
R
d
, bias b and an activation function
σ
:
R
→
R
. The feed-forward network is then expressed as a composition of n layers,
PNG
media_image4.png
42
417
media_image4.png
Greyscale
where
θ
represents the set of all trainable parameters” (Leiteritz, page 2, right column, paragraph 2)
“Given data in the form of input-output pairs
{
x
i
,
y
i
}
i
=
1
N
(training samples) we can learn the parameters of the network by minimizing the mean squared error (MSE) loss function
PNG
media_image5.png
96
399
media_image5.png
Greyscale
” (Leiteritz, page 2, right column, paragraph 3)
… to account for physical causality in the partial differential equation:
“a PINN does not guarantee a physically valid solution. But it encourages the solution to be close to one” (Leiteritz, page 2, right column, paragraph 4)
“to form a new physical loss term
PNG
media_image2.png
46
316
media_image2.png
Greyscale
” (Leiteritz, page 2, right column, paragraph 5)” (Leiteritz, page 2, right column, paragraph 5). This physical loss term encourages solutions close to those physically valid, thus accounting for physical causality.
“This is then added to the MSE loss function (equation 3), resulting in a physics-informed loss
PNG
media_image6.png
113
548
media_image6.png
Greyscale
” (Leiteritz, page 3, left column, paragraph 1)
predicting, using the physics-informed neural network, movement of at least one component of the mechanical system: “To demonstrate the effect of the aforementioned variants, we use a PINN to predict the motion (movement) of a simple 1-D harmonic oscillator” (Leiteritz, page 3, right column, paragraph 3).
Leiteritz relates to physics-informed neural networks simulating mechanical systems and is analogous to the claimed invention.
Regarding claim 2, the rejection of claim 1 is incorporated. Leiteritz further discloses a method, wherein training the physics-informed neural network comprises iteratively training the physics-informed neural network over a plurality of training iterations: “In each iteration of the PINN’s training, the physical loss has to be evaluated at every collocation point. And this, again, requires automatic differentiation.” (Leiteritz, page 3, left column, paragraph 2)
Regarding claim 3, the rejection of claim 2 is incorporated. Leiteritz further discloses a method, wherein training the physics-informed neural network comprises using a gradient descent algorithm: “Given data in the form of input-output pairs
{
x
i
,
y
i
}
i
=
1
N
we can learn the parameters of the network by minimizing the mean squared error (MSE) loss function
PNG
media_image5.png
96
399
media_image5.png
Greyscale
with regard to the parameters
θ
, typically using some form of stochastic gradient descent method such as ADAM” (Leiteritz, page 2, right column, paragraph 3)
Regarding claim 6, the rejection of claim 1 is incorporated. Leiteritz further discloses a method, wherein differentiating the partial differential equation comprises using automatic differentiation: “To this end, we assume that our data is the product of a physical process that follows some dynamics (time-dependent) which can be described by a PDE (partial differential equation) in the general form of
PNG
media_image1.png
82
531
media_image1.png
Greyscale
Given that this information is available, PINNs exploit it by substituting the solution u with the prediction of the network
ϕ
(
x
;
θ
)
and evaluating the differential operator using automatic differentiation to form a new physical loss term
PNG
media_image2.png
46
316
media_image2.png
Greyscale
” (Leiteritz, page 2, right column, paragraph 5)
Claim Rejections - 35 USC § 103
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Leiteritz (How to Avoid Trivial Solutions in Physics-Informed Neural Networks, published 12/10/2021, arXiv:2112.05620v1) in view of Lee (METHOD AND SYSTEMS FOR SCAN CONVERSION WITH HIGHER RESOLUTION, published 5/14/2020, US 2020/0151513 A1).
Regarding claim 4, the rejection of claim 2 is incorporated. While Leiteritz fails to disclose the further limitations of the claim, Lee discloses a method, wherein training the physics-informed neural network comprises updating the plurality of weights in each iteration of the training iterations: “At each iteration of training, the cost defined by the cost function is calculated and the error is back-propagated to update the parameters or weights
w
i
of the neural network:
w
i
←
w
i
+
∆
w
i
” (Lee, [0042])
Lee relates to iteratively training neural networks and is analogous to the claimed invention. The primary reference teaches a method of iteratively training a neural network. The claimed invention improves upon this method by updating the network weights for each training iteration. Lee teaches a method of disclosing neural network weights at each iteration of training, applicable to the primary reference. A person of ordinary skill in the art before the effective filing date of the claimed invention would have recognized that updating the weights of the primary reference’s network during each training iteration would lead to the predictable result of optimizing weights according to loss values calculated at each iteration, and would improve the known device by gradually optimizing the network weights during each iteration of training (MPEP 2143 I. (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results).
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Leiteritz (How to Avoid Trivial Solutions in Physics-Informed Neural Networks, published 12/10/2021, arXiv:2112.05620v1) in view of Raissi (Physics Informed Deep Learning (Part I): Data-driven Solutions of Nonlinear Partial Differential Equations, published 11/30/2017, arXiv:1711.10561v1).
Regarding claim 7, the rejection of claim 1 is incorporated. While Leiteritz fails to disclose the further limitations of the claim, Raissi discloses a method, wherein the partial differential equation characterizes one of: a conservation law, a diffusion process, an advection-diffusion-reaction system, and a kinetic equation: “let us consider parametrized and nonlinear partial differential equations of the general form
PNG
media_image7.png
44
265
media_image7.png
Greyscale
where
u
(
t
,
x
)
denotes the latent (hidden) solution and
N
[
∙
;
λ
]
is a nonlinear operator parametrized by
λ
. This setup encapsulates a wide range of problems in mathematical physics including conservation laws, diffusion processes, advection-diffusion-reaction systems, and kinetic equations” (Raissi, page 3, paragraph 4).
Raissi relates to physics-informed networks for simulating mechanical systems and is analogous to the claimed invention. The primary reference teaches a method of simulating the dynamics of a PDE with a neural network. The claimed invention improves upon this method by applying it to PDEs related to conservation, diffusion, advection-diffusion reaction systems, and kinetics. Raissi teaches a method of applying physics-informed networks to PDEs related to conservation, diffusion, advection-diffusion reaction systems, and kinetics, applicable to the primary reference. A person of ordinary skill in the art before the effective filing date of the claimed invention would have recognized that applying the primary reference’s physics-informed network to PDEs of these different fields would lead to the predictable result of simulating each system, and would improve the known device by enabling it to simulate dynamics for a variety of different fields (MPEP 2143 I. (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results).
Claims 8-10, 13, 15-17, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Leiteritz (How to Avoid Trivial Solutions in Physics-Informed Neural Networks, published 12/10/2021, arXiv:2112.05620v1) in view of Mahmoudabadbozchelou (RHEOLOGY-INFORMED NEURAL NETWORKS FOR COMPLEX FLUIDS, filed 1/21/2022, US 2022/0228960 A1).
Regarding claim 8, Leiteritz discloses operations comprising:
training a physics-informed neural network using a plurality of training samples:
“In this work, we investigate the prediction performance of PINNs (physics-informed neural network[s]) with respect to the number of collocation points used to enforce the physics-based penalty terms” (Leiteritz, page 1, left column, Abstract)
“In this section we propose two ways to improve the standard collocation point sampling and loss function of PINNs. This enables the efficient use of PINNs in data-scarce simulation settings. We assume that a fixed number of “classical” training data samples, which typically only consist of initial and boundary information, is available to compute the MSE loss.” (Leiteritz, page 3, left column, paragraph 2)
“To consolidate the effects of regular grid sampling further, we did an extensive study by varying the number of collocation points
n
c
from 10 to 50 (plurality of training samples). For each number of collocation points we trained the network 30 times and calculated the resulting mean-squared error (MSE) to the analytical solution in 1000 equidistantly placed points on the time domain” (Leiteritz, page 5, left column, paragraph 2)
… wherein training the physics-informed neural network includes:
differentiating at least one partial differential equation characterizing a time-dependent behavior of a mechanical system: “To this end, we assume that our data is the product of a physical process that follows some dynamics (time-dependent) which can be described by a PDE (partial differential equation) in the general form of
PNG
media_image1.png
82
531
media_image1.png
Greyscale
Given that this information is available, PINNs exploit it by substituting the solution u with the prediction of the network
ϕ
(
x
;
θ
)
and evaluating the differential operator using automatic differentiation to form a new physical loss term
PNG
media_image2.png
46
316
media_image2.png
Greyscale
” (Leiteritz, page 2, right column, paragraph 5)
minimizing a loss function specifying an error of the physics-informed neural network with respect to the training samples by assigning a plurality of weights in a residual loss value:
“Following Raissi et al. (2019), we start with a fully connected feed-forward network to build a PINN (physics-informed neural network). Let
PNG
media_image3.png
41
261
media_image3.png
Greyscale
be a single layer of the network with inputs
x
∈
R
d
, learnable weights
w
∈
R
d
, bias b and an activation function
σ
:
R
→
R
. The feed-forward network is then expressed as a composition of n layers,
PNG
media_image4.png
42
417
media_image4.png
Greyscale
where
θ
represents the set of all trainable parameters” (Leiteritz, page 2, right column, paragraph 2)
“Given data in the form of input-output pairs
{
x
i
,
y
i
}
i
=
1
N
(training samples) we can learn the parameters of the network by minimizing the mean squared error (MSE) loss function
PNG
media_image5.png
96
399
media_image5.png
Greyscale
” (Leiteritz, page 2, right column, paragraph 3)
… to account for physical causality in the partial differential equation:
“a PINN does not guarantee a physically valid solution. But it encourages the solution to be close to one” (Leiteritz, page 2, right column, paragraph 4)
“to form a new physical loss term
PNG
media_image2.png
46
316
media_image2.png
Greyscale
” (Leiteritz, page 2, right column, paragraph 5)” (Leiteritz, page 2, right column, paragraph 5). This physical loss term encourages solutions close to those physically valid, thus accounting for physical causality.
“This is then added to the MSE loss function (equation 3), resulting in a physics-informed loss
PNG
media_image6.png
113
548
media_image6.png
Greyscale
” (Leiteritz, page 3, left column, paragraph 1)
predicting, using the physics-informed neural network, movement of at least one component of the mechanical system: “To demonstrate the effect of the aforementioned variants, we use a PINN to predict the motion (movement) of a simple 1-D harmonic oscillator” (Leiteritz, page 3, right column, paragraph 3).
Leiteritz relates to physics-informed neural networks simulating mechanical systems and is analogous to the claimed invention.
While Leiteritz fails to disclose the further limitations of the claim, Mahmoudabadbozchelou discloses [a] system comprising: at least one processor; and a physics-informed neural network trainer implemented on the at least one processor and configured to perform operations:
“A comprehensive machine-learning algorithm, namely a Multi-Fidelity Neural Network (MFNN) (physics-informed neural network) architecture, is disclosed for data-driven constitutive meta-modelling of complex fluids. The physics-based neural networks are informed by underlying rheological constitutive models through synthetic generation of low-fidelity model-based data points. The performance of these rheologically-informed algorithms is investigated and compared against classical Deep Neural Networks (DNN).” (Mahmoudabadbozchelou, [0010])
“In order to evaluate the ability of the MFNN (physics-informed neural network) and the DNN algorithms, in each section the networks are trained on 18 (of 19 total) samples' experimental data, and asked to make predictions for the 19th sample.” (Mahmoudabadbozchelou, [0055])
“The methods, operations, modules, and systems described herein may be implemented in one or more computer programs executing on a programmable computer system. FIG. 21 is a simplified block diagram illustrating an exemplary computer system 100, on which the computer programs may operate as a set of computer instructions. The computer system 100 includes at least one computer processor 102” (Mahmoudabadbozchelou, [0083])
Mahmoudabadbozchelou relates to using physics-informed neural networks to make predictions in physical systems and is analogous to the claimed invention. The primary reference teaches a method of training and utilizing a physics-informed neural network. The claimed invention improves upon this method by storing it in the form of instructions on computer hardware. Mahmoudabadbozchelou teaches generic computer hardware for running operations to train and utilize physics-informed neural networks, applicable to the primary reference. A person of ordinary skill in the art would have recognized that storing the primary reference’s method as computer instructions on Mahmoudabadbozchelou’s hardware would lead to the predictable result of the method being executable by a computing system, and would improve the known device by allowing it to be performed with real data (MPEP 2143 I. (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results).
The analysis of claims 9-10 & 13 mirrors that of claims 2-3 & 6, with the exception that claims 9-10 & 13 are directed to generic computer hardware which executes the methods of claims 2-3 & 6. This generic hardware is taught by Mahmoudabadbozchelou, as discussed regarding claim 8. Thus, claims 9-10 & 13 are rejected under the same rationales used for claims 2-3 & 6, respectively.
Regarding claim 15, Leiteritz discloses operations comprising:
training a physics-informed neural network using a plurality of training samples:
“In this work, we investigate the prediction performance of PINNs (physics-informed neural network[s]) with respect to the number of collocation points used to enforce the physics-based penalty terms” (Leiteritz, page 1, left column, Abstract)
“In this section we propose two ways to improve the standard collocation point sampling and loss function of PINNs. This enables the efficient use of PINNs in data-scarce simulation settings. We assume that a fixed number of “classical” training data samples, which typically only consist of initial and boundary information, is available to compute the MSE loss.” (Leiteritz, page 3, left column, paragraph 2)
“To consolidate the effects of regular grid sampling further, we did an extensive study by varying the number of collocation points
n
c
from 10 to 50 (plurality of training samples). For each number of collocation points we trained the network 30 times and calculated the resulting mean-squared error (MSE) to the analytical solution in 1000 equidistantly placed points on the time domain” (Leiteritz, page 5, left column, paragraph 2)
… wherein training the physics-informed neural network includes:
differentiating at least one partial differential equation characterizing a time-dependent behavior of a mechanical system: “To this end, we assume that our data is the product of a physical process that follows some dynamics (time-dependent) which can be described by a PDE (partial differential equation) in the general form of
PNG
media_image1.png
82
531
media_image1.png
Greyscale
Given that this information is available, PINNs exploit it by substituting the solution u with the prediction of the network
ϕ
(
x
;
θ
)
and evaluating the differential operator using automatic differentiation to form a new physical loss term
PNG
media_image2.png
46
316
media_image2.png
Greyscale
” (Leiteritz, page 2, right column, paragraph 5)
minimizing a loss function specifying an error of the physics-informed neural network with respect to the training samples by assigning a plurality of weights in a residual loss value:
“Following Raissi et al. (2019), we start with a fully connected feed-forward network to build a PINN (physics-informed neural network). Let
PNG
media_image3.png
41
261
media_image3.png
Greyscale
be a single layer of the network with inputs
x
∈
R
d
, learnable weights
w
∈
R
d
, bias b and an activation function
σ
:
R
→
R
. The feed-forward network is then expressed as a composition of n layers,
PNG
media_image4.png
42
417
media_image4.png
Greyscale
where
θ
represents the set of all trainable parameters” (Leiteritz, page 2, right column, paragraph 2)
“Given data in the form of input-output pairs
{
x
i
,
y
i
}
i
=
1
N
(training samples) we can learn the parameters of the network by minimizing the mean squared error (MSE) loss function
PNG
media_image5.png
96
399
media_image5.png
Greyscale
” (Leiteritz, page 2, right column, paragraph 3)
… to account for physical causality in the partial differential equation:
“a PINN does not guarantee a physically valid solution. But it encourages the solution to be close to one” (Leiteritz, page 2, right column, paragraph 4)
“to form a new physical loss term
PNG
media_image2.png
46
316
media_image2.png
Greyscale
” (Leiteritz, page 2, right column, paragraph 5)” (Leiteritz, page 2, right column, paragraph 5). This physical loss term encourages solutions close to those physically valid, thus accounting for physical causality.
“This is then added to the MSE loss function (equation 3), resulting in a physics-informed loss
PNG
media_image6.png
113
548
media_image6.png
Greyscale
” (Leiteritz, page 3, left column, paragraph 1)
predicting, using the physics-informed neural network, movement of at least one component of the mechanical system: “To demonstrate the effect of the aforementioned variants, we use a PINN to predict the motion (movement) of a simple 1-D harmonic oscillator” (Leiteritz, page 3, right column, paragraph 3).
Leiteritz relates to physics-informed neural networks simulating mechanical systems and is analogous to the claimed invention.
While Leiteritz fails to disclose the further limitations of the claim, Mahmoudabadbozchelou discloses [a] non-transitory computer readable medium storing executable instructions that when executed by at least one processor of a computer control the computer to perform operations:
“A comprehensive machine-learning algorithm, namely a Multi-Fidelity Neural Network (MFNN) (physics-informed neural network) architecture, is disclosed for data-driven constitutive meta-modelling of complex fluids. The physics-based neural networks are informed by underlying rheological constitutive models through synthetic generation of low-fidelity model-based data points. The performance of these rheologically-informed algorithms is investigated and compared against classical Deep Neural Networks (DNN).” (Mahmoudabadbozchelou, [0010])
“A computer program product for predicting one or more rheological properties of a non-Newtonian fluid using a multi-fidelity neural network framework, said computer program product residing on a non-transitory computer readable medium having a plurality of instructions stored thereon which, when executed by a computer processor, cause that computer processor to: (a) receive, at a physics-informed low fidelity neural network, a plurality of low fidelity parameter inputs related to the non-Newtonian fluid; (b) generate, by the physics-informed low fidelity neural network, one or more synthetically generated parameters of the non-Newtonian fluid based on the plurality of low fidelity parameter inputs” (Mahmoudabadbozchelou, [0083])
Mahmoudabadbozchelou relates to using physics-informed neural networks to make predictions in physical systems and is analogous to the claimed invention. The primary reference teaches a method of training and utilizing a physics-informed neural network. The claimed invention improves upon this method by storing it in the form of instructions on computer hardware. Mahmoudabadbozchelou teaches generic computer hardware for running operations to train and utilize physics-informed neural networks, applicable to the primary reference. A person of ordinary skill in the art would have recognized that storing the primary reference’s method as computer instructions on Mahmoudabadbozchelou’s hardware would lead to the predictable result of the method being executable by a computing system, and would improve the known device by allowing it to be performed with real data (MPEP 2143 I. (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results).
The analysis of claims 16-17 & 20 mirrors that of claims 2-3 & 6, with the exception that claims 16-17 & 20 are directed to generic computer hardware which executes the methods of claims 2-3 & 6. This generic hardware is taught by Mahmoudabadbozchelou, as discussed regarding claim 15. Thus, claims 16-17 & 20 are rejected under the same rationales used for claims 2-3 & 6, respectively.
Claims 11 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Leiteritz (How to Avoid Trivial Solutions in Physics-Informed Neural Networks, published 12/10/2021, arXiv:2112.05620v1) in view of Mahmoudabadbozchelou (RHEOLOGY-INFORMED NEURAL NETWORKS FOR COMPLEX FLUIDS, filed 1/21/2022, US 2022/0228960 A1), and further in view of Lee (METHOD AND SYSTEMS FOR SCAN CONVERSION WITH HIGHER RESOLUTION, published 5/14/2020, US 2020/0151513 A1).
The analysis of claims 11 & 18 mirrors that of claim 4, with the exception that claims 11 & 18 are directed to generic computer hardware which executes the methods of claim 4. The generic hardware of claims 11 & 18 is taught by Mahmoudabadbozchelou, as discussed regarding claims 8 & 15, respectively. Thus, claims 11 & 18 are rejected under the same rationales used for claim 4, respectively.
Claim 14 is rejected under 35 U.S.C. 103 as being unpatentable over Leiteritz (How to Avoid Trivial Solutions in Physics-Informed Neural Networks, published 12/10/2021, arXiv:2112.05620v1) in view of Mahmoudabadbozchelou (RHEOLOGY-INFORMED NEURAL NETWORKS FOR COMPLEX FLUIDS, filed 1/21/2022, US 2022/0228960 A1), and further in view of Raissi (Physics Informed Deep Learning (Part I): Data-driven Solutions of Nonlinear Partial Differential Equations, published 11/30/2017, arXiv:1711.10561v1).
The analysis of claim 14 mirrors that of claim 7, with the exception that claim 14 is directed to generic computer hardware which executes the methods of claim 7. This generic hardware is taught by Mahmoudabadbozchelou, as discussed regarding claim 8. Thus, claim 14 is rejected under the same rationales used for claim 7.
Allowable Subject Matter
Claims 5, 12, and 19 are objected to as being dependent upon rejected base claims, but would be allowable if rewritten in independent form including all of the limitations of the base claims and any intervening claims.
Regarding claim 5, “in at least one iteration of the training iterations, each of the weights in the residual loss value is inversely exponentially proportional to a magnitude of a residual from a previous iteration.” is not taught by the prior art of record. The closest prior arts of record are Niaki (Physics-Informed Neural Network for Modelling the Thermochemical Curing Process of Composite-Tool Systems During Manufacture, published 6/14/2021, arXiv:2011.13511v2) and Maddu (Inverse-Dirichlet Weighting Enables Reliable Training of Physics Informed Neural Networks, published 7/2/2021, arXiv:2107.00940v1).
Niaki discloses a method, wherein, in at least one iteration of the training iterations, each of the weights in the residual loss value is inversely exponentially proportional to a magnitude of a residual from a previous iteration:
PNG
media_image8.png
418
897
media_image8.png
Greyscale
(Niaki, page 12, paragraph 2)
Niaki does not disclose loss weights inversely exponentially proportional to a magnitude of a residual from a previous iteration.
Maddu discloses a method, wherein, in at least one iteration of the training iterations, each of the weights in the residual loss value is inversely exponentially proportional to a magnitude of a residual from a previous iteration:
PNG
media_image9.png
532
800
media_image9.png
Greyscale
(Maddu, page 4, left column, paragraph 3)
PNG
media_image10.png
472
759
media_image10.png
Greyscale
(Maddu, page 3, right column, paragraph 3)
Maddu does not disclose loss weights inversely exponentially proportional to a magnitude of a residual from a previous iteration.
Therefore, the prior art of record, individually or in combination, does not disclose the entirety of claim 5 as a whole. Substantially similar claims 12 and 19 are not disclosed in their entirety by the prior art of record on these same grounds.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Oh (Toward the Fully Physics-Informed Echo State Network - an ODE Approximator Based on Recurrent Artificial Neurons, published 11/13/2020, arXiv:2011.06769v1) discloses loss for a physics-informed neural network that takes physical causality into account.
Torun (Causal and Passive Parameterization of S-Parameters Using Neural Networks, published October 2020, IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 68, NO. 10) discloses a method of maintaining physical causality in a neural network.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Aaron P Gormley whose telephone number is (571)272-1372. The examiner can normally be reached Monday - Friday 12:00 PM - 8:00 PM 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, Michelle T Bechtold can be reached at (571) 431-0762. 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.
/AG/Examiner, Art Unit 2148 /MICHELLE T BECHTOLD/Supervisory Patent Examiner, Art Unit 2148
Prosecution Insights