DETAILED ACTION
1. This office action is in response to the Application No. 18579451 filed on 01/15/2024. Claims 1-16 are presented for examination and are currently pending.
Notice of Pre-AIA or AIA Status
2. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
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.
3. Claims 1-16 are rejected under 35 U.S.C 101 because the claimed invention is directed towards an abstract idea without significantly more.
Step 1
Independent claim 1 is directed to a method, and falls into one of the four statutory categories.
Step 2A, Prong 1
Claim 1 recites the following abstract ideas:
determining an output vector by inputting the plurality of latent vectors into a second neural network comprising a graph neural network, (Mental process directed to determining an output vector by inputting the plurality of latent vectors a graph neural network, which can be performed with a pen and paper)
minimising a loss function by tuning the edge probabilities of the graph, at least one parameter of the first neural network and at least one parameter of the second neural network, wherein the loss function comprises a function of the graph and a measure of difference between the input vector and the output vector (Mental process directed to minimizing a loss function by tuning the edge probabilities using a function of the graph and a measure of difference between the input vector and the output vector which can be done with a pen and paper).
Step 2A, Prong 2
Claim 1 recites the following additional elements:
receiving an input vector comprising values of variables (This limitation is directed to insignificant extra solution activity of data transmission. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(g));
using a first neural network to encode the values of the variables of the input vector into a plurality of latent vectors (This limitation is directed to mere instruction to apply an abstract idea. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f));
wherein the graph neural network is parametrized by a graph comprising edge probabilities indicating causal relationships between the variables (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(h)); and
Step 2B
receiving an input vector comprising values of variables (This limitation is directed to insignificant extra solution activity of data transmission and it is well understood routine and conventional. This does not amount to significantly more than judicial exception. See MPEP 2106.05(d)(II), example i);
using a first neural network to encode the values of the variables of the input vector into a plurality of latent vectors (This limitation is directed to mere instruction to apply an abstract idea. This limitation does not amount to significantly more than the judicial exception. See MPEP 2106.05(f));
wherein the graph neural network is parametrized by a graph comprising edge probabilities indicating causal relationships between the variables (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not amount to significantly more than the judicial exception. See MPEP 2106.05(h)); and
4. Dependent claim 2 is directed to a method, and falls into one of the four statutory categories.
Claim 2 do no recite any abstract ideas.
Claim 2 recites the following additional elements:
wherein the method of claim 1 is repeated for a plurality of further input vectors to provide further tuning of the edge probabilities of the graph, the at least one parameter of the first neural network and the at least one parameter of the second neural network (This limitation is directed to mere instructions to carry out the judicial exception. This limitation does not amount to significantly more than the judicial exception. See MPEP 2106.05(f))).
Claim 2 recites the following additional elements:
wherein the method of claim 1 is repeated for a plurality of further input vectors to provide further tuning of the edge probabilities of the graph, the at least one parameter of the first neural network and the at least one parameter of the second neural network (This limitation is directed to mere instructions to carry out the judicial exception. This limitation does not amount to significantly more than the judicial exception. See MPEP 2106.05(f)).
5. Dependent claim 3 is directed to a method, and falls into one of the four statutory categories.
Claim 3 recite the following abstract ideas:
wherein the method comprises: after minimising the loss function by tuning the edge probabilities of the graph, the at least one parameter of the first neural network and the at least one parameter of the second neural network: setting the edge probabilities of the graph, the at least one parameter of the first neural network and the at least one parameter of the second neural network (Mental process directed to setting the edge probabilities of the graph which can be done with pen and paper);
Claim 3 recites the following additional elements:
receiving a further input vector comprising the variables of the input vector, the further input vector having at least one missing value for at least one of the variables and at least one observed value for at least one of the variables (This limitation is directed to insignificant extra solution activity of data gathering. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(g)); and
applying the first neural network and the second neural network to the further input vector to obtain the at least one missing value (This limitation is directed to mere instructions of the judicial exception. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)).
Claim 3 recites the following additional elements:
receiving a further input vector comprising the variables of the input vector, the further input vector having at least one missing value for at least one of the variables and at least one observed value for at least one of the variables (This limitation is directed to insignificant extra solution activity of data gathering and it is well understood routine and conventional. This does not amount to significantly more than judicial exception. See MPEP 2106.05(d)(II), example i); and
applying the first neural network and the second neural network to the further input vector to obtain the at least one missing value (This limitation is directed to mere instructions of the judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f)).
6. Dependent claim 4 is directed to a method, and falls into one of the four statutory categories.
Claim 4 recite the following abstract ideas:
wherein the function of the graph increases the value of the loss function when a cyclic relationship is present in the graph (Mathematical concepts directed to a function of the graph that increases the value of the loss function when a cyclic relationship is present in the graph).
Claim 4 do not recites any additional elements.
7. Dependent claim 5 is directed to a method, and falls into one of the four statutory categories.
Claim 5 recite the following abstract ideas:
wherein the function of the graph comprises a measure of a difference between two distributions, wherein the first distribution is an estimate of a posterior function of the graph and the second distribution is a predefined user function of the graph (Mathematical concepts directed to the function of the graph that measure of a difference between an estimate of a posterior function of the graph and the second distribution is a predefined user function of the graph).
Claim 5 do not recites any additional elements.
8. Dependent claim 6 is directed to a method, and falls into one of the four statutory categories.
Claim 6 recite the following abstract ideas:
wherein the loss function only operates on variables that are present in the input vector (Mental process directed to operating the loss function only when an input vector is available which can be done by observing the available input vector and making a judgement on when the function will be operated).
Claim 6 do not recites any additional elements.
9. Dependent claim 7 is directed to a method, and falls into one of the four statutory categories.
Claim 7 do not recite any abstract ideas.
Claim 7 recite the following additional elements:
wherein the using the first neural network to encode the values of the variables of the input vector into the plurality of latent vectors comprises: using the first neural network to encode each variable of the input vector into a respective latent vector (This limitation is directed to mere instructions to apply a judicial exception. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)).
Claim 7 recite the following additional elements:
wherein the using the first neural network to encode the values of the variables of the input vector into the plurality of latent vectors comprises: using the first neural network to encode each variable of the input vector into a respective latent vector (This limitation is directed to mere instructions to apply a judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f)).
10. Dependent claim 8 is directed to a method, and falls into one of the four statutory categories.
Claim 8 recite the following abstract ideas:
organizing the values of the variables into a number of groups, wherein the number of groups is less than a number of variables in the input vector (Mental process directed to organizing values of the variables into groups by observation and making a judgement that the number of groups is less than the number of variables in the input vector), and
Claim 8 recite the following additional elements:
wherein the using the first neural network to encode the values of the variables of the input vector into the plurality of latent vectors comprises: using the first neural network to encode each group into a respective latent vector (This limitation is directed to mere instructions to apply a judicial exception. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)).
Claim 8 recite the following additional elements:
wherein the using the first neural network to encode the values of the variables of the input vector into the plurality of latent vectors comprises: using the first neural network to encode each group into a respective latent vector (This limitation is directed to mere instructions to apply a judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f)).
11. Dependent claim 9 is directed to a method, and falls into one of the four statutory categories.
Claim 9 do not recite any abstract ideas.
Claim 9 recite the following additional elements:
wherein each group comprises at least one related variable (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(h)).
Claim 9 recite the following additional elements:
wherein each group comprises at least one related variable (This limitation is directed to a particular type or source of data, which is field of use. This does not amount to significantly more than judicial exception. See MPEP 2106.05(h)).
12. Dependent claim 10 is directed to a method, and falls into one of the four statutory categories.
Claim 10 do not recite any abstract ideas.
Claim 10 recite the following additional elements:
wherein the variables comprise at least one data value representing at least one sensor value of at least one device (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(h)).
Claim 10 recite the following additional elements:
wherein the variables comprise at least one data value representing at least one sensor value of at least one device (This limitation is directed to a particular type or source of data, which is field of use. This does not amount to significantly more than judicial exception. See MPEP 2106.05(h)).
13. Dependent claim 11 is directed to a method, and falls into one of the four statutory categories.
Claim 11 recite the following abstract ideas:
wherein the method comprises: using the causal relationships to diagnose a patient (Mental process directed to using the causal relationships to diagnose a patient, which can be done by observing the causal relationships and making a judgement on the diagnosis).
wherein the tuning the edge probabilities of the graph function provides causal relationships between a plurality of health conditions (Mental process directed to the edge probabilities being the causal relationship between a plurality of health conditions which is done by observing the causal relationship and making a judgement on what health condition is probable),
Claim 11 recite the following additional elements:
wherein: the at least one device comprises a health monitoring device for monitoring a patient (This limitation is directed to a computer component. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)) and
Claim 11 recite the following additional elements:
wherein: the at least one device comprises a health monitoring device for monitoring a patient (This limitation is directed to a computer component. This does not amount to significantly more than judicial exception. See MPEP 2106.05(h). See MPEP 2106.05(f)) and
14. Dependent claim 12 is directed to a method, and falls into one of the four statutory categories.
Claim 12 recites the following abstract ideas:
wherein the method comprises, after tuning the edge probabilities of the graph, the at least one parameter of the first neural network and the at least one parameter of the second neural network: setting the edge probabilities of the graph, the at least one parameter of the first neural network and the at least one parameter of the second neural network (Mental process directed to setting the edge probabilities of the graph which can be done with pen and paper);
Claim 12 recite the following additional elements:
wherein the at least one device comprises a health monitoring device for monitoring a patient (This limitation is directed to a computer component. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)) and
receiving a further input vector comprising the variables of the input vector, the further input vector having at least one missing value for at least one of the variables and at least one observed value for at least one of the variables (This limitation is directed to insignificant extra solution activity of data gathering. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(g)); and
applying the first neural network and the second neural network to the further input vector (This limitation is directed to mere instructions of the judicial exception. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f))
to obtain the at least one missing value, the missing value representing a health condition (This limitation is directed to insignificant extra solution activity of data gathering. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(g)).
Claim 12 recite the following additional elements:
wherein the at least one device comprises a health monitoring device for monitoring a patient (This limitation is directed to a computer component. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f)) and
receiving a further input vector comprising the variables of the input vector, the further input vector having at least one missing value for at least one of the variables and at least one observed value for at least one of the variables (This limitation is directed to insignificant extra solution activity of data gathering and it is well understood routine and conventional . This does not amount to significantly more than judicial exception. See MPEP 2106.05(d)(II), example i); and
applying the first neural network and the second neural network to the further input vector (This limitation is directed to mere instructions of the judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f))
to obtain the at least one missing value, the missing value representing a health condition (This limitation is directed to insignificant extra solution activity of data gathering and it is well understood routine and conventional. This does not amount to significantly more than judicial exception. See MPEP 2106.05(d)(II), example i).
15. Dependent claim 13 is directed to a method, and falls into one of the four statutory categories.
Claim 13 recites the following abstract ideas:
wherein the tuning the edge probabilities of the graph provides causal relationships between a plurality of sensor measurements (Mental process directed to tuning the edge probabilities being the causal relationship between a plurality of sensor measurements which is done by observing the causal relationship between the sensor measurements); and
wherein the method comprises: using the causal relationships to determine at least one fault in the at least one device (Mental process directed to using the causal relationships to determine a fault one device which can be done by observing and making a judgement on the device that is faulty).
Claim 13 do not recite any additional elements.
16. Dependent claim 14 is directed to a method, and falls into one of the four statutory categories.
Claim 14 recite the following abstract ideas:
wherein the method comprises, after minimising the loss function by tuning the edge probabilities of the graph, the at least one parameter of the first neural network and the at least one parameter of the second neural network:setting the edge probabilities of the graph, the at least one parameter of the first neural network and the at least one parameter of the second neural network (Mental process directed to setting the edge probabilities of the graph which can be done with pen and paper);
using the missing value to determine a fault in the at least one device (Mental process directed to using the missing value to determine a fault one device which can be done by observing the values that are missing and making a judgement on the device that is faulty).
Claim 14 recites the following additional elements:
receiving a further input vector comprising the variables of the input vector, the further input vector having at least one missing value for at least one of the variables and at least one observed value for at least one of the variables (This limitation is directed to insignificant extra solution activity of data gathering. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(g));
applying the first neural network and the second neural network to the further input vector to obtain the at least one missing value(This limitation is directed to mere instructions of the judicial exception. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)),
the missing value representing a state of the device (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(h)); and
Claim 14 recites the following additional elements:
receiving a further input vector comprising the variables of the input vector, the further input vector having at least one missing value for at least one of the variables and at least one observed value for at least one of the variables (This limitation is directed to insignificant extra solution activity of data gathering and it is well understood routine and conventional. This does not amount to significantly more than judicial exception. See MPEP 2106.05(d)(II), example i);
applying the first neural network and the second neural network to the further input vector to obtain the at least one missing value(This limitation is directed to mere instructions of the judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f)),
the missing value representing a state of the device (This limitation is directed to a particular type or source of data, which is field of use. This does not amount to significantly more than judicial exception. See MPEP 2106.05(h)); and
With regards to claim 16, it is substantially similar to claim 3, and is rejected in the
same manner and reasoning applying.
17. Independent claim 15 is directed to a system, and falls into one of the four statutory categories.
With regards to claim 15, it is substantially similar to claim 1, and is rejected in the same manner and reasoning applying.
Claim 15 further recites “a computer program embodied on computer-readable storage, the program comprising code configured so as when run on at least one processor to perform the operations of;” this limitation is directed to a computer component. This does not integrate the abstract idea into a practical application nor does it amount to significantly more than judicial exception. See MPEP 2106.05(f).
18. Independent claim 16 is directed to a system, and falls into one of the four statutory categories.
With regards to claim 16, it is substantially similar to claim 1, and is rejected in the same manner and reasoning applying.
Claim 16 further recites “computer system comprising: storage comprising at least one memory unit and a processing apparatus comprising at least one processing unit; wherein the storage stores code arranged to run on the processing apparatus, the code being configured so as when thus run to perform the operations of:” this limitation is directed to a computer component. This does not integrate the abstract idea into a practical application nor does it amount to significantly more than judicial exception. See MPEP 2106.05(f).
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
19. Claims 1-9, 15 and 16 are rejected under 35 U.S.C 102(a)(1) as being anticipated by Levy et al. ("Estimating latent positions in social and biological networks using Graph Neural Networks in R with GCN4R." bioRxiv preprint doi: https://doi.org/10.1101/2020.11.02.364935; this version posted November 5, 2020).
Regarding claim 1, Levy teaches a computer-implemented method (The operations performed on the neighbors of node 𝑖 to update their information is done by scattering option SCATTER, which parallelize executions across the CPU/GPU, pg. 7, first para.)
comprising: receiving an input vector comprising values of variables (Figure 1: Illustration of GCN4R framework; graph, with adjacency matrix A and attribute matrix X are input into model, Fig. 1, pg. 8. The Examiner notes a matrix comprises of values);
using a first neural network to encode the values of the variables of the input vector (where an encoder maps the node-level covariates of the input graph to low-dimensional representations (pg. 7, last para.); covariate information X is updated using successive applications of GCNConv operators (encoder) to yield lower dimensional representations of the nodes, Fig. 1, pg. 8) into a plurality of latent vectors (latent vectors Z, Fig. 1, pg. 8);
determining an output vector (The decoder outputs a probability score for the presence of a possible edge, pg. 16, last para.) by inputting the plurality of latent vectors into a second neural network comprising a graph neural network (encoder-decoder architecture of the GNN (pg. 11, fourth para.);The distribution of the latent vectors is the posterior distribution 𝑃(𝑍|𝑋), where the decoder represents the likelihood function or data generating process given the latent data 𝑃(𝑋|𝑍), pg. 8, first para., Fig. 1),
wherein the graph neural network is parametrized by a graph comprising edge probabilities indicating causal relationships between the variables (Note that these probabilities may vary across all edges of the graph (pg. 8, first para.); We perturbed edges of the input graph with networks to assess and remove statistically insignificant edges (pg. 21, last para. to pg. 22, first para.); The instant specification, pg. 21, line 16, discloses: “The generator (also known as decoder takes ... G (Graph) as input)”); and
minimising a loss function by tuning the edge probabilities of the graph (The Kullback Leibler Divergence, 𝐿 , is minimized between the distribution of latent variables and the “guide” multivariate normal distribution to approximate the true posterior distribution in the absence of a conjugate-prior, pg. 8, first para.; The instant specification, pg. 15, lines 14-2, discloses: “In some examples the loss function may also comprise a further function for regularizing G 921. The function may comprise a measure of difference between an estimated value of a posterior function of G 921, q(G), and a prior value of G 921, p(G) ... The measure of difference may be determined using any suitable algorithm, for example a Kullback-Liebler (KL) divergence function”),
at least one parameter of the first neural network and at least one parameter of the second neural network (The GNN modeling approach introduced in GCN4R only covers a small subset of all possible network learning tasks using GNNs ... The latent parameters as introduced by the encoder-decoder architecture of the GNN follows a variational bayes learning framework (pg. 11, fourth para.); Given that one or more attention layers may be learned at a given layer, the matrix multiplication of these weights across layers may give an indication of which node has influenced the prediction of another within k-steps (for k GCN layers), pg. 28, second para.) ,
wherein the loss function comprises a function of the graph and a measure of difference between the input vector and the output vector (Various objective functions (L) such as mean squared error ... are utilized to update the parameters of these models via backpropagation, pg. 3, first para.; The instant specification, pg. 16, lines 6-12, discloses: “In some examples, tuning the values of one or more of G (e.g. edge probabilities of graph G 921), θGNN (parameters of decoder GNN 921), θ (parameters of decoder 938) and ɸ (parameters of encoder 934) in order to minimise the loss function L ... Differentiating L and computing the gradient of each parameter with respect to L (using backpropagation)”).
Regarding claim 2, Levy teaches a method according to claim 1, Levy teaches wherein the method of claim 1 is repeated for a plurality of further input vectors to provide further tuning of the edge probabilities of the graph (the operator is the GCNConv32. Multiple applications of the operator update the node-level predictors to form latent embedding 𝑍⃗ for the nodes in the graph (pg. 7, last para.); an edge (i, j), corresponding to nodes i and j, is an indicator function that is valued 1 when there exists a relationship between i and j and 0 otherwise (pg. 5, second to the last para.); We hope to see that all losses are decreasing over training iterations, indicating that the model is training towards convergence, pg. 16, first para.),
the at least one parameter of the first neural network and the at least one parameter of the second neural network (Attention based methods seek to derive an importance score for each node based on centrality measures that are weighted by the learned attention weights. These importance scores may be derived for each layer of the network and speak to how much information one node may receive from neighboring nodes (pg. 24, third to the last para.); Given that one or more attention layers may be learned at a given layer, the matrix multiplication of these weights across layers may give an indication of which node has influenced the prediction of another within k-steps (for k GCN layers), pg. 28, last para.).
Regarding claim 3, Levy teaches the method according to claim 1, wherein the method comprises: after minimising the loss function by tuning the edge probabilities of the graph (GCN4R parameters includes many tunable default parameters, pg. 14, second para.),
the at least one parameter of the first neural network and the at least one parameter of the second neural network: setting the edge probabilities of the graph (positive edges are randomly removed with probability 𝑝pos and negative edges are randomly flipped to positive edges with probability 𝑝neg and can be set such that the density of the network is preserved (pg. 10, last para.),
the at least one parameter of the first neural network and the at least one parameter of the second neural network (The GNN modeling approach introduced in GCN4R only covers a small subset of all possible network learning tasks using GNNs ... The latent parameters as introduced by the encoder-decoder architecture of the GNN follows a variational bayes learning framework (pg. 11, fourth para.); Given that one or more attention layers may be learned at a given layer, the matrix multiplication of these weights across layers may give an indication of which node has influenced the prediction of another within k-steps (for k GCN layers), pg. 28, second para.));
receiving a further input vector comprising the variables of the input vector (Figure 1: Illustration of GCN4R framework; graph, with ... attribute matrix X are input into model, Fig. 1, pg. 8),
the further input vector having at least one missing value for at least one of the variables and at least one observed value for at least one of the variables (When the adjacency matrix is a sparse matrix, calculations can be streamlined (pg. 5, second to the last para.); imputation of missing characteristics of observed nodes, pg. 5, first para. The instant specification (US20240338559 [0146], discloses: “To make the model adapt to different sparsity levels in the training data X, during training we drop a random percentage of the observed values”. The Examiner notes a sparse matrix indicates missing values); and
applying the first neural network and the second neural network to the further input vector to obtain the at least one missing value (Here, we see that layer 1 (Figure 5A-B) tends to have sparse attention weights between peers, while layer 2 of the network (Figure 5C-D) is highly interconnected, pg. 18, second to the last para.).
Regarding claim 4, Levy teaches the method according to claim 1, Levy teaches wherein the function of the graph increases the value of the loss function (Loss function increases from about 1.25 to about 1.45, Top graph, Fig. 11(E), pg. 26) when a cyclic relationship is present in the graph (input graph in Fig. 8 has one cyclic relationship).
Regarding claim 5, Levy teaches the method according to claim 1, Levy teaches wherein the function of the graph comprises a measure of a difference between two distributions (The GCN4R package provides additional subroutines to extract measures of model fit, pg. 16, second to the last para.),
wherein the first distribution is an estimate of a posterior function of the graph (The distribution of the latent vectors is the posterior distribution 𝑃(𝑍|𝑋), where the decoder represents the likelihood function or data generating process given the latent data 𝑃(𝑋|𝑍), pg. 8, first para.) and
the second distribution is a predefined user function of the graph (The GCN4R package operates on single graphs in the R environment. The user selects a backbone convolutional operator, which sets 𝛾 and ɸ, pg. 7, last para.).
Regarding claim 6, Levy teaches the method according to claim 1, Levy teaches wherein the loss function only operates on variables that are present in the input vector (Here the model may be trained using a loss function that is the negative log likelihood of a multinomial outcome for classification tasks, or the mean squared error (MSE) for regression tasks. When training and predicting on a single graph, or setting aside nodes for testing from other graphs, we refer to these tasks as “semi-supervised” tasks, pg. 12, second to the last para. The Examiner notes input graph in Fig. 1 is the received input).
Regarding claim 7, Levy teaches the method according to claim 1, Levy teaches wherein the using the first neural network to encode the values of the variables of the input vector into the plurality of latent vectors comprises: using the first neural network (where an encoder maps the node-level covariates of the input graph to low-dimensional representations (pg. 7, last para.); covariate information X is updated using successive applications of GCNConv operators (encoder) to yield lower dimensional representations of the nodes, Fig. 1, pg. 8) to encode each variable of the input vector into a respective latent vector (latent vectors Z, Fig. 1).
Regarding claim 8, Levy teaches the method according to claim 1, Levy teaches comprising: organizing the values of the variables into a number of groups (Given that members of formed communities should have similar connections and attributes, the attention weights visualized in Figure 5 may reflect the efforts of the model to “pull” lawyers that appear spuriously connected towards a community, pg. 18, second to the last para.),
wherein the number of groups (Three different networks were featured in this study: one defined by whether advice was received by individuals in a network (advice network; directed), the second by whether the lawyer considered another lawyer a friend (friendship network; directed), the final by whether two individuals were direct coworkers (coworker network; undirected), pg. 12, second to the last para. The Examiner notes the three groups are advice network, friendship network, and coworker network) is less than a number of variables in the input vector (The aim of the original study was to study cooperative relationships amongst 71 lawyers that had formed amongst competitive law firms, pg. 12, second to the last para.), and
wherein the using the first neural network to encode the values of the variables of the input vector into the plurality of latent vectors comprises: using the first neural network to encode each group into a respective latent vector (Figure 13: An Example of Social Influence Results from Prediction of Lawyer's Law Firm (color of nodes in networks) from Seniority, Age, and Gender: Visualization of attention weights for GCN: A) Layer 1, B) Layer 2, C) Layer 3, D) Layer 4; these attention matrices are matrix multiplied together to form: E) the 4-step social influence network; this GCN social influence network, pg. 28).
Regarding claim 9, Levy teaches the method as claimed in claim 8, Levy teaches wherein each group (advice network, friendship network, and coworker, pg. 12, second to the para.) comprises at least one related variable (Each node was assigned attributes based on their age, gender, practice (whether they were a corporate or litigation lawyer), status (partner or associate of firm), seniority in company (number of years) and law school, pg. 12, second to the last para.).
Regarding claim 15, claim 15 is similar to claim 1. It is rejected in the same manner and reasoning applying. Further, Levy teaches a computer program embodied on computer-readable storage, the program comprising code configured so as when run on at least one processor to perform the operations of: (The operations performed on the neighbors of node 𝑖 to update their information is done by scattering option SCATTER, which parallelize executions across the CPU/GPU (sending “messages” outwards and gathering the results), pg. 7, first para.)
Regarding claim 16, claim 16 is similar to claim 1. It is rejected in the same manner and reasoning applying. Further, Levy teaches a computer system comprising: storage comprising at least one memory unit and a processing apparatus comprising at least one processing unit; wherein the storage stores code arranged to run on the processing apparatus, the code being configured so as when thus run to perform the operations of (The operations performed on the neighbors of node 𝑖 to update their information is done by scattering option SCATTER, which parallelize executions across the CPU/GPU (sending “messages” outwards and gathering the results), pg. 7, first para. The Examiner notes that CPU/GPU include cores (which is a processing unit) and DRAM which is a memory unit):
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.
20. Claims 10-12 are rejected under 35 U.S.C. 103 as being unpatentable over Levy et al. ("Estimating latent positions in social and biological networks using Graph Neural Networks in R with GCN4R." bioRxiv preprint doi: https://doi.org/10.1101/2020.11.02.364935; this version posted October 27, 2021) in view of Ahmedt-Aristizabal et al. ("Graph-based deep learning for medical diagnosis and analysis: past, present and future." Sensors 21.14 (2021): 4758, published: 12 July 2021).
Regarding claim 10, Levy teaches the method according to claim 1, levy does not explicitly teach the limitations of claim 10.
Ahmedt-Aristizabal teaches wherein the variables comprise at least one data value (Brain graph of fMRI and EEG data for brain responses and emotion analysis, respectively, Fig. 1, pg. 2) representing at least one sensor value of at least one device (diagnostic imaging methods such as functional magnetic resonance imaging (fMRI), magnetic resonance imaging (MRI), computed tomography (CT), and other diagnostic tools including the electroenchephalogram (EEG) (pg. 1, introduction, first para.); These signals are captured with 306 sensors (electrodes) distributed across the scalp that record the cortical activation, pg. 14, third para.).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Modified Levy to incorporate the teachings of Ahmedt-Aristizabal for the benefit of using deep learning on graphs for node-level classification of disease prediction (Ahmedt-Aristizabal, pg. 12, second to the last para.)
Regarding claim 11, Modified Levy teaches the method as claimed in claim 10, Ahmedt-Aristizabal teaches wherein: the at least one device comprises a health monitoring device for monitoring a patient (The required information for diagnosis is obtained from a patient’s medical history, and various medical tests that capture the patient’s functional and anatomical structures through diagnostic imaging methods such as functional magnetic resonance imaging (fMRI), pg. 1, introduction) and
wherein the tuning the edge probabilities of the graph function provides causal relationships between a plurality of health conditions (At the functional level, the graph nodes represent brain regions of interest (ROI), while edges capture the relationships between the regions and their activities, computed via an fMRI correlation matrix, pg. 3, third para.),
wherein the method comprises: using the causal relationships to diagnose a patient (illustrated in Figure 3: (a) Individual graph: nodes are brain regions and edges are functional correlations between time series observations from those regions, pg. 10, last para.).
The same motivation to combine dependent claim 10 applies here.
Regarding claim 12, Modified Levy teaches the method as claimed in claim 10, Ahmedt-Aristizabal teaches wherein the at least one device comprises a health monitoring device for monitoring a patient (The required information for diagnosis is obtained from a patient’s medical history, and various medical tests that capture the patient’s functional and anatomical structures through diagnostic imaging methods such as functional magnetic resonance imaging (fMRI), pg. 1, introduction) and wherein the method comprises, after tuning the edge probabilities of the graph (Then, a GCN predicts the refined subgraph that corresponds to the structure of interest in a supervised setting, where edge probabilities are predicted from learnt edge embeddings, pg. 29, first para.),
the at least one parameter of the first neural network and the at least one parameter of the second neural network: setting the edge probabilities of the graph, the at least one parameter of the first neural network and the at least one parameter of the second neural network (the edge construction strategy can be further improved by incorporating techniques to learn the edge weights such as self-attention weight features (pg. 13, third para.); module works as a filter between the CNN encoder and the GCN decoder to extract more effective semantic and spatial features, pg. 30, first para.);
receiving a further input vector comprising the variables of the input vector, the further input vector having at least one missing value for at least one of the variables and at least one observed value for at least one of the variables (A sparse representation is established that takes into account local and long-range relations between high and low uncertainty elements (pg. 30, second para.); rs-fMRI can observe dysfunction in brain connectivity on BOLD signals, pg. 13, fourth para.); and
applying the first neural network and the second neural network (module works as a filter between the CNN encoder and the GCN decoder to extract more effective semantic and spatial features, pg. 30, first para.) to the further input vector to obtain the at least one missing value, the missing value representing a health condition (a residual learning architecture with graph convolutions to capture brain longitudinal changes to predict missing DMRI data over time, pg. 24, first para.).
The same motivation to combine dependent claim 10 applies here.
20. Claims 13 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Levy et al. ("Estimating latent positions in social and biological networks using Graph Neural Networks in R with GCN4R." bioRxiv preprint doi: https://doi.org/10.1101/2020.11.02.364935; this version posted October 27, 2021) in view of Ahmedt-Aristizabal et al. ("Graph-based deep learning for medical diagnosis and analysis: past, present and future." Sensors 21.14 (2021): 4758, published: 12 July 2021) and further in view of Ozonat et al. (US20200351171)
Regarding claim 13, Modified Levy teaches the method as claimed in claim 10, Ahmedt-Aristizabal teaches wherein the tuning the edge probabilities of the graph provides causal relationships (... edges capture the relationships between the regions and their activities, computed via an fMRI correlation matrix, pg. 3, third para.) between a plurality of sensor measurements (diagnostic imaging methods such as functional magnetic resonance imaging (fMRI), magnetic resonance imaging (MRI), computed tomography (CT), and other diagnostic tools including the electroenchephalogram (EEG) (pg. 1, introduction, first para.); and
The same motivation to combine dependent claim 10 applies here.
Modified Levy does not explicitly teach wherein the method comprises: using the causal relationships to determine at least one fault in the at least one device.
Ozonat teaches wherein the method comprises: using the causal relationships (incorporating the structure of the graph, such as the relationship or dependencies between nodes of the graph (e.g., whether two nodes are directly and/or indirectly related) [0053]) to determine at least one fault in the at least one device (Accordingly, when a failure or anomaly is detected in the outputs Y1 to Y14, it is possible to determine the impacting or triggering nodes by tracing the relationships, and analyzing the observable outputs and hidden states of the connected nodes that function as primary and supplemental indicators [0086]).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Modified Levy to incorporate the teachings of Ozonat for the benefit of designing a sensor based system that can detect anomalies and failures (Ozonat [0016])
Regarding claim 14, Modified Levy teaches the method as claimed in claim 10, Ahmedt-Aristizabal teaches wherein the method comprises, after minimising the loss function by tuning the edge probabilities of the graph (Then, a GCN predicts the refined subgraph that corresponds to the structure of interest in a supervised setting, where edge probabilities are predicted from learnt edge embeddings, pg. 29, first para.),
the at least one parameter of the first neural network and the at least one parameter of the second neural network: setting the edge probabilities of the graph, the at least one parameter of the first neural network and the at least one parameter of the second neural network (the edge construction strategy can be further improved by incorporating techniques to learn the edge weights such as self-attention weight features (pg. 13, third para.); module works as a filter between the CNN encoder and the GCN decoder to extract more effective semantic and spatial features, pg. 30, first para.);
receiving a further input vector comprising the variables of the input vector, the further input vector having at least one missing value for at least one of the variables and at least one observed value for at least one of the variables (A sparse representation is established that takes into account local and long-range relations between high and low uncertainty elements (pg. 30, second para.); rs-fMRI can observe dysfunction in brain connectivity on BOLD signals, pg. 13, fourth para.);
applying the first neural network and the second neural network (module works as a filter between the CNN encoder and the GCN decoder to extract more effective semantic and spatial features, pg. 30, first para.) to the further input vector to obtain the at least one missing value, the missing value representing a state of the device (a residual learning architecture with graph convolutions to capture brain longitudinal changes to predict missing DMRI data over time, pg. 24, first para.);
The same motivation to combine dependent claim 10 applies here.
Modified Levy does not explicitly teach using the missing value to determine a fault in the at least one device.
Ozonat teaches using the missing value to determine a fault in the at least one device (anomalies are detected when ... operator identifies that the output of a given sensor is above or below the threshold value set for that sensor [0016]; Accordingly, when a failure or anomaly is detected in the outputs Y1 to Y14, it is possible to determine the impacting or triggering nodes by tracing the relationships, and analyzing the observable outputs).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Modified Levy to incorporate the teachings of Ozonat for the benefit of designing a sensor based system that can detect anomalies and failures (Ozonat [0016])
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MORIAM MOSUNMOLA GODO whose telephone number is (571)272-8670. The examiner can normally be reached Monday-Friday 8:00am-5:00pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Michelle T. Bechtold can be reached on (571) 431-0762. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/M.G./Examiner, Art Unit 2148
/MICHELLE T BECHTOLD/Supervisory Patent Examiner, Art Unit 2148
Prosecution Insights