Detailed Action
This action is in response to the RCE filed on 03/27/2026 for the amended claims filed 03/27/2026 for application 17/696,685, in which:
Claims 1, 8, and 15 are independent claims.
Claims 6, 13, and 19 are cancelled.
Claims 1, 8, and 15 are currently amended.
Claim 21 has been newly added.
Claims 1-5, 7-12, 14-18, and 20-21 are currently pending.
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 03/27/2026 has been entered.
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 .
Response to Arguments
Applicant's arguments filed 03/27/2026 have been fully considered but they are not persuasive.
Regarding the 35 USC § 101 Rejections:
Applicant's arguments regarding the 35 U.S.C. 101 rejections of the previous office action have been fully considered, but are unpersuasive.
Applicant asserts (Pages 8-9), that the Claims have been amended to advance prosecution over the 101 rejections and supported via the specification.
Examiner respectfully disagrees. 35 U.S.C. § 101 rejections for the amended claims are directed to an abstract idea (Step 2A Prong 1) and do not integrate the abstract idea into a practical application (Step 2A Prong 2). The claims recites abstract ideas a-g; where the abstract ideas are evaluations/judgements that can be performed in the human mind (or by a human using pen and paper). The independent claims are no more detailed than accessing/manipulating data, performing abstract ideas and implementing updates within a computer device where the computer implemented method performs the access, adding, performing, selecting, and updating based on determinations; thus, the Claims are not a technical solution to a technical problem as the independent claim is merely performing abstract ideas with specific restrictions within a computer. The additional elements noted within Step 2A Prong 2 are unable to amount to significantly more than the judicial exception (when evaluated individually and holistically) as they fall within MPEP 2106.05. Thus, the additional elements are not able to integrate the abstract ideas in a practical application. The claims are directed towards the improvement of an abstract idea. Therefore, the claims do not integrate the judicial exception into a practical application. For the reasons given above and in the rejections below, the rejection to all Claims (including Claim 1, similar independent claims, and all dependent Claims) are maintained. More specific details are discussed below within the 35 USC § 101 Rejections.
Applicant asserts (Pages 9-12), that the claims recite limitations that cannot practically be performed in the human mind. Therefore, the claimed limitations are not directed to "mental processes" under 35 USC 101, and are not directed to a judicial exception. The examiner is reminded to consult the specification to determine whether the disclosed invention improves technology or a technical field, and evaluate the claim to ensure it reflects the disclosed improvement. The specification does not need to explicitly set forth the improvement, but it must describe the invention such that the improvement would be apparent to one of ordinary skill in the art. The claim itself does not need to explicitly recite the improvement described in the specification. Therefore, Applicant submits that the specification discloses an improvement to technology. The claims describe a particular solution and a particular way to efficiently recommend a model and hyperparameters for a given dataset. The claim as a whole integrates a judicial exception into a practical application since it improves the functioning of a computer or other technology, and is therefore directed to patent eligible subject matter.
Examiner respectfully disagrees. Applicant's arguments fail to specifically point out how the language of the claims cannot be practically performed in the human mind and merely states they are not able to be performed. The pending Claims are directed to a judicial exception due to reciting limitations which fall within the “mental processes” group of abstract ideas; where the judicial exception is unable to be directed to significantly more than the judicial exception due to the pending Claims not including additional elements that contribute to an “inventive concept”. Due to the additional elements of the independent claims falling under MPEP 2106.05, the judicial exception is not integrated into a practical application and the specific details are discussed below within the 35 USC § 101 Rejections. The claims are directed towards the improvement of an abstract idea. Improvements to an abstract idea are still considered to an abstract idea. Additionally, the Claims does not reflect alleged improvement in the functioning of a computer or hardware processor. Therefore, the claims do not integrate the judicial exception into a practical application nor amount to significantly more. The office action establishes a proper and well-supported prima facie case as the claims are explained to be not patentable via the Patent Subject Matter Eligibility steps within MPEP 2106; thus, the additional elements noted within Step 2A Prong 2 are unable to amount to significantly more than the judicial exception (when evaluated individually and holistically). The limitations are unable to provide the alleged improvement as they are currently being evaluated as either abstract idea(s) or additional elements that fall within MPEP 2106.05. The claim is not patent eligible. Although the Claims are interpreted in light of the specification, limitations from the specification are not read into the Claims.
MPEP 2106.05(a) recites:
After the examiner has consulted the specification and determined that the disclosed invention improves technology, the claim must be evaluated to ensure the claim itself reflects the disclosed improvement in technology … the claim must include the components or steps of the invention that provide the improvement described in the specification
…
It is important to note, the judicial exception alone cannot provide the improvement. The improvement can be provided by one or more additional elements. See the discussion of Diamond v. Diehr, 450 U.S. 175, 187 and 191-92, 209 USPQ 1, 10 (1981)) in subsection II, below.
Applicant fails to show how any alleged technical improvement would be provided by anything more than the judicial exception on its own. Additionally, applicant fails to show how the claim includes components or steps that would provide the alleged improvement described in the specification. By MPEP 2106.05(f)(1), "the claim recites only the idea of a solution or outcome, i.e. the claim fails to recite details of how a solution to a problem is accomplished". The rejection follows the steps of the analysis as laid out in the MPEP which was followed for the previous and current examination (see MPEP 2106). Moreover, the examiner maintains that the Claim does not impose any meaningful limits on the judicial exception. As noted in the rejection, the Claim does not include additional elements that are sufficient to amount to an integration of the identified abstract idea(s) into a practical application, thus the claim is directed to an abstract idea.
Applicant asserts (Page 13), that at least the following limitations are not conventional, well understood, or routine: for each iterative step: determining, by the computing device, a selection probability for each model node; wherein determining the selection probability for each model node comprises: determining a selection weight, and determining a plurality of selection criteria; selecting, by the computing device, a particular model node, the particular model node being selected based at least in part on the selection probability, wherein the model node comprises hyperparameters, wherein the hyperparameters are configured to control a structure of a model; updating, by the computer device, the selection criteria based at least in part on the particular model node; and updating, by the computer device, at least one of the plurality of features or the plurality of weights based at least in part on the particular model node; and training, by the computer device, a selected model from a last iterative step on at least one dataset that is represented by the one or more test dataset nodes. Therefore, Applicant submits that the claims recite significantly more than the judicial exception. For at least the above reasons, Applicant respectfully requests reconsideration of the 35 USC 101 rejection.
Examiner agrees as the noted limitations are not listed as conventional, well understood, or routine. The only limitation that falls within this category, which is noted within the office action, is the limitation accessing … a graph comprising one or more model nodes, one or more dataset nodes, and one or more edges, wherein the one or more model nodes having at least one of a plurality of features or a plurality of weights … which is not noted within the limitations recited within the above remarks. Applicant's arguments amount to a general allegation that the recited limitations are not conventional, well understood, or routine without specifically pointing out how the language of the recited limitations are not conventional, well understood, or routine. Applicant’s arguments regarding the other independent and dependent claims rely upon the same assertions as with respect to Claim 1, and are thus likewise unpersuasive. Therefore, for the reasons given above and in the rejections below, the rejection to all Claims (including Claim 1, similar independent claims, and all dependent Claims) are maintained. More specific details are discussed below within the 35 USC § 101 Rejections.
Regarding the 35 USC § 103 Rejections:
Applicant's arguments regarding the 35 U.S.C. 103 rejections of the previous office action have been fully considered, but are unpersuasive.
Applicant asserts (Page 14), Page 1, paragraph 2 of Luxon describes constructing an initial molecular system. Page 34 describes graph structures that enable inter and intra model prediction evaluations. However, this does not teach or suggest performing a series of iterative steps until a threshold is reached, that for each iterative step, a selection probability for each model node is determined, or that determining the selection probability for each model includes determining a selection weight and determining a plurality of selection criteria. Further, the cited art does not teach or suggest the claimed hyperparameters in combination with the claimed limitations. Specifically, the cited art does not teach or suggest wherein the model node comprises hyperparameters, wherein the hyperparameters are configured to control a structure of a model as now recited in claim 1.
Examiner respectfully disagrees. As noted by Luxon within Paragraph 2 within Page 1 that the “The design-test-build sequence is iterated upon until the system performance requirements are satisfied”. The series of iterative steps for optimizing performance and design is done until performance requirements are satisfied; thus, Luxon explicitly teaches performing a series of iterative steps. Luxon also does teach the newly amended limitation that the model nodes comprise hyperparameters as noted in the updated rejection. However, the hyperparameters configured to control the structure of the model is not explicitly stated. However, the newly referenced prior art from Wang does teach … wherein the hyperparameters are configured to control a structure of a model … as the training leads to adaptively controlling the learning rates to update model parameters which controls the structure of the model. Thus, amended Claim 1 is taught by the Luxon/Needham/Färber and newly cited Wang. More specific details are discussed below within the 35 USC § 101 Rejections.
Applicant asserts (Pages 14-15), for at least the above reasons, claim 1 and its dependent claims should be deemed allowable. To the extent that independent claims 8 and 15 recite similar subject matter, claims 8 and 15 and their dependent claims should be deemed allowable for at least the same reasons. Claim 21 should be deemed allowable by virtue of its dependency on independent claim 1 for at least the reasons set forth above. Further, Applicant submits that the art cited by the Examiner does not teach or suggest the limitations of claim 21. Therefore, claim 21 should be deemed allowable.
Examiner respectfully disagrees. Applicant’s arguments regarding the other independent and dependent claims rely upon the same assertions as with respect to Claim 1, and are thus likewise unpersuasive. Therefore, for the reasons given above and in the rejections below, the rejection to all Claims (including Claim 1, similar independent claims, and all dependent Claims) are maintained. More specific details are discussed below within the 35 USC § 101 Rejections.
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 1-5, 7-12, 14-18, and 20-21 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding Claim 1:
Subject Matter Eligibility Analysis Step 1:
Claim 1 recites a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 1 further recites the method comprising of:
adding … one or more test dataset nodes and one or more test edges to the graph (a human being can mentally apply evaluation to add test dataset nodes and test edges to a graph)
performing … a series of iterative steps until a threshold is reached (a human being can mentally apply evaluation to perform a series of iterative steps until a threshold is reached)
for each iterative step: determining … a selection probability for each model node … (a human being can mentally apply evaluation to determine a selection probability for each model node)
… wherein the determining the selection probability for each model node comprises: determining a selection weight, and determining a plurality of selection criteria (a human being can mentally apply evaluation to determine a specific selection weight and a plurality of specific selection criteria)
selecting … a particular model node, the particular model node being selected based at least in part on the selection probability (a human being can mentally apply evaluation to select a model node based on the selection probability)
updating … the selection criteria based at least in part on the particular model node (a human being can mentally apply evaluation to update the selection criteria based on a specific model node)
updating … at least one of the plurality of features or the plurality of weights based at least in part on the particular model node (a human being can mentally apply evaluation to update a plurality of features or weights based on a particular model node)
Claim 1 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the additional elements recited consists of:
A computer-implemented method, comprising: (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h))
accessing … a graph comprising one or more model nodes, one or more dataset nodes, and one or more edges, wherein the one or more model nodes having at least one of a plurality of features or a plurality of weights … (which is insignificant extra-solution activity of data gathering, by MPEP 2106.05(g))
… by a computing device … (to perform a mental process and the performance of an abstract idea on a computer is no more than instructions to “apply it” on a computer, by MPEP 2106.05(f))
… for a machine learning model that is represented by a model node of the one or more model nodes, and wherein an edge weight of an edge that connects two dataset nodes represents a similarity of datasets represented by the two dataset nodes (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h))
wherein the model node comprises hyperparameters, wherein the hyperparameters are configured to control a structure of a model (to perform a mental process and the performance of an abstract idea on a computer is no more than instructions to “apply it” on a computer, by MPEP 2106.05(f))
training … a selected model from a last iterative step on at least one dataset that is represented by the one or more test dataset nodes (to perform a mental process and the performance of an abstract idea on a computer is no more than instructions to “apply it” on a computer, by MPEP 2106.05(f))
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements recited, alone or in combination, do not provide significantly more than the abstract idea itself. Additional elements a and d are only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which cannot provide significantly more. Additional element b falls within MPEP 2106.05(d) as well-understood, routine and conventional activities of receiving or transmitting data over a network (MPEP 2106.05(d)(II): buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014)). Additional elements c and e-f are merely applying the abstract idea on a computer (MPEP 2106.05(f)) which cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claim 2:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 2 recites the method of Claim 1. Claim 1 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 2 does not recite any additional abstract ideas and only inherits the abstract ideas from Claim 1. Claim 2 thus recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the sole additional element recited consists of the plurality of features comprise a plurality of hyperparameters (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h)).
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception because the new sole additional element recited, alone or in
combination, does not provide significantly more than the abstract idea itself. The additional element is
only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which
cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claim 3:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 3 recites the method of Claim 1. Claim 1 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 3 does not recite any additional abstract ideas and only inherits the abstract ideas from Claim 1. Claim 3 thus recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the sole additional element recited consists of the threshold can be at least one of a time period, a central processing unit (CPU) allocation, or a random access memory (RAM) allocation (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h)).
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception because the new sole additional element recited, alone or in
combination, does not provide significantly more than the abstract idea itself. The additional element is
only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which
cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claim 4:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 4 recites the method of Claim 1. Claim 1 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 4 does not recite any additional abstract ideas and only inherits the abstract ideas from Claim 1. Claim 4 thus recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the sole additional element recited consists of a model node, from the one or more model nodes, and a first dataset node, from the one or more dataset nodes, are connected by a model edge, of the one or more edges, in accordance with the dataset having been evaluated by a model that is represent by the model node (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h)).
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception because the new sole additional element recited, alone or in
combination, does not provide significantly more than the abstract idea itself. The additional element is
only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which
cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claim 5:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 5 recites the method of Claim 4. Claim 4 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 5 does not recite any additional abstract ideas and only inherits the abstract ideas from Claim 4. Claim 5 thus recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the sole additional element recited consists of the model edge has a selection weight, the selection criteria being based at least in part on the selection weight (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h)).
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception because the new sole additional element recited, alone or in
combination, does not provide significantly more than the abstract idea itself. The additional element is
only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which
cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claim 7:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 7 recites the method of Claim 4. Claim 4 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 7 further recites … determining a similarity weight for the one or more test edges (a human being can mentally apply evaluation to determine a similarity weight for test edges). Claim 7 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the sole additional element recited consists adding the one or more test edges further comprises … (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h)).
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception because the new sole additional element recited, alone or in
combination, does not provide significantly more than the abstract idea itself. The additional element is
only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which
cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claims 8-12:
Claims 8-12 incorporate substantively all the limitations of Claims 1-5 in a non-transitory computer-readable storage medium (thus, a manufacture) and further recites storing a set of instructions, that, when executed by one or more processors of a recommendation system computing device, cause the one or more processors to perform instructions comprising (these claim limitations appear to perform a mental process and the performance of an abstract idea on a computer is no more than instructions to “apply it” on a computer, by MPEP 2106.05(f)) and does not appear to integrate the abstract idea into a particular application; thus, the claim is subject-matter ineligible as it does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements, alone or in combination, do not provide significantly more than the abstract idea itself); thus, Claims 8-12 are rejected for reasons set forth in the rejections of Claims 1-5, respectively.
Regarding Claim 14:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 14 recites the non-transitory computer-readable storage medium of Claim 9. Claim 9 is a non-transitory computer-readable storage medium, thus a manufacture, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 14 further recites … determining a similarity weight for the one or more test edges (a human being can mentally apply evaluation to determine a similarity weight for test edges). Claim 14 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the sole additional element recited consists adding the one or more test edges further comprises … (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h)).
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception because the new sole additional element recited, alone or in
combination, does not provide significantly more than the abstract idea itself. The additional element is
only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which
cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claims 15-18 and 20:
Claims 15-18 and 20 incorporate substantively all the limitations of Claims 1-2, 4-5 and 7 in a system (thus, a machine) and further recites comprising: memory storing computer-executable instructions; and one or more processors configured to access the memory, and execute the computer-executable instructions to at least (these claim limitations appear to perform a mental process and the performance of an abstract idea on a computer is no more than instructions to “apply it” on a computer, by MPEP 2106.05(f)) and does not appear to integrate the abstract idea into a particular application; thus, the claim is subject-matter ineligible as it does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements, alone or in combination, do not provide significantly more than the abstract idea itself); thus, Claims 15-18 and 20 are rejected for reasons set forth in the rejections of Claims 1-2, 4-5 and 7, respectively.
Regarding Claim 21:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 21 recites the method of Claim 2. Claim 2 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 21 does not recite any additional abstract ideas and only inherits the abstract ideas from Claim 2. Claim 21 thus recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the sole additional element recited consists of wherein the plurality of hyperparameters comprise parameters that tum a learning algorithm into a trained model, and wherein the plurality of hyperparameters are set before training the learning algorithm starts (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h)).
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception because the new sole additional element recited, alone or in
combination, does not provide significantly more than the abstract idea itself. The additional element is
only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which
cannot provide significantly more. Thus, the claim is subject-matter ineligible.
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.
Claims 1-5, 7-12, 14-18, and 20-21 are rejected under 35 U.S.C. 103 as being unpatentable over Luxon et al., “Information Architecture for a Chemical Modeling Knowledge Graph”, in view of Needham et al., “A Comprehensive Guide to Graph Algorithms in Neo4j”, in view of Färber et al., “The data set knowledge graph: Creating a linked open data source for data sets”, in view of Wang et al., “KGAT: Knowledge Graph Attention Network for Recommendation”.
Regarding Claim 1:
Luxon teaches:
A computer-implemented method, comprising:
(Luxon, Page 25, Figure 6; Page 30, Paragraph 1, “ … The pipeline supports parallel computing across multiple CPU threads for Scikit-Learn algorithms and GPU computing for neural network models. The model inputs and outputs are automatically stored in a Neo4j graph database.” Luxon’s Chapter 2: Methodology provides Figure 6 which is a data process flow diagram that utilizes the machine learning pipeline for model generation (which is interpreted by the examiner as the method). The pipeline supports computing across multiple CPU threads + GPU computation for NN models which implies the machine learning pipeline/process is computing on a computing device; thus, a computer-implemented method).
accessing, by a computing device, a graph (Luxon, Page 50, Figure 17; Page 53, Paragraph 1, “The interactive user interface enables intuitive exploration and manipulation of the graph”; Page 26, Paragraph 3, “The Neo4j software platform … was used for creating the property graph database. Graph queries and visualization were carried out using Neo4j Browser and Neo4j Bloom ....”. Figure 17 shows an example of the UI for a user accessing a graph (by querying via the Neo4j Browser) within the interactive user interface (Neo4j Bloom) for a case study) comprising one or more model nodes, one or more dataset nodes, and one or more edges (Luxon, Page 19, Figure 5b, Caption, “B: The proposed approach connects model inputs and outputs as a descriptive graph. Database queries access the full model information and context for richer model analyses”; Page 27, Figure 7. Figure 5b depicts a simplified connected graph example showing a graph comprising one or model nodes (ex: Model 1 node, Model 2 node, Model 3 node), dataset nodes (ex: Data node which will be split into multiple nodes and shown in Figure 7), and one or more edges (ex: USE_DATA, HAS_SCORE, USES_SPLIT, etc.). Figure 7 shows the Dataset node being split within the DB schema where nodes represent entities in the machine learning workflow/pipeline/method), wherein the one or more model nodes having at least one of a plurality of features … for a machine learning model that is represented by a model node of the one or more model nodes, … (Luxon, Page 27, Figure 7; Page 28, Figure 8, Caption, “… In the property graph model, properties can be stored on graph nodes and relationships to further describe and quantify the structures”. Figure 8 depicts the storage of model nodes (Figure 8: MLModel (A)) and edges both of which contain properties (which is interpreted as features by the examiner as they describe the model/edge such as seed, feat_time, etc. Figure 7 also depicts the DB schema where the ML Model node is the backbone of the architecture with edges to Feature Method -> Feature, and Feature List to depict the ML Model process of using feature methods to calculate features based off the Dataset containing a Molecule to create a Feature List (plurality of features));
adding, by the computing device (Luxon, Page 30, Paragraph 2, “Adding a single machine learning model to the graph will typically create three or four new unique nodes and thousands of new relationships”. adding ML Models to the graph adds test nodes via the pipeline), one or more test dataset nodes and one or more test edges to the graph (Luxon, Page 27, Figure 7; Page 26, Paragraph 4, “When a new model was run and imported to the graph, the necessary nodes were created if they did already exist. A new model always created a unique MLModel, training set, test set, and if applicable, validation set nodes”. Figure 7 depicts the one or more test dataset nodes and one or more test edges to the graph);
performing, by the computing device, a series of iterative steps until a threshold is reached (Luxon, Page 1, Paragraph 2, “Prior to the design cycle, system requirements and criteria must be established to define the desired performance of the system and provide a target for the optimization process … system is then built to the design specifications … performance … is evaluated against the system requirements. Performance information is used to inform the next iteration ... The design-test-build sequence is iterated upon until the system performance requirements are satisfied”. The design-test-build sequence is a series of iterative steps which are performed and iterated upon until the predetermined requirements are satisfied (which is interpreted by the examiner as reaching a threshold));
for each iterative step: determining, by the computing device, … for each model node,
(Luxon, Page 34, Figure 10; Page 35, Figure 11; Page 31, Paragraph 2, “One can quickly identify top performing models by thick green arrows …”; Page 35, Paragraph 1, “The visualization can be used to identify troublesome molecules, both by prediction error, shown by the arrow color, and by prediction uncertainty, shown by arrow thickness”; Page 24, Paragraph 4, “Cross validation (CV) hyperparameter tuning was performed using exhaustive grid and random search from the Scikit-Learn package and Bayesian optimization from the skopt package”. Figure 10 shows a sub-graph and represents the architecture that users would use for evaluation. The graph architecture highlights predictive performance via color and thickness of edges where the predictive performance is the probability for a ML Model node being able to predict/select a particular dataset node correctly (molecule)). The predictive performance is calculated for the full graph architecture (for each model node) as the user/researcher is able to view/evaluate and manipulate the graph and the optimization method for each model node is done through Bayesian Optimization. Figure 11 shows an example of comparing two trained models nodes (dark purple) from the graph database that use different ML algorithms (B and C nodes) and visually representing the overall predictive performance (Green D and Orange E edges) for the shared test dataset (light blue node));
…
selecting, by the computing device, a particular model node, the particular model node being selected based at least in part on the … , wherein the model node comprises hyperparameters, …
(Luxon, Page 13, Figure 3; Page 25, Figure 6: TUNE?; Page 35, Figure 11; Page 27, Figure 7; Page 32, Table 3 & Figure 9. During training the goal is to adjust the learnable model weights until a completion criteria (when the predictive performance has been optimized based on the selection criteria) via hyperparameter tuning which is shown in Figure 3. Figure 3 depicts the process of model tuning via 3-fold cross validation where the model is based on the cross validation iterations. Figure 6 shows the Data process flow diagram once new data is imported which contains a step of TUNE? where the decision step selects models that need to be tuned (thus, interpreted as a particular model node). Figure 7 depicts the database schema where the ML model node has a relationship with Tuning Method via USES_TUNING; Table 3 and Figure 9 show that each node has labels which contains properties including tuning (which is for hyperparameter optimization as noted within Figure 6) and thus, the examiner interprets the model node to comprise hyperparameters as the properties (including hyperparameters) are stored on the node (noted within the caption of Table 3) and tuned);
updating, by the computer device, the selection criteria based at least in part on the particular model node; and
(Luxon, Page 13, Figure 3; Page 25, Figure 6: HYPERPARAMETER OPTIMIZATION; Page 12, Paragraph 4, “There are also non-learnable algorithm parameters that can be adjusted by the modeler to increase model performance. These non-learnable parameters are called hyperparameters and the optimization of them is important to maximizing performance …The objective is to find which point in the parameter grid provides the best model performance”. Figure 6 shows TUNE? -> HYPERPARAMETER OPTIMIZATION -> MODEL TRAINER; where the selected model (the particular model node) from TUNE? is optimized for performance via hyperparameter optimization (shown in Figure 3). The selection criteria is updated for the particular model node as the objective is to update parameters to optimize the selection of a molecule; where the optimization is based on finding the point in the parameter grid that provides the best model performance);
updating, by the computer device, at least one of the plurality of features or the plurality of weights based at least in part on the particular model node; and
(Luxon, Page 12, Paragraph 3, “… the algorithm adjusts learnable model weights until a completion criteria, which varies by algorithm, has been met”. The training is updating the plurality of weights for learnable and non-learnable parameters (by hyperparameter optimization) which adjusts/updates the weights to optimize the particular model node);
training, by the computer device, a selected model from a last iterative step on at least one dataset that is represented by the one or more test dataset nodes.
(Luxon, Page 13, Figure 3; Page 25, Figure 6: Model Storage - EXPORT; Page 35, Figure 11; Page 53, Paragraph 1, “The interactive user interface enables intuitive exploration and manipulation of the graph”. Figure 6 shows the last iterative step where EXPORT (which is interpreted as providing by the examiner) exports the Graph Database and an example can be seen within the UI via Figure 11 (where the user is able to manipulate and explore the graph); thus, providing the selected particular model node for presentation which is trained by at least one dataset that is represented by the one or more test dataset nodes as the datasets are split into training and test data; thus, the training data represents the same type of data as the test data (just a different split));
Luxon teaches a knowledge graph utilizing the Neo4j’s graph database system for calculating predictive performance … but does not explicitly disclose determining a selection probability or an explicit edge weight connecting two dataset nodes. These recited limitations are not taught explicitly by Luxon:
… and wherein an edge weight of an edge that connects two dataset nodes represents a similarity of datasets represented by the two dataset nodes;
… a selection probability …
… wherein the determining the selection probability … comprises: determining a selection weight, and
determining a plurality of selection criteria;
… wherein the hyperparameters are configured to control a structure of a model …
However, Needham explicitly teaches:
… a selection probability …
(Needham, Page 34, Chapter 7: Centrality Algorithms & Table – “Centrality algorithms are used to find the most influential nodes in a graph … ”; Page 46, Paragraph 2, “Two common strategies for selecting … nodes are: Random - Nodes are selected uniformly, at random, with defined probability of selection … Degree …”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to utilize the machine learning pipeline of Luxon which provides an interactive user interface for graph evaluation via Neo4j to find the best model (in terms of predictive performance), with the use of the Needham’s centrality algorithms for selection probability, and similarity weight graph generation for Label Propagation to assign similarity weights between dataset nodes. One having ordinary skill in the art would have been motivated to implement this change before the effective filing date of the claimed invention, as this leads to more metrics to provide the user, detecting influence + importance, performance optimization, utilization of the Neo4j graph database management algorithms, computationally inexpensive, and has been used in the past for reviewing chemical similarities (see Needham, Page 34, Chapter 7: Centrality Algorithms & Table – “Centrality algorithms are used to find the most influential nodes in a graph … ”; Page 42, Paragraph 1, “Betweenness Centrality is a way of detecting the amount of influence a node has over the flow of information in a graph. It is often used to find nodes that serve as a bridge from one part of a graph to another … Each node receives a score, based on the number of these shortest paths that pass through the node. Nodes that most frequently lie on these shortest paths will have a higher betweenness centrality score.”; Page 59, Paragraph 1, “The Label Propagation algorithm (LPA) is a fast algorithm for finding communities in a graph … LPA is a relatively new algorithm and was only proposed by Raghavan et al. in 2007, in "Near linear time algorithm to detect community structures in large-scale networks"”; Page 60, Paragraph 2, “Label Propagation has been used to estimate potentially dangerous combinations of drugs to co-prescribe to a patient, based on the chemical similarity and side effect profiles. The study is found in "Label Propagation Prediction of Drug-Drug Interactions Based on Clinical Side Effects"”).
Needham/Luxon do not explicitly disclose:
… and wherein an edge weight of an edge that connects two dataset nodes represents a similarity of datasets represented by the two dataset nodes;
… wherein the determining the selection probability … comprises: determining a selection weight, and
determining a plurality of selection criteria;
… wherein the hyperparameters are configured to control a structure of a model …
However, Färber teaches:
… and wherein an edge weight of an edge that connects two dataset nodes represents a similarity of datasets represented by the two dataset nodes;
(Färber, Page 1329, Paragraph 3, “The edges between the data sets of the knowledge graph represent the similarity of the data sets. The similarities between the data sets are constructed using their metadata and the SOM algorithm …”; Page 1342, Paragraph 2, “… we have developed a rule-based approach for author name disambiguation that is adapted to the metadata of data sets. To calculate the similarity between two author names and, thus, to know the candidates for author name disambiguation, we use the Jaro-Winkler similarity …” . Färber utilizes the teaching of Ojo and Sennaike (2020) where there are edge weights between datasets (whether that be between two distinct datasets or split datasets based on category) are represented based on a similarity score (weight of relevance)).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to utilize the machine learning pipeline of Luxon/Needham with the use of the Färber’s explicit teachings of edge weights between datasets to assign similarity weights between dataset nodes to be able to distinguish different datasets and relevancy. One having ordinary skill in the art would have been motivated to implement this change before the effective filing date of the claimed invention, as this leads to resolving ambiguities, categorical evaluations, improvement in recommendation systems/evaluations/analysis (see Färber, Page 1352, Paragraph 3, “To resolve the ambiguities of the author names, we adapted an author name disambiguation approach to data set metadata. In addition to explicit evidence, implicit evidence was taken into account in the form of latent topic modeling … We outlined potential use cases of the created knowledge graph and showed that the DSKG can be used in particular in the context of search and recommender systems, as well as for scientific impact quantification …”).
Needham/Luxon/Färber do not explicitly disclose:
… wherein the determining the selection probability … comprises: determining a selection weight, and
determining a plurality of selection criteria;
… wherein the hyperparameters are configured to control a structure of a model …
However, Wang teaches:
… wherein the determining the selection probability … comprises: determining a selection weight, and
(Wang, Page 951, Column 1, Paragraph 2, “… 1) the nodes that have high-order relations … which imposes computational overload to the model, and 2) the high-order relations contribute unequally to a prediction, which requires the model to carefully weight (or select) them”; Page 952, Figure 2, Column 2, Bullet Point #2, “Output: a prediction function that predicts the probability … that user u would adopt item i”; Page 953, Equation (4):
PNG
media_image1.png
23
272
media_image1.png
Greyscale
. Figure 2 shows the KGAT (Knowledge Graph Attention Network) model which utilizes an attention mechanism (interpreted as a selection mechanism) to learn the weights of neighboring nodes to determine the most likely for a user to adopt; thus, determining selection weights (via Equation 4) for a selection probability (via Equation 5)).
determining a plurality of selection criteria;
(Wang, Page 953, Equation (5):
PNG
media_image2.png
57
376
media_image2.png
Greyscale
, Column 2, Paragraph 1, “As a result, the final attention score is capable of suggesting which neighbor nodes should be given more attention to capture collaborative signals. When performing propagation forward, the attention flow suggests parts of the data to focus on, which can be treated as explanations behind the recommendation”. Equation 5 shows the final attention score for a selection probability which comprises the selection criteria of head embedding, relation embedding and tail embedding; thus, a plurality of selection criteria).
… wherein the hyperparameters are configured to control a structure of a model …
(Wang, Page 954, Column 2, Paragraph 2, “Training: … adaptively control the learning rate w.r.t. the absolute value of gradient. In particular, for a batch of randomly sampled (h, r , t , t ′), … and then update model parameters by using the gradients of the prediction loss”. The training leads to adaptively controlling the learning rates of a model to update model parameters which controls the structure of the model; thus, the hyperparameters control/update/tune the model parameters and interpreted by the examiner to control a structure of a model with respects to gradients of the prediction loss).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to utilize the machine learning pipeline of Luxon/Needham/Färber with the use of the Wang’s explicit teachings of determining selection probabilities, weight, criteria, and utilizing the hyperparameters to control a structure of a model. One having ordinary skill in the art would have been motivated to implement this change before the effective filing date of the claimed invention, as this leads to analyzing influence + explainability, updating of node representations, recommendation performance/accuracy, and exploiting structural knowledge (see Wang, Page 958, Column 1, Paragraph 1, “In this work, we explore high-order connectivity with semantic relations in CKG for knowledge-aware recommendation. We devised a new framework KGAT, which explicitly models the high order connectivities in CKG in an end-to-end fashion. At it core is the attentive embedding propagation layer, which adaptively propagates the embeddings from a node’s neighbors to update the node’s representation. Extensive experiments on three real-world datasets demonstrate the rationality and effectiveness of KGAT. This work explores the potential of graph neural networks in recommendation, and represents an initial attempt to exploit structural knowledge with information propagation mechanism … with CKG, we can investigate how social influence affects the recommendation. Another exciting direction is the integration of information propagation and decision process, which opens up research possibilities of explainable recommendation”).
Regarding Claim 2:
Luxon/Needham/Färber/Wang teach Claim 1. Luxon further teaches:
wherein the plurality of features comprise a plurality of hyperparameters.
(Luxon, Page 13, Figure 3; Page 19, Figure 5B; Page 12, Paragraph 4, “There are also non-learnable algorithm parameters … called hyperparameters and the optimization of them is important to maximizing performance … Hyperparameter tuning begins with specifying which algorithm parameters to adjust and over what range …”. There are a plurality of hyperparameters that are optimized through the hyperparameter optimization for the model features where the combination of the algorithm feature parameters are updated and evaluated to determine the optimal parameters).
Regarding Claim 3:
Luxon/Needham/Färber/Wang teach Claim 1. Luxon further teaches:
wherein the threshold can be at least one of a time period …
(Luxon, Page 25, Figure 6; Page 13, Paragraph 1, “… for a given set of algorithm parameters, the model will be trained thrice – each iteration using a two folds for training and the third fold for predictions (validation). This process is repeated until all folds have been used as validation folds and the performance of the algorithm is evaluated on the composite performance of all three iterations”. The design-test-build sequence is iterated until the predetermined requirements are satisfied (which is interpreted by the examiner as reaching a threshold) where the model tuning/training is trained three times which is interpreted as a time period as an iteration is a period of time).
Regarding Claim 4:
Luxon/Needham/Färber/Wang teach Claim 1. Luxon further teaches:
wherein a model node, from the one or more model nodes, and a first dataset node, from the one or more dataset nodes, are connected by a model edge, of the one or more edges, in accordance with the dataset having been evaluated by a model that is represented by the model node
(Luxon, Page 19, Figure 5B; Page 25, Figure 6: Database Matching - DATABASE/FEATURIZER; Page 35, Figure 11. Figure 5B shows a simplified connected graph depicting one or more model nodes with a USE_DATE edge between the model nodes and a dataset node. Figure 11 shows a comparison of two particular model nodes where both models share a dataset where a model edge is noted as USES_DATASET and a PREDICTS edge as the dataset was evaluated by the model using the light blue dataset node).
Regarding Claim 5:
Luxon/Needham/Färber/Wang teach Claim 4. Luxon further teaches:
wherein the model edge has a selection weight, the selection criteria being based at least in part on the selection weight.
(Luxon, Figures 10-12; Page 31, Paragraph 2, “This sub-graph provides a visual, semi-quantitative comparison of models through relationship color and weight. One can quickly identify top performing models by thick green arrows (A)”; Page 35, Paragraph 1, “The visualization can be used to identify troublesome molecules, both by prediction error, shown by the arrow color, and by prediction uncertainty, shown by arrow thickness. Figures 10-12 shows different thickness arrows in color to provide a visualization. Figure 11 shows the model node (purple nodes) edges (PREDICTS) which gives a pictorial representation of predictive performance where the PREDICTS edge has a selection weight (interpreted as the strength of the predictive performance) and a selection criteria based on the selection weight (interpreted as minimizing the prediction error for selecting the correct molecule)).
Regarding Claim 7:
Luxon/Needham/Färber/Wang teach Claim 4. Luxon further teaches:
wherein adding the one or more test edges further comprises determining a similarity weight for the one or more test edges.
(Luxon, Page 35, Figure 11. Figure 11 shows the two models evaluations (predictions) for the test datasets which were trained on the shared dataset; where the test dataset edge arrows are indicating prediction error and uncertainty via color and thickness, respectively. Incorporating Needham would then be able to provide a similarity weight between test dataset nodes via Neo4j’s algorithms which are already used by Luxon).
Regarding Claims 8-12:
Claims 8-12 incorporate substantively all the limitations of Claims 1-5 in a non-transitory computer-readable storage medium (thus, a manufacture) and further recites storing a set of instructions, that, when executed by one or more processors of a recommendation system computing device, cause the one or more processors to perform instructions comprising (Luxon, Page 25, Figure 6; Page 30, Paragraph 1, “The developed machine learning pipeline automates the process of parameterizing, training, optimizing, and evaluating … The pipeline supports parallel computing across multiple CPU threads for Scikit-Learn algorithms and GPU computing for neural network models. The model inputs and outputs are automatically stored in a Neo4j graph database”. The machine learning pipeline learns, trains, optimizes and evaluates on a computing device in which the CPU (processor), memory, and CRM are inherent); thus, Claims 8-12 are rejected for reasons set forth in the rejections of Claims 1-5, respectively.
Regarding Claim 14:
Luxon/Needham/Färber/Wang teach Claim 9. Luxon further teaches:
wherein adding the one or more test edges further comprises determining a similarity weight for the one or more test edges.
(Luxon, Page 35, Figure 11. Figure 11 shows the two models evaluations (predictions) for the test datasets which were trained on the shared dataset; where the test dataset edge arrows are indicating prediction error and uncertainty via color and thickness, respectively. Incorporating Needham would then be able to provide a similarity weight between test dataset nodes via Neo4j’s algorithms which are already used by Luxon).
Regarding Claims 15-18 and 20:
Claims 15-18 and 20 incorporate substantively all the limitations of Claims 1-2, 4-5 and 7 in a system (thus, a machine) and further recites comprising: memory storing computer-executable instructions; and one or more processors configured to access the memory, and execute the computer-executable instructions to at least (Luxon, Page 25, Figure 6; Page 30, Paragraph 1, “The developed machine learning pipeline automates the process of parameterizing, training, optimizing, and evaluating … The pipeline supports parallel computing across multiple CPU threads for Scikit-Learn algorithms and GPU computing for neural network models. The model inputs and outputs are automatically stored in a Neo4j graph database”. The machine learning pipeline learns, trains, optimizes and evaluates on a computing device in which the CPU (processor), memory, and CRM are inherent); thus, Claims 15-18 and 20 are rejected for reasons set forth in the rejections of Claims 1-2, 4-5 and 7, respectively.
Regarding Claim 21:
Luxon/Needham/Färber/Wang teach Claim 2. Wang further teaches:
wherein the plurality of hyperparameters comprise parameters that tum a learning algorithm into a trained model, and wherein the plurality of hyperparameters are set before training the learning algorithm starts.
(Wang, Page 955, Column 2, Paragraph 2, “Parameter Settings. We implement our KGAT model in Tensorflow. The embedding size is fixed to 64 for all models, except RippleNet 16 due to its high computational cost. We optimize all models with Adam optimizer, where the batch size is fixed at 1024. The default Xavier initializer [8] to initialize the model parameters. We apply a grid search for hyper-parameters: the learning rate is tuned amongst …”. Section 4.2.3 within Wang describes the initialization of the hyperparameters before training and initializing the hyperparameters to then be tuned via training where the learning algorithm of KGAT is implemented until the model is trained for the different scenarios).
The motivation of Claim 1’s combination of Luxon/Needham/Färber/Wang is still maintained.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to IBRAHIM RAHMAN whose telephone number is (703)756-1646. The examiner can normally be reached M-F 8am-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, Kakali Chaki can be reached at (571) 272-3719. 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.
/I.R./Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights