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 .
Claim Status
Claims 1-26 are pending and examined herein.
Claims 1-26 are rejected.
Claim 23 is objected to.
Priority
Claims 1-26 are granted the claim to the benefit of priority to U.S. Provisional application 63/234768 filed 19 August 2021 and 63/167255 filed 29 March 2021. Thus, the effective filling date of claims 1-26 is 29 March 2021.
Information Disclosure Statement
The information disclosure statement (IDS) was received on 27 September 2023. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement has been considered by the examiner.
Drawings
The drawings received 25 September 2023 are accepted.
Claim Objections
Claim 23 is objected to because of the following informalities:
Claim 23 recites “the one or more property calculus…” in lines 3-4 but should read “the one or more property calculators”.
Appropriate correction is required.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier.
Such claim limitations are: “a density functional theory engine to calculate a self-consistent field” in claims 1 and 19.
There is no associated structure (i.e., computer hardware and algorithm) for a density functional theory engine to perform the entire claimed function of “to calculate a self-consistent field” (i.e., does not show the associated algorithm for computing the self-consistent field).
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
Claim Rejections - 35 USC § 112
112/a
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 1-26 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Claims 1 and 19 recite “a density functional theory engine to calculate a self-consistent field”. The MPEP provides “When a claim containing a computer-implemented 35 U.S.C. 112(f) claim limitation is found to be indefinite under 35 U.S.C. 112(b) for failure to disclose sufficient corresponding structure (e.g., the computer and the algorithm) in the specification that performs the entire claimed function, it will also lack written description under section 112(a)” (MPEP 2181 IV). The instant disclosure provides “DFT engine 200 uses charge density distribution 190 o run at least one round of SCF to compute final charge density distribution 210 and estimated error in charge density distribution 190. Optionally, multiple rounds of SCF may also be run for DFT convergence by feeding final charge density distribution 210 back into DFT engine 200 multiple times”. The instant disclosure does not provide an associated structure (i.e., computer hardware and algorithm) for performing the claimed function of a density theory engine to calculate a self-consistent field (e.g., does not provide a corresponding structure such as steps taken to calculate a self-consistent field). Dependent claims 2-18 and 20-26 are rejected by virtue of their dependency on a rejected claim without alleviating the issue.
112/b
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 1-26 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 3 and 21 recite “wherein the set of atomic moieties and positions is in the form of an array of scalars dependent upon the atomic moiety” which renders the metes and bounds of the claim indefinite. The indefiniteness arises because it is unclear if this limitation is meant to encompass that all the scalars in the array are dependent upon one atomic moiety (if so which atomic moiety of the set of atomic moieties is the array of scalars dependent upon) or if this is meant to encompass that each scalar in the array of scalars is associated with an atomic moiety of the set of atomic moieties. The specification does not provide a clear and precise definition of the limitation, nor would one skilled in the art recognize the metes and bounds of said limitation. For the sake of furthering examination, this limitation will be interpreted as each scalar in the array of scalars is associated with an atomic moiety of the set of atomic moieties.
Claims 4 and 21 recite “the charge density distribution at each atomic position”. There is insufficient antecedent basis for this limitation in the claim. The metes and bounds of the claims are rendered indefinite because the claims does not make clear what “the charge density distribution at each atomic position” is referring to in the claims (e.g., it is unclear if the “charge density distribution at each atomic position” is meant to refer to and be the same as the “initial charge density distribution” recited in claim 1). Dependent claims 5 and 7 are rejected by virtue of its dependency on a rejected claim without alleviating the indefiniteness
Claims 6, 8, 10, 12, and 13 recite “the equivariant neural network” in line 1 of claim 6, in lines 2-3 of claim 8, line 1 of claim 10, line 1 of claim 12, and lines 2-3 of claim 13. There is insufficient antecedent basis for these limitations in the claim. The metes and bounds of the claims are rendered indefinite because the claims does not make clear what “the equivariant neural network” is referring to in the claims (e.g. claim 6 depends from claim 3 (which depends from claim 1) which provides basis for a “neural network) and claim 8 depends from claim 1 which provides basis for a “neural network”). Dependent claims 9, 11, and 14-18 are rejected by virtue of its dependency on a rejected claim without alleviating the indefiniteness. For the sake of furthering examination, claim 6 will be interpreted as “wherein the neural network is an equivariant neural network that is equivariant in terms of rotation and translation” and claim 8 will be interpreted as “inputting the initial charge density distribution produced by the neural network…”.
Claims 14 and 17 recite “the equivariant operator” in line 1 (of claim 14) and lines 1-2 (of claim 17) and claim 18 recites “the operator”. There is insufficient antecedent basis for these limitations in the claim. The metes and bounds of these claims are rendered indefinite because the claim does not make clear what “the equivariant operator” or “the operator” are referring to in the claims (e.g., the claims from which 14, 17, and 18 depend on do not set out what “the equivariant operator” or “the operator” are and it is further unclear if “the operator” is meant to refer to “the equivariant operator” or meant to be a different operator). Dependent claims 15 and 16 are rejected by virtue of their dependency on a rejected claim without alleviating the indefiniteness. For the sake of furthering examination, claim 14 will be interpreted as reciting “wherein the equivariant neural network implements an equivariant operator to map a scalar field of electronic density to a scalar field of total energy” and claim 18 will be interpreted as reciting “the equivariant operator”.
Claim 16 recites “wherein external nuclear potential, mean field, and approximate exchange -correlation energies are directly calculated” which renders the metes and bounds of the claim indefinite. The indefiniteness arises because it is unclear if this limitation is meant to further limit an active step of the method (e.g., the step of receiving information from the neural network) or if this limitation is meant to be an intended use of the method. Dependent claims 17 and 18 are rejected by virtue of their dependency on a rejected claim without alleviating the indefiniteness. For the sake of furthering examination this limitation will be interpreted as being an intended use of the method.
Claim 18 recites “the operator is trained to predict energy and have its functional derivative approach a zero field with respect to charge density distribution” which renders the metes and bounds of the claim indefinite. Indefiniteness arises because it is unclear if “its functional derivative approach a zero field…” is referring to the energy being predicted or if “its functional derivative approach a zero field…” is referring to the operator itself. Thus, it is unclear which limitation of the claim has its functional derivation approach a zero field with respect to charge density distribution. The specification does not provide a clear and precise definition of the limitation, nor would one skilled in the art recognize the metes and bounds of said limitation. For the sake of furthering examination this limitation will be interpreted as “the equivariant operator is trained to predict energy that converges to a final predicted energy”.
112/b: Indefiniteness based on 112/f interpretation:
Claim limitation “a density functional theory engine to calculate a self-consistent field” in claims 1 and 19 invokes 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. However, the written description fails to disclose the corresponding structure, material, or acts for performing the entire claimed function and to clearly link the structure, material, or acts to the function.
Claims 1 and 19 recite “a density functional theory engine to calculate a self-consistent field”. The instant disclosure provides “DFT engine 200 uses charge density distribution 190 o run at least one round of SCF to compute final charge density distribution 210 and estimated error in charge density distribution 190. Optionally, multiple rounds of SCF may also be run for DFT convergence by feeding final charge density distribution 210 back into DFT engine 200 multiple times”. The instant disclosure does not provide an associated structure (i.e., computer hardware and algorithm) for performing the claimed function of a density theory engine to calculate a self-consistent field (e.g., does not provide a corresponding structure such as steps taken to calculate a self-consistent field). Dependent claims 2-18 and 20-26 are rejected by virtue of their dependency on a rejected claim without alleviating the indefiniteness. For the sake of furthering examination, claims 1 and 19 will be interpreted as “inputting the initial charge density distribution into a self-consistent field process” and claims 2 and 20 will be interpreted as “iteratively repeating the self-consistent field process until convergence”.
Therefore, the claim is indefinite and is rejected under 35 U.S.C. 112(b) or pre-AIA 35 U.S.C. 112, second paragraph.
Applicant may:
(a) Amend the claim so that the claim limitation will no longer be interpreted as a limitation under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph;
(b) Amend the written description of the specification such that it expressly recites what structure, material, or acts perform the entire claimed function, without introducing any new matter (35 U.S.C. 132(a)); or
(c) Amend the written description of the specification such that it clearly links the structure, material, or acts disclosed therein to the function recited in the claim, without introducing any new matter (35 U.S.C. 132(a)).
If applicant is of the opinion that the written description of the specification already implicitly or inherently discloses the corresponding structure, material, or acts and clearly links them to the function so that one of ordinary skill in the art would recognize what structure, material, or acts perform the claimed function, applicant should clarify the record by either:
(a) Amending the written description of the specification such that it expressly recites the corresponding structure, material, or acts for performing the claimed function and clearly links or associates the structure, material, or acts to the claimed function, without introducing any new matter (35 U.S.C. 132(a)); or
(b) Stating on the record what the corresponding structure, material, or acts, which are implicitly or inherently set forth in the written description of the specification, perform the claimed function. For more information, see 37 CFR 1.75(d) and MPEP §§ 608.01(o) and 2181.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-26 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
(Step 1)
Claims 1-18 fall under the statutory category of a process and claim 19-26 falls under the statutory category of a machine.
(Step 2A Prong 1)
Under the BRI, the instant claims recite judicial exceptions that are an abstract idea of the type that is in the grouping of a “mental process”, such as procedures for evaluating, analyzing or organizing information, and forming judgement or an opinion. The instant claims further recite judicial exceptions that are an abstract idea of the type that is in the grouping of a “mathematical concept”, such as mathematical relationships and mathematical equations.
Independent claims 1 and 19 recite mathematical concept of generating an initial charge density distribution from the set of atomic moieties and positions and inputting the initial charge density into a self-consistent field process.
Dependent claims 2 and 20 recite mathematical concepts of repeating self-consistent field process until convergence is achieved. Dependent claims 3, 4, 5, 7, 12-14, 16-18, 21, and 26 are interpreted as furthering limiting the abstract idea (i.e., mathematical concept and mental process) recited in the claims by further limiting the abstract content of the data being used as input for the analysis, the abstract output being produced by the analysis, and the content of the information being learned by the network. Dependent claim 8 recites a mathematical concept of inputting the produced charge density distribution into one or more property calculators. Dependent claim 9 further limits the mathematical concept in claim 8 by further limiting the one or more property calculators to include a force calculator and a multipole calculator. Dependent claim 23 recites a mathematical concept of inputting the charge density into one or more property calculators which include a force calculator and a multipole calculator (which is interpreted as performing a calculation using a property calculator). Dependent claim 15 recites a mathematical concept of iteratively refining the initial charge density distribution via gradient descent.
The claims recite mathematical concepts of mathematical calculations as generating an initial charge density distribution from the set of atomic moieties and positions (which encompasses calculating an initial charge density distribution based on atoms and positions of these atoms) and inputting the initial charge density into a self-consistent field process (which encompasses calculating a self-consistent field which is a process of refining a charge density distribution until convergence utilizing a series of mathematical calculations through multiple iterations), inputting the produced charge density distribution into one or more property calculators wherein the property calculators include is a force calculator and a multipole calculator (which encompasses a mathematical calculation for calculation these properties using the initial charge density distribution). Thus, claims 1-26 recite abstract ideas.
(Step 2A Prong 2)
Claims found to recite a judicial exception under Step 2A, Prong 1 are then further analyzed to determine if the claims as a whole integrate the recited judicial exception into a practical application or not (Step 2A, Prong 2). Integration into a practical application is evaluated by identifying whether there are any additional elements recited in the claim and evaluating those additional elements to determine whether they integrate the exception into a practical application.
The additional element in claims 1 of inputting data into a neural network and receiving data from a neural network, the additional element in claims 6 and 22 of wherein the equivariant neural network is equivariant in terms of rotation and translation, the additional element in claim 7 of inputting information into a scalar neural network, the additional element in claim 10 and 24 of the equivariant neural network implements an equivariant operator, the additional element in claims 11 and 25 a tensor field convolution linear layer, a local product by linear layer, and a local non-linear layer, and the additional element in claims 19 inputting data into an equivariant neural network and receiving data from an equivariant neural network do not integrate the judicial exceptions into a practical application because these additional elements amount to generally linking the judicial exceptions of generating a charge density distribution from a set of atomic moieties and positions to the particular technological environment of neural networks and equivariant neural networks (see MPEP 2106.05(h) and Example 47 claim 2). Further, these additional elements are not integrated into a practical application because they amount to mere instructions to apply the judicial exceptions to a generic computer (see MPEP 2106.05(f) and Example 47 claim 2).
The additional element in claim 19 of using a generic computer (i.e., a processor and software to cause the processor to perform judicial exceptions) does not integrate the judicial exceptions into a practical application because this is applying the judicial exceptions to a generic computer without an improvement to computer functionality/technology (see MPEP 2106.04(d)(1)). The generic computer only interacts with the recited judicial exceptions by being invoked as a tool to perform the judicial exceptions.
Thus, the additional elements do not integrate the judicial exceptions into a practical application and claims 1-26 are directed to the abstract idea.
(Step 2B)
Claims found to be directed to a judicial exception are then further evaluated to determine if the claims recite an inventive concept that provides significantly more than the judicial exception itself (Step 2B). The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception because:
The additional element in claims 1 of inputting data into a neural network and receiving data from a neural network, the additional element in claims 6 and 22 of wherein the equivariant neural network is equivariant in terms of rotation and translation, the additional element in claim 7 of inputting information into a neural network, the additional element in claim 10 and 24 of the equivariant neural network implements an equivariant operator which operates between tensor fields, the additional element in claims 11 and 25 a tensor field convolution linear layer, a local product by linear layer, and a local non-linear layer, and the additional element in claims 19 inputting data into an equivariant neural network and receiving data from an equivariant neural network are conventional as show by Thomas et al. (arXiv preprint arXiv:1802.08219 (2018)) (see pages 4-6 which include equivariant operators of an equivariant neural network with these layers, see pages 7-8 for processing information with a equivariant neural networks to learn mappings) and Batzner et al. (Preprint at https://arxiv. org/abs/2101.03164 v3 (2021)) (see pages 6, 7, and 8 Fig. 2 for equivariant neural network with operators implemented as these layers for processing information using neural networks and equivariant neural networks).
The additional element in claim 19 of using a generic computer (i.e., a processor and software to cause the processor to perform judicial exceptions) is conventional as shown by MPEP 2106.05(b) and MPEP 2106.05(d)(II)).
Thus, the additional elements are not sufficient to amount to significantly more than the judicial exception because they are conventional.
Claim Rejections - 35 USC § 102
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 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.
Claims 1- 3 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Kamal et al. (Machine Learning: Science and Technology, Volume 1, Number 2, 18 March 2020; cited in IDS received 27 September 2023).
Claim 1 is directed to a method comprising: inputting a set of atomic moieties and positions to a neural network
Kamal et al. shows inputting a set of atomic moieties and positions to an input layer of a neural network (Kamal et al. page 2 Fig. 1a and page 4 section 2.3 Fingerprint).
receiving, from the neural network, an initial charge density distribution
Kamal et al. shows receiving from the neural network an initial charge density distribution as the output of the neural network predicting the electronic charge density distribution of a molecule based on the set of atomic moieties and positions which define the molecule (Kamal et al. page 2 Fig. 1a).
and inputting the initial charge density distribution into a self-consistent field process.
Kamal et al. shows the charge density model can be used to quickly estimate the electronic charge density of new configurations and can even serve as a starting point for self-consistent computations involved in the KS-DFT routines (Kamal et al. page 2 last paragraph). Kamal et al. shows inputting the initial charge density distribution as an initial guess an approximate speedup was observed for self-consistent field convergence (SCF) which shows using the initial charge density distribution to calculate (SCF) (Kamal et al. page 8 paragraph 2).
Claim 2 is directed to repeating the self-consistent field process until convergence.
Kamal et al. shows inputting the initial charge density distribution as an initial guess an approximate speedup was observed for self-consistent field convergence (SCF) which shows using the initial charge density distribution to calculate (SCF) (Kamal et al. page 8 paragraph 2).
Claim 3 is directed to wherein the set of atomic moieties and positions is in the form of an array of scalars where each scalar in the array of scalars is associated with an atomic moiety of the set of atomic moieties
Kamal et al. shows that the set of atomic moieties and positions is in the form of an array of scalars (the array of S1,1,… Sk,1) which each depend on the respective atomic moiety which is used as input to the neural network (Kamal et al. page 2 Fig 1(a)).
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 4-7 and 10-13 are rejected under 35 U.S.C. 103 as being unpatentable over Kamal et al. (Machine Learning: Science and Technology, Volume 1, Number 2, 18 March 2020; cited in IDS received 27 September 2023) as applied to claims 1 and 3 under 35 U.S.C. 102 above, and further in view of Batzner (Preprint at https://arxiv. org/abs/2101.03164 v3 (2021)).
Claim 4 is directed to wherein the neural network is an equivariant neural network outputting a set of tensor features representing the charge density distribution at each atomic position.
Kamal et al. does not show wherein the neural network is an equivariant neural network outputting a set of tensor features representing the charge density distribution at each atomic position.
Like Kamal et al., Batzner et al. shows utilizing a neural network to learn a mapping between atomic moieties and positions to properties using quantum chemical level theory. Batzner et al. shows every atom in the equivariant neural network is associated with a feature, which are tensors of different order: scalars, vectors, and higher-order tensors, that are geometric objects that convolutions operate on (Batzner et al. page 5 left col.). Batzner et al. further shows that after every convolution, output tensors of a particular rotation order stemming from different tensor products are concatenated on a per-atom basis which shows outputting a set of tensor features representing the predicted value at each atomic position (Batzner et al. page 7 left col.).
Claim 5 is directed to wherein the set of tensor features is computed at each atomic position via an equivariant convolution.
Batzner et al. shows that the convolutions are equivariant functions (i.e., if a feature at a layer is rotated, then the output of the convolution from the layer rotates accordingly) (Batzner et al. page 5 left col.).
Claims 6 is directed to wherein the neural network is equivariant in terms of rotation and translation.
Batzner et al. shows that the neural network is equivariant with respect to SE(3) which is the group of rotations and translations in 3D space (Batzner et al. page 4 right col.).
Claim 7 is directed to inputting the set of tensor features to a scaler neural network to predict charge density at each atomic position.
Batzner et al. shows the output block (which is interpreted as a scaler neural network) intakes the set of tensor features produced by the series of interactions blocks to produce a scalar atomic output representing a predicted value at each atomic position (Batzner et al. page 7 right col. and page 8 Fig. 2).
Claims 10 is directed to wherein the neural network implements an equivariant operator operating between sets of tensor fields. Claims 11 is directed to wherein the equivariant operator is implemented as a composition of a tensor field convolution linear layer, a local product by linear layer and a local nonlinear layer.
Batzner et al. shows that the neural network implements interaction blocks, each interaction block includes a composition of several layers, that allow for interactions between sets of tensor fields (Batzner et al. page 5 left col. and page 8 Fig. 2). Batzner et al. shows the layers include a convolutional layer which processes tensor fields to produce a tensor product, a self-interaction layer which is interpreted as being a local product by linear layer, and a non-linearity layer (Batzner et al. page 8 Fig. 2).
Claim 12 is directed to wherein the output of the neural network is a convolution of the input and a characteristic impulse response. Claim 13 is directed to wherein the impulse response is a product of a scaler radial function and a spherical harmonic, such that the neural network is rotationally equivariant.
Batzner et al. shows that the equivariant neural network utilizes convolutional filters, which performs a convolution operation on the input and the values/weights that make up the convolutional filter, that are constrained to be a product of learnable radial functions and spherical harmonics which provides rotational equivariance (Batzner et al. page 5 right col. and page 8 Fig. 2).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the invention to have modified the neural network used to learn the mapping between a set of atomic moiety and positions and charge density distribution of Kamal et al. to use the equivariant neural network Batzner et al. because this would allow for a process to employ a computationally and data efficient equivariant neural network (i.e., relies on a smaller set of reference samples for training) to efficiently predict a charge density distribution which is a fundamental quantity that may be used in downstream self-consistency calculations (Batzner et al. page 14 section “computational efficiency” and page 14 right col.). One would have reasonable expectation of success for this modification because Kamal et al. shows the ability of a neural network model to learn the mapping between an input of atomic moieties and positions and the output of an electron charge density distribution (Kamal et al. page 2 figure 1) and Batzner et al. shows that the equivariant neural network learns the mapping between atomic moieties and positions to an energy which can be extended to other quantities of interest by adjusting the loss function during training (Batzner et al. page 4 right col. and page 17 left col.).
Claims 8 and 9 are rejected under 35 U.S.C. 103 as being unpatentable over Kamal et al. in as applied to claim 1 under 35 U.S.C. 102 above, and further in view of Veit et al. (The Journal of chemical physics 153.2 (2020)).
Claim 8 is directed to inputting the charge density distribution produced by the neural network to one or more property calculators. Claim 9 is directed to wherein the one or more property calculators include a forces calculator and a multipole moments calculator.
Kamal et al. does not explicitly show inputting the initial charge density distribution produced by the neural network into one or more property calculators which includes a forces calculator and a multipole moments calculator.
Like Kamal et al., Veit et al. shows processing charge density distribution information for a molecule. Veit et al. shows inputting an electron charge density distribution into a dipole moment calculator which is interpreted as particular type multipole moments calculator (Veit et al. page 3 of the document left col.).
It would have been obvious to one or ordinary skill in the art before the effective filling date of the invention to have combined the neural network that efficiently estimates an initial charge density distribution of Kamal et al. with the dipole moments calculator which utilizes a charge density distribution as input of Veit et al. because this would provide a comprehensive process which can efficiently predict the initial electron charge density distribution and then use the initial charge density distribution to predict an observable property of a dipole moment for a molecule (Veit et al. page 3 of the document left col.). One would have a reasonable expectation of success because Kamal et al. shows producing an initial charge density distribution and provides that the model may be used to predict properties like dipoles and quadrupoles moments in structures (Kamal et al. page 2 last paragraph - page 3 first paragraph) while Veit et al. shows a dipole calculator, as a mathematical equation, to calculate a dipole moment of a molecule using the charge density distribution of a molecule.
Claims 14-18 are rejected under 35 U.S.C. 103 as being unpatentable over Kamal et al. in view of Batzner et al. as applied to claim 6 above, and further in view of Li et al. (arXiv:2009.08551v2, 17 Nov. 2020).
Claims 14 is directed to wherein the equivariant neural network implements an equivariant operator to map a scalar field of electronic density to a scalar field of total energy.
Kamal et al. in view of Batzner et al. shows an equivariant neural network which implements equivariant operators to map between atomic moieties and positions and a scalar field of electronic density. Kamal et al. in view of Batzner et al. does not show that the equivariant operator implemented by the equivariant neural network maps a scalar field of electronic density to a scalar field of total energy.
Like Kamal et al. in view of Batzner et al., Li et al. shows using a neural network to perform density functional theory calculations. Li et al. shows the ability of a neural network to learn the mapping between an initial electronic charge density and total energy (Li et al. page 2 Fig. 2).
Claim 15 is directed to iteratively refining the initial charge density distribution via gradient descent.
Li et al. shows iteratively refining the initial charge density distribution (Li et al. page 2 Fig. 2, page 3 left col., and supplemental material page 3 section D).
Claim 16 is directed to wherein external nuclear potential, mean field, and approximate exchange-correlation energies are directly calculated.
The BRI of the claim is interpreted as being an intended use of the method. Kamal et al. in view of Batzner et al. shows the ability of a equivariant neural network model to predict quantities of interest by adding them to the loss function during training which indicates the ability for this neural network to include other values to be predicted (Batzner et al. page 4 right col. and page 17 left col.).
Claim 17 is directed to wherein a trainable portion of the equivariant operator learns a deviation of the gradient descent. Claim 18 is directed to wherein the equivariant operator is trained to predict energy that converges to a final predicted energy.
Li et al. shows a neural network, which includes an operator in the form of multiple layers, is trained to perform KS-DFT self-consistent calculations to iteratively refine the electronic charge density which is interpreted as the layers learn a deviation of the gradient descent which iteratively refines the initial charge density distribution in self-consistent field calculations (Li et al. page 2 left col. and page 2 Fig. 2). Li et al. shows a neural network, which includes an operator in the form of multiple layers, is trained to perform KS-DFT self-consistent calculations to predict energy (Li et al. page 2 left col. and page 2 Fig. 2). Li et al. shows that the neural network, which includes an operator in the form of multiple layers, is trained to predict energy which converges to a final predicted energy (Li et al. page 3 left col.).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the invention to modify the equivariant neural network that includes an equivariant operator to learn the mapping between atomic moieties and charge density distribution of Kamal et al. in view of Batzner et al. to incorporate an additional operator with the ability to map electronic density to total energy show in Li et al. because this would provide a comprehensive process of using an equivariant neural network with multiple operators to map a molecule to an initial electronic density then map this electronic density to total energy of molecule with the ability to quickly converge to a final predicted energy (Li et al. page 2 Fig. 2). One would have a reasonable expectation of success because Kamal et al. in view of Batzner et al. shows the ability of the equivariant operators to learn mappings while Li et al. shows the use of operators to map electronic charge density to total energy.
Claims 19-22, 24, and 25 are rejected under 35 U.S.C. 103 as being unpatentable over Kamal et al. (Machine Learning: Science and Technology, Volume 1, Number 2, 18 March 2020; cited in IDS received 27 September 2023) in view of Batzner (Preprint at https://arxiv. org/abs/2101.03164 v3 (2021)).
Claim 19 is directed to a system comprising: a processor; and software, which, when executed on the processor, causes the system to perform the functions of: inputting a set of atomic moieties and positions to an equivariant neural network
Kamal et al. shows inputting a set of atomic moieties and positions to an input layer of a neural network (Kamal et al. page 2 Fig. 1a).
receiving, from the neural network, an initial charge density distribution
Kamal et al. shows receiving from the neural network an initial charge density distribution as the output of the neural network predicting the electronic charge density distribution of a molecule based on the set of atomic moieties and positions which define the molecule (Kamal et al. page 2 Fig. 1a).
and inputting the initial charge density distribution into a self-consistent field process.
Kamal et al. shows the charge density model can be used to quickly estimate the electronic charge density of new configurations and can even serve as a starting point for self-consistent computations involved in the KS-DFT routines (Kamal et al. page 2 last paragraph). Kamal et al. shows inputting the initial charge density distribution as an initial guess an approximate speedup was observed for self-consistent field convergence (SCF) which shows using the initial charge density distribution to calculate (SCF) (Kamal et al. page 8 paragraph 2).
Kamal et al. does not show an equivariant neural network.
Like Kamal et al., Batzner et al. shows utilizing a neural network to learn a mapping between atomic moieties and positions to properties using quantum chemical level theory. Batzner et al. shows that the equivariant neural network learns the mapping between atomic moieties and positions to an energy which can be extended to other quantities of interest by adjusting the loss function during training (Batzner et al. page 4 right col. and page 17 left col.).
Claim 20 is directed to repeating the self-consistent field process until convergence.
Kamal et al. shows inputting the initial charge density distribution as an initial guess an approximate speedup was observed for self-consistent field convergence (SCF) which shows using the initial charge density distribution to calculate (SCF) (Kamal et al. page 8 paragraph 2).
Claim 21 is directed to wherein the set of atomic moieties and positions is in the form of an array of scalars where each scalar in the array of scalars is associated with an atomic moiety of the set of atomic moieties, and further wherein the equivariant neural network outputs a set of tensor features representing the charge density distribution at each atomic position.
Kamal et al. shows that the set of atomic moieties and positions is in the form of an array of scalars (the array of S1,1,… Sk,1) which each depend on the respective atomic moiety (Kamal et al. page 2 Fig 1(a)).
Kamal et al. shows that wherein the equivariant neural network outputs a set of tensor features representing the charge density distribution at each atomic position
Batzner et al. further shows that after every convolution, output tensors are concatenated on a per-atom basis which shows outputting a set of tensor features representing the predicted value at each atomic position (Batzner et al. page 7 left col.).
Claim 22 is directed to wherein the neural network is equivariant in terms of rotation and translation.
Batzner et al. shows that the neural network is equivariant with respect to SE(3) which is the group of rotations and translations in 3D space (Batzner et al. page 4 right col.).
Claim 24 is directed to wherein the neural network implements an equivariant operator operating between sets of tensor fields. Claim 25 is directed to wherein the equivariant operator is implemented as a composition of a tensor field convolution linear layer, a local product by linear layer and a local nonlinear layer.
Batzner et al. shows that the neural network implements interaction blocks, each interaction block includes a composition of several layers, that allow for interactions between sets of tensor fields (Batzner et al. page 5 left col. and page 8 Fig. 2). Batzner et al. shows the layers include a convolutional layer which processes tensor fields to produce a tensor product, a self-interaction layer which is interpreted as being a local product by linear layer, and a non-linearity layer (Batzner et al. page 8 Fig. 2).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the invention to have modified the neural network that learns the mapping between a set of atomic moiety and positions and charge density distribution of Kamal et al. to use the equivariant neural network Batzner et al. because this would allow for a system that employs a computationally and data efficient equivariant neural network (i.e., relies on a smaller set of reference samples for training) to efficiently predict a charge density distribution which is a fundamental quantity that may be used in downstream self-consistency calculations (Batzner et al. page 14 section “computational efficiency” and page 14 right col.). One would have reasonable expectation of success for this modification because Kamal et al. shows the ability of a neural network model to learn the mapping between an input of atomic moieties and positions and the output of an electron charge density distribution (Kamal et al. page 2 figure 1) and Batzner et al. shows that the equivariant neural network learns the mapping between atomic moieties and positions to an energy which can be extended to other quantities of interest by adjusting the loss function during training (Batzner et al. page 4 right col. and page 17 left col.).
Claim 23 is rejected under 35 U.S.C. 103 as being unpatentable over Kamal et al. in view of Batzner et al. as applied to 19 above, and further in view of Veit et al. (The Journal of chemical physics 153.2 (2020)).
Claim 23 is directed to input the charge density distribution produced by the equivariant neural network to one or more property calculators, the one or more property calculus including a forces calculator and a multipole moments calculator.
Kamal et al. in view of Batzner et al. does not explicitly show inputting the initial charge density distribution produced by the neural network into one or more property calculators which includes a forces calculator and a multipole moments calculator.
Like Kamal et al. in view of Batzner et al., Veit et al. shows processing charge density distribution information for a molecule. Veit et al. shows inputting an electron charge density distribution into a dipole moment calculator which is interpreted as particular type multipole moments calculator (Veit et al. page 3 of the document left col.).
It would have been obvious to one or ordinary skill in the art before the effective filling date of the invention to have combined the neural network that efficiently estimates an initial charge density distribution of Kamal et al. in view of Batzer et al. with the dipole moments calculator which utilizes a charge density distribution as input of Veit et al. because this would provide a comprehensive process which can efficiently predict the initial electron charge density distribution and then use the initial charge density distribution to predict an observable property of a dipole moment for a molecule (Veit et al. page 3 of the document left col.). One would have a reasonable expectation of success because Kamal et al. in view of Batzer et al. shows producing an initial charge density distribution and provides that the model may be used to predict properties like dipoles and quadrupoles moments in structures (Kamal et al. page 2 last paragraph - page 3 first paragraph) while Veit et al. shows a dipole calculator, as a mathematical equation, to calculate a dipole moment of a molecule using the charge density distribution of a molecule.
Claims 26 is rejected under 35 U.S.C. 103 as being unpatentable over Kamal et al. in view of Batzner et al. as applied to claim 24 above, and further in view of Li et al. (arXiv:2009.08551v2, 17 Nov. 2020).
Claims 26 is directed to wherein the equivariant neural network implements an equivariant operator to map a scalar field of electronic density to a scalar field of total energy.
Kamal et al. in view of Batzner et al. shows an equivariant neural network which implements equivariant operators to map between atomic moieties and positions and a scalar field of electronic density. Kamal et al. in view of Batzner et al. does not show that the equivariant operator implemented by the equivariant neural network maps a scalar field of electronic density to a scalar field of total energy.
Like Kamal et al. in view of Batzner et al., Li et al. shows using a neural network to perform density functional theory calculations. Li et al. shows the ability of a neural network to learn the mapping between an initial electronic charge density and total energy (Li et al. page 2 Fig. 2).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the invention to modify the equivariant neural network that includes an equivariant operator to learn the mapping between atomic moieties and charge density distribution of Kamal et al. in view of Batzner et al. to incorporate an additional operator with the ability to map electronic density to total energy show in Li et al. because this would provide a comprehensive process of using an equivariant neural network with multiple operators to map a molecule to an initial electronic density then map this electronic density to total energy of molecule with the ability to quickly converge to a final predicted energy (Li et al. page 2 Fig. 2). One would have a reasonable expectation of success because Kamal et al. in view of Batzner et al. shows the ability of the equivariant operators to learn mappings while Li et al. shows the use of operators to map electronic charge density to total energy.
Conclusion
No claims are allowed.
This Office action is a Non-Final action. A shortened statutory period for reply to this action is set to expire THREE MONTHS from the mailing date of this action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JONATHAN EDWARD HAYES whose telephone number is (571)272-6165. The examiner can normally be reached M-F 9am-5pm.
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, Olivia Wise can be reached at 571-272-2249. 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.
/JONATHAN EDWARD HAYES/Examiner, Art Unit 1685
Prosecution Insights