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 .
This office action is in response to submission of application on 10/18/2023.
Claims 1-20 are presented for examination.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
Claims 8-13, 15, 17-18, and 20 are rejected under 35 U.S.C. 112(b) as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor, or a joint inventor, regards as the invention.
Claims 8-13, 15, 17-18, and 20 recite the phrase “one or more insights” which is indefinite because it is subjective. An insight could be the shape of a cluster, or the size of a cluster, or something else. MPEP 2173(b)(IV) recites “When a subjective term is used in the claim, the examiner should determine whether the specification supplies some objective standard for measuring the scope of the term. Some objective standard must be provided in order to allow the public to determine the scope of the claim. A claim term that requires the exercise of subjective judgment without restriction may render the claim indefinite.”
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-3, 7-11, and 13-20 are rejected under 35 U.S.C. § 103 as being unpatentable over Deutsch, et al (Zero Shot Learning via Multi-Scale Manifold Regularization, herein Deutsch), and Silva, et al (US 2021/0124780 A1, Graph Search and Visualization For Fraudulent Transaction Analysis, herein Silva).
Regarding claim 1,
Deutsch teaches [a computing platform comprising: a communication interface; at least one processor; at least one non-transitory computer-readable medium; and program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor] to cause the computing platform to:
based on a set of data points defining a feature space, construct a graph structure associated with the set of data points (Deutsch, Figure 1, and page 7113, column 2, paragraph 1, line 1 “We formalize ZSL as the problem of learning a map from the data space X to a semantic descriptions Y (Sect. 3), after making the underlying assumptions and the resulting
limitations explicit (Sec. 2). Our first contribution is to cast the inference process as imposing a differentiable structure on the map h : X → Y , supported on a discrete graph. To address this problem, we use the multi-scale graph transform (Sect. 3.1), which allows us to enforce global regularity without sacrificing local structure.”
PNG
media_image1.png
314
618
media_image1.png
Greyscale
In other words, data space X is a set of data points defining a feature space, and, from Figure 1, construct affinity graph of visual features is construct a graph structure associated with the set of data points.) ;
generate a multi-scale representation that represents the feature space, wherein each feature of a plurality of features from the feature space is represented in a respective plurality of scales with respect to the graph structure (Deutsch, page 7113, column 2, paragraph 2, line 3 “To address this problem, we use the multi-scale graph transform (Sect. 3.1), which allows us to enforce global regularity without sacrificing local structure.” And, page 7115, column 1, paragraph 1, line 1 “In order to characterize the global smoothness of a function f-r ∈ RN, we define its Graph Laplacian quadratic form with respect to the graph as:
PNG
media_image2.png
68
576
media_image2.png
Greyscale
where fr is the graph signal which correspond to an arbitrary dimension r of the semantic representation y, and L denotes the combinatorial graph Laplacian, defined as L = D−W, with D the diagonal degree matrix with entries dii = d(i). The degree d(i) of vertex i is defined as the sum of weights of edges that are connected to i.” In other words, multi-scale graph transform is generate a multi-scale representation that represents the feature space, fr is feature, and di is respective plurality of scales with respect to the graph structure.) ;
regularize the multi-scale representation; based on the regularized multi-scale representation, identify a plurality of clusters associated with the set of data points (Deutsch, page 7115, column 2, paragraph 5, line 7 “Using the regularized semantic representation
(h(xti)) = y∗i for each instance i in the testing set, we perform clustering into ci, i = 1..ct classes by globally partitioning the regularized graph (constructed from y∗i ) using Spectral Clustering [19] or Affinity Propagation [25]).” In other words, regularized semantic representation is regularize the multi-scale representation, and perform clustering is identify a plurality of clusters associated with the set of data points.); and
[transmit, to a client station, data regarding the plurality of clusters and thereby cause an indication of the plurality of clusters to be presented at a user interface of the client station.]
Thus far, Deutsch does not explicitly teach a computing platform comprising: a communication interface; at least one processor; at least one non-transitory computer-readable medium; and program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor. However, it would be obvious to one of ordinary skill in the art that a computing platform is required in order to execute the method. For clarity, Silva is combined with Deutsch.
Silva teaches a computing platform comprising: a communication interface; at least one processor; at least one non-transitory computer-readable medium; and program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor (Silva, paragraph [0027], line 1 “The invention can be implemented in numerous ways, including as a process; an apparatus; a system; a composition of matter; a computer program product embodied on a computer readable storage medium; and/or a processor, such as a processor configured to execute instructions stored on and/or provided by a memory coupled to the processor.” And, claim 10, line 1 “A system, comprising: a communication interface configured to receive a query graph; and one or more processors coupled to the communication interface …” In other words, system is computing platform, communication interface is communication interface, one or mor processors is at least one processor, memory is at least one non-transitory, computer readable medium, and instructions stored on memory is instructions stored on at least one non-transitory computer-readable medium.)
Silva teaches transmit, to a client station, data regarding the plurality of clusters and thereby cause an indication of the plurality of clusters to be presented at a user interface of the client station (Silva, paragraph [0132], line 4 “For example, through the network interface 1216, the processor 1202 can receive information (e.g., data objects or program instructions) from another network, or output information to another network in the course of performing method/process steps.” And, paragraph [0127], line 1 “ FIG. 11 shows an example of a graphical user interface for displaying graphs within a case manager tool.” In other words, output information to another network is transmit to a client station, and graphical user interface is user interface.) .
Both Deutsch and Silva are directed to processing data and generating graphs from the data, among other things. Deutsch teaches based on a set of data points defining a feature space, construct a graph structure associated with the set of data points; generate a multi-scale representation that represents the feature space, wherein each feature of a plurality of features from the feature space is represented in a respective plurality of scales with respect to the graph structure; regularize the multi-scale representation; based on the regularized multi-scale representation, identify a plurality of clusters associated with the set of data points; but does not explicitly teach a computing platform comprising: a communication interface; at least one processor; at least one non-transitory computer-readable medium; and program instructions stored on the at least one non-transitory computer-readable medium; or transmit, to a client station, data regarding the plurality of clusters and thereby cause an indication of the plurality of clusters to be presented at a user interface of the client station. Silva teaches a computing platform comprising: a communication interface; at least one processor; at least one non-transitory computer-readable medium; and program instructions stored on the at least one non-transitory computer-readable medium; and transmit, to a client station, data regarding the plurality of clusters and thereby cause an indication of the plurality of clusters to be presented at a user interface of the client station.
In view of the teaching of Deutsch, it would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Silva into Deutsch. This would result in a computing platform comprising: a communication interface; at least one processor; at least one non-transitory computer-readable medium; and program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor to cause the computing platform to: based on a set of data points defining a feature space, construct a graph structure associated with the set of data points; generate a multi-scale representation that represents the feature space, wherein each feature of a plurality of features from the feature space is represented in a respective plurality of scales with respect to the graph structure; regularize the multi-scale representation; based on the regularized multi-scale representation, identify a plurality of clusters associated with the set of data points; and transmit, to a client station, data regarding the plurality of clusters and thereby cause an indication of the plurality of clusters to be presented at a user interface of the client station.
One ordinary skill in the art would be motivated to do this because graphs are useful for visualizing and analyzing data and finding new ways of being able to display, process, and transmit graphs would be beneficial. (Silva, paragraph [0002], line 1 “Graphs are useful for visualizing and analyzing data, and can help navigate large datasets. For example, graphs can be used to visualize entities and parameters associated with transactions for detecting security attacks and fraud. However, it is challenging to explore and search graphs that represent large datasets because conventional searching techniques are typically slow and do not efficiently
identify similarities between graphs or patterns within a graph.”)
Regarding claim 2,
The combination of Deutsch and Silva teaches the computing platform of claim 1, wherein
the graph structure comprises a graph Laplacian, and wherein the program instructions that are executable by the at least one processor to cause the computing platform to generate a multi-scale representation that represents the feature space, wherein each feature of a plurality of features from the feature space is represented in a respective plurality of scales with respect to the graph structure (Deutsch, page 7113, column 2, paragraph 2, line 3. See mapping of claim 1. In other words, L denotes the combinatorial graph Laplacian is the graph structure comprises a graph Laplacian, and, multi-scale graph is multi-scale graph representation that represents the feature space represented in a respective plurality of scales with respect to the graph structure.) comprise program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor to cause the computing platform to:
compute coefficients corresponding to a polynomial approximation for a Spectral Graph Wavelets (SGW) transform of the graph Laplacian; and initialize multi-scale graph embedding coordinates for the multi-scale representation by computing a respective SGW for each coordinate dimension (Deutsch, page 7115, column 1, paragraph 2, line 3 “Spectral
graph wavelets (SGWs) [13] define a scaling operator in the Graph Fourier domain, based on the eigenvectors of the graph Laplacian L, which can be thought of as an analog of the Fourier transform for functions on weighted graphs.” page 7115, column 1, paragraph 3, line 1 “ SGWs can be computed with a fast algorithm based on approximating the scaled generating kernels by low order polynomials. The wavelet coefficients at each scale can then be computed as a polynomial of L applied to the input data.” In other words, wavelet coefficients can be computed is compute coefficients, SGW is Spectral Graph Wavelets, Fourier transform is transform, and L is graph Laplacian.)
Regarding claim 3,
The combination of Deutsch and Silva teaches the computing platform of claim 1, wherein
the multi-scale representation represents signal of the multi-scale representation in a vertex domain and a spectral domain (Deutsch, page 7115, column 1, paragraph 4 line 1 “After the graph is constructed using the proposed representation of semantic attributes as graph signals, we compute the SGW transform using low-order polynomials of the Laplacian. This way, the SGW coefficients are localized in the vertex domain, since for any two points i ...” In other words, multi-scale representation is previously mapped, signal is signal, vertex domain is vertex domain and SGW transform is spectral domain.)
Regarding claim 7,
The combination of Deutsche and Silva teaches the computing platform of claim 1, wherein the program instructions that are executable by the at least one processor to cause the computing platform to, based on the regularized multi-scale representation, identify a plurality of clusters associated with the set of data points comprise program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor to cause the computing platform to: (Examiner notes that the above limitations are previously mapped in claim 1.)
use an unsupervised machine learning model to output the plurality of clusters (Deutsch, page 7112, column 1, paragraph 3, line 11 “Our approach is performed in a transductive setting, using all unlabeled data where the classification and learning process on the transfer data is entirely unsupervised.” In other words, our approach… is entirely unsupervised is use an unsupervised machine learning model to output the plurality of clusters.).
Regarding claim 8,
The combination of Deutsch and Silva teaches the computing platform of claim 1, wherein: the feature space comprises a plurality of original points; each original point is associated with a plurality of feature dimensions; the regularized multi-scale representation comprises a plurality of coordinates;
each respective coordinate in the regularized multi-scale representation is associated with a corresponding feature dimension of the feature space (Deutsch, Algorithm 1,
PNG
media_image3.png
656
460
media_image3.png
Greyscale
In other words, assign the corresponding coordinate values is a plurality of coordinate values, each associated with a feature dimension of the feature space.); and the program instructions are further executable by the at least one processor to cause the computing platform to: based on the plurality of clusters and the corresponding feature dimensions associated with the coordinates of the regularized multi-scale representation,
derive one or more insights for the plurality of clusters (Deutsch, page 7112, column 2, paragraph 2, line 1 “While our suggested approach is similar in scope to other ”visual-semantic alignment” methods (see [5, 10] and references therein) it is, to the best of our knowledge, the first to use multi-scale localized representations that respect the global (non-flat) geometry of the data space and the fine scale structure of the local semantic representation. Moreover,
learning the relationship between visual features and semantic attributes is unified into a single process, whereas in most ZSL approaches it is divided into a number of independent steps [10].” Examiner notes that the word “insights” is subjective. See paragraph 6. of the instant office action. There is no specific guidance in the specification that identifies exactly what an insight is. Therefore, examiner is interpreting that “one or more insights” is any type of learning. In other words, learning the relationship between visual features and semantic attributes is derive one or more insights for the plurality of clusters.)
Regarding claim 9,
The combination of Deutsch and Silva teaches the computing platform of claim 8, wherein the program instructions that are executable by the at least one processor to cause the computing platform to, based on the plurality of clusters and the corresponding feature dimensions associated with the coordinates of the regularized multi-scale representation,
derive one or more insights for the plurality of clusters comprise program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor to cause the computing platform to: for each respective cluster of the plurality of clusters,
identify one or more feature dimensions from the feature space that have a threshold level of significance for the respective cluster (Deutsch, Algorithm 1, Equations (3) and (4), and page 7115, column 1, paragraph 1, line 4 “…where fr is the graph signal which correspond to an arbitrary dimension r of the semantic representation y, and L denotes the combinatorial graph Laplacian, defined as L = D−W, with D the diagonal degree matrix with entries
dii = d(i).”
PNG
media_image4.png
64
386
media_image4.png
Greyscale
PNG
media_image5.png
44
422
media_image5.png
Greyscale
In other words, dimension r is one or more feature dimensions from the feature space, and if xj Є kNN(xi) is a threshold level of significance for the respective cluster. ) .
Regarding claim 10,
The combination of Deutsch and Silva teaches the computing platform of claim 8, wherein: the computing platform further comprises program instructions that are executable by the at least one processor to cause the computing platform to: for each respective coordinate in the regularized multi-scale representation,
identifying a respective corresponding feature dimension of the feature space associated with the respective coordinate (Deutsch, Algorithm 1, step 3, “ Assign the corresponding coordinate values of the semantic representation in dimension r.” In other words, dimension r is one or more feature dimensions from the feature space, and corresponding coordinate values is associated with the respective coordinate.); and the program instructions that are executable by the at least one processor to cause the computing platform to
regularize the multi-scale representation to, based on the plurality of clusters and the corresponding feature dimensions associated with the coordinates of the regularized multi-scale representation (Deutsch, page 7115, column 2, paragraph 5, line 7 “Using the regularized semantic representation (h(xti)) = y∗i for each instance i in the testing set, we perform clustering into ci, i = 1..ct classes by globally partitioning the regularized graph (constructed from y∗i ) using Spectral Clustering [19] or Affinity Propagation [25]).” In other words, regularized semantic representation is regularize the multi-scale representation, and perform clustering is identify a plurality of clusters associated with the set of data points.) ,
derive one or more insights for the plurality of clusters comprise program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor (Deutsch, page 7112, column 2, paragraph 2, line 1 “While our suggested approach is similar in scope to other ”visual-semantic alignment” methods (see [5, 10] and references therein) it is, to the best of our knowledge, the first to use multi-scale localized representations that respect the global (non-flat) geometry of the data space and the fine scale structure of the local semantic representation. Moreover, learning the relationship between visual features and semantic attributes is unified into a single process, whereas in most ZSL approaches it is divided into a number of independent steps [10].” See mapping of claim 8 for interpretation of “insights”. In other words, learning the relationship between visual features and semantic attributes is derive one or more insights for the plurality of clusters.) to cause the computing platform to: based on the plurality of clusters and the identified corresponding feature dimensions of the feature space associated with the respective coordinates,
derive one or more insights for the plurality of clusters (Deutsch, page 7112, column 2, paragraph 2, line 1 “While our suggested approach is similar in scope to other ”visual-semantic alignment” methods (see [5, 10] and references therein) it is, to the best of our knowledge, the first to use multi-scale localized representations that respect the global (non-flat) geometry of the data space and the fine scale structure of the local semantic representation. Moreover, learning the relationship between visual features and semantic attributes is unified into a single process, whereas in most ZSL approaches it is divided into a number of independent steps [10].” See mapping of claim 8 for interpretation of “insights”. In other words, learning the relationship between visual features and semantic attributes is derive one or more insights for the plurality of clusters.) .
Regarding claim 11,
The combination of Deutsch and Silva teaches the computing platform of claim 8, wherein the program instructions that are executable by the at least one processor to cause the computing platform to, based on the plurality of clusters and the corresponding feature dimensions associated with the coordinates of the regularized multi-scale representation,
derive one or more insights for the plurality of clusters comprise program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor to cause the computing platform to: for each respective cluster,
assign a clustering label to the respective cluster; and use the assigned clustering labels to derive the one or more insights for the plurality of clusters (Deutsch, Figure 1, In other words, determine labels using Spectral Clustering is assigning a clustering label to the respective cluster, and labels of unseen test set is use the clustering labels to derive one or more insights for the plurality of clusters.).
Regarding claim 13,
The combination of Deutsch and Silva teaches the computing platform of claim 8, wherein the computing platform further comprises program instructions that are executable by the at least one processor to cause the computing platform to:
transmit, to a second client station, data defining the one or more insights and thereby cause an indication of the one or more insights to be presented at a user interface of the second client station (Silva, paragraph [0132], line 4 “For example, through the network interface 1216, the processor 1202 can receive information (e.g., data objects or program instructions) from another network, or output information to another network in the course of performing method/process steps.” And, paragraph [0127], line 1 “ FIG. 11 shows an example of a graphical user interface for displaying graphs within a case manager tool.” Examiner notes that outputting to “another network” is the same as transmitting to one or more client stations since there could be numerous clients on the network. In other words, output information to another network is transmit to a second client station, and graphical user interface is user interface of the second client station.) .
Regarding claim 14,
The combination of Deutsch and Silva teaches the computing platform of claim 13, wherein
the client station and the second client station are the same client station (Silva, paragraph [0132], line 4. See above mapping regarding client stations.)
Regarding claim 15,
The combination of Deutsch and Silva teaches the computing platform of claim 14, wherein
the indication of the plurality of clusters and the indication of the one or more insights are presented at a same time (Deutsch, page 7112, column 2, paragraph 2, line 1 “While our suggested approach is similar in scope to other ”visual-semantic alignment” methods (see [5, 10] and references therein) it is, to the best of our knowledge, the first to use multi-scale localized representations that respect the global (non-flat) geometry of the data space and the fine scale structure of the local semantic representation. Moreover, learning the relationship between visual features and semantic attributes is unified into a single process, whereas in most ZSL approaches it is divided into a number of independent steps [10].” In other words, learning the relationship between visual features is one or more insights, and the relationship between visual features and semantic attributes is unified into a single process is the indication of the plurality of clusters and the indication of the one or more insights are presented at the same time.)
Claims 16-18 are non-transitory computer-readable medium claims that correspond to computing platform claims 1, and 8-9, respectively. Otherwise, they are not patentably distinct. The combination of Deutsch and Silva teaches a non-transitory computer-readable medium (Silva, paragraph [0027], line 1 “The invention can be implemented in numerous ways, including as a process; an apparatus; a system; a composition of matter; a computer program product embodied on a computer readable storage medium.” In other words, a computer readable storage medium is a non-transitory computer-readable medium.) Therefore, claims 16-18 are rejected for the same reasons as claims 1, and 8-9, respectively.
Claims 19-20 are method claims that correspond to computer platform claims 1 and 8, respectively. Otherwise, they are not patentably distinct. The combination of Deutsch and Silva teaches a method (Silva, claim 19, line 1 “method, comprising…” In other words, method is method.) Therefore, claims 19-20 are rejected for the same reasons as claims 1 and 8, respectively.
Claim 4 is rejected under 35 U.S.C. § 103 as being unpatentable over Deutsch, Silva, and McInnes, et al (UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction, herein McInnes).
Regarding claim 4,
The combination of Deutsch and Silva teaches the computing platform of claim 1, wherein the program instructions that are executable by the at least one processor to cause the computing platform to regularize the multi-scale representation comprise program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor to cause the computing platform to:
Thus far, the combination of Deutsch and Silva does not explicitly teach optimize features of the multi-scale representation by using stochastic gradient descent with respect to the graph structure.
McInnes teaches optimize features of the multi-scale representation by using stochastic gradient descent with respect to the graph structure ( McInnes, page 12, paragraph 2, line 1 “Similar to t-SNE we can optimize the embedding Y with respect to fuzzy set cross entropy C by using stochastic gradient descent.” In other words, embedding Y is features, and optimize…by using stochastic gradient descent is optimize…by using stochastic gradient descent.)
Both McInnes and the combination of Deutsch and Silva are directed to manifold learning, among other things. The combination of Deutsch and Silva teach the computer platform of claim 1, but does not explicitly teach optimize features of the multi-scale representation by using stochastic gradient descent with respect to the graph structure. McInnes teaches optimize features of the multi-scale representation by using stochastic gradient descent with respect to the graph structure.
In view of the teaching of the combination of Deutsch and Silva it would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of McInnes into the combination of Deutsch and Silva. This would result in the apparatus of claim 1, and optimizing features of the multi-scale representation by using stochastic gradient descent with respect to the graph structure.
One of ordinary skill in the art would be motivated to do this because dimension reduction is an important part of data science, and finding an algorithm that is scalable and efficient is important for increasing sizes of datasets. (McInnes, page 1, paragraph 2, line 1 “Dimension reduction plays an important role in data science, being a fundamental technique in both visualization and as pre-processing for machine learning. Dimension reduction techniques are being applied in a broadening range of fields and on ever increasing sizes of datasets. It is thus desirable to have an algorithm that is both scalable to massive data and able to cope with the diversity of data available.”).
Claim 5 is rejected under 35 U.S.C. § 103 as being unpatentable over Deutsch, Silva, and Qin, et al (Skeleton-based action recognition by part-aware graph convolutional networks, herein Qin).
Regarding claim 5,
The combination of Deutsch and Silva teaches the computing platform of claim 1, wherein the program instructions that are executable by the at least one processor to cause the computing platform to regularize the multi-scale representation comprise program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor to cause the computing platform to: for each respective feature of the multi-scale representation,
Thus far, the combination of Deutsch and Silva does not explicitly teach (i) concatenate scales and feature dimensions corresponding to the respective feature and (ii) optimize the respective feature of the multi-scale representation using the concatenated scales and feature dimensions corresponding to the respective feature.
Qin teaches (i) concatenate scales and feature dimensions corresponding to the respective feature and (ii) optimize the respective feature of the multi-scale representation using the concatenated scales and feature dimensions corresponding to the respective feature (Qin, abstract, line 3 “For scale invariance on multi-scale data, an Inception-like structure is introduced, which can concatenate feature maps from different convolution kernels.” In other words, multi-scale data is scale, feature maps is respective feature, and concatenate maps from different kernels is concatenate scales and feature dimensions. )
Both Qin and the combination of Deutsch and Silva are directed to machine learning, and representing data as graphs, among other things. The combination of Deutsch and Silva teaches the computing platform of claim 1, but does not explicitly teach (i) concatenate scales and feature dimensions corresponding to the respective feature and (ii) optimize the respective feature of the multi-scale representation using the concatenated scales and feature dimensions corresponding to the respective feature. Qin teaches (i) concatenate scales and feature dimensions corresponding to the respective feature and (ii) optimize the respective feature of the multi-scale representation using the concatenated scales and feature dimensions corresponding to the respective feature.
In view of the teaching of the combination of Deutsch and Silva, it would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Qin into the combination of Deutsch and Silva. This would result in the computing platform of claim 1, and (i) concatenating scales and feature dimensions corresponding to the respective feature and (ii) optimizing the respective feature of the multi-scale representation using the concatenated scales and feature dimensions corresponding to the respective feature.
One of ordinary skill in the art would be motivated to do this because it is difficult to efficiently process multi-scale data and new methods are needed. (Qin, abstract, line 3 “For scale invariance on multi-scale data, an Inception-like structure is introduced, which can concatenate feature maps from different convolution kernels. In contrast to methods based
on LSTMs, the model presented is capable of extracting both temporal and spatial features from input data.”)
Claim 12 is rejected under 35 U.S.C. § 103 as being unpatentable over Deutsch, Silva, and Marcilio, et al (Explaining dimensionality reduction results using Shapley values, herein Marcilio).
Regarding claim 12,
The combination of Deutsch and Silva teaches the computing platform of claim 11, wherein: the coordinates of the regularized multi-scale representation comprise regularized wavelet coefficients; and the program instructions that are executable by the at least one processor to cause the computing platform to use the assigned clustering labels to derive the one or more insights for the plurality of clusters comprise program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor to cause the computing platform to:
Thus far, the combination of Deutsch and Silva does not explicitly teach determine Shapley values of the regularized wavelet coefficients by using the assigned clustering labels and employing a Shapley algorithm to the regularized multi-scale representation; and based on the determined Shapley values and the corresponding original feature dimensions associated with the coordinates of the regularized multi-scale representation, identify, for each respective cluster of at least a subset of the plurality of clusters, respective one or more original features from the original feature space that have a threshold level of significance for the respective cluster.
Marcilio teaches determine Shapley values of the regularized wavelet coefficients by using the assigned clustering labels and employing a Shapley algorithm to the regularized multi-scale representation (Marcilio, Algorithm 2, and, column 1, paragraph 3, line 1 “After the cluster definition, we can generate Shapley values for each data sample. Thus, we need to define a model f that returns the prediction probabilities for a data sample x based on the cluster definition – discussed in the previous section.”
PNG
media_image6.png
148
390
media_image6.png
Greyscale
PNG
media_image7.png
204
386
media_image7.png
Greyscale
In other words, generate Shapley values is determine Shapley values, after cluster definition is using cluster labels, and Shapley algorithm is Shapley algorithm.)
Marcilio teaches based on the determined Shapley values and the corresponding original feature dimensions associated with the coordinates of the regularized multi-scale representation, identify, for each respective cluster of at least a subset of the plurality of clusters, respective one or more original features from the original feature space that have a threshold level of significance for the respective cluster (Marcilio, page 4, column 1, paragraph 4, line 1 “To return the prediction probabilities for a data sample x, we measure the distance from x to each cluster centroid. Fig. 1 (2.a – bottom) illustrates such a process for three cluster centroids (
PNG
media_image8.png
22
120
media_image8.png
Greyscale
) and, consequently, a three-dimensional distance vector. To convert the distances into probabilities, we apply an L1 normalization. The Shapley estimator (in our case, KernelSHAP (Lundberg & Lee, 2017)) uses these probabilities (for each data sample) to generate explanations discussed in Section 3. Notice that while estimating the Shapley values using KernelSHAP accounts for most of the dataset, we only compute the estimation for 20% of the data. The result of this procedure will be a matrix of dimensions n × M for each cluster, where n corresponds to 20% of the dataset size and M represents the dimensionality of the dataset – each cell i, j of the matrix will contain the Shapley value of the datapoint i for the feature j.” And, page 4, column 2, paragraph 3, line 1 “The estimated Shapley values correspond to each feature’s contribution to the dimensionality reduction result. Thus, a feature with a high absolute Shapley value contributes a lot to the projected dataset cluster formation. In this case, each data point used for Shapley values estimation contains the correspondent Shapley value. Negative Shapley values mean that a feature contributes to the cluster formation, and positive Shapley values mean that a feature does not contribute to cluster formation.” In other words, correspond to feature’s contribution is corresponding feature, matrix of dimensions for each cluster is dimensions associated with each respective cluster, and Shapley value that is positive is a threshold level of significance for the respective cluster.)
Both Marcilio and the combination of Deutsch and Silva are directed to machine learning, and dimensionality reduction of data, among other things. The combination of Deutsch and Silva teaches the computer platform of claim 11, but does not explicitly teach determine Shapley values of the regularized wavelet coefficients by using the assigned clustering labels and employing a Shapley algorithm to the regularized multi-scale representation; and based on the determined Shapley values and the corresponding original feature dimensions associated with the coordinates of the regularized multi-scale representation, identify, for each respective cluster of at least a subset of the plurality of clusters, respective one or more original features from the original feature space that have a threshold level of significance for the respective cluster. Marcilio teaches determine Shapley values of the regularized wavelet coefficients by using the assigned clustering labels and employing a Shapley algorithm to the regularized multi-scale representation; and based on the determined Shapley values and the corresponding original feature dimensions associated with the coordinates of the regularized multi-scale representation, identify, for each respective cluster of at least a subset of the plurality of clusters, respective one or more original features from the original feature space that have a threshold level of significance for the respective cluster.
In view of the teaching of the combination of Deutsch and Silva, it would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Marcilio into the combination of Deutsch and Silva. This would result in the computer platform of claim 11, and determining Shapley values of the regularized wavelet coefficients by using the assigned clustering labels and employing a Shapley algorithm to the regularized multi-scale representation; and based on the determined Shapley values and the corresponding original feature dimensions associated with the coordinates of the regularized multi-scale representation, identify, for each respective cluster of at least a subset of the plurality of clusters, respective one or more original features from the original feature space that have a threshold level of significance for the respective cluster.
One of ordinary skill in the art would be motivated to do this because interpreting teach feature’s contributions can help improve the value of dimensionality reduction techniques. (Marcilio, abstract, line 1 “Dimensionality reduction (DR) techniques have been consistently supporting high-dimensional data analysis in various applications. Besides the patterns uncovered by these techniques, the interpretation of DR results based on each feature’s contribution to the low-dimensional representation supports new finds through exploratory analysis. Current literature approaches designed to interpret DR techniques do not explain the features’ contributions well since they focus only on the low-dimensional representation or do not consider the relationship among features.”)
The prior art made of record and not used is considered pertinent to applicant’s disclosure:
Makarov, et al “Survey on graph embeddings and their applications to machine learning problems on graphs” discloses the core concepts of graph embeddings and provides several taxonomies for their description and explains three types of graph embedding models based on matrix factorization, random-walks and deep learning approaches.
Mishne, et al “Data-Driven Tree Transforms and Metrics” discloses multiscale data-driven transforms and metrics based on trees. Their construction is implemented in an iterative refinement procedure that exploits the co-dependencies between features and observations.
Kosman, et al, US 2023/0101250 A1, “Method for Generating a Graph Structure for Training a Graph Neural Network” discloses a method for generating a graph structure for training a graph neural network. The method includes: obtaining data representing a computational graph, wherein the computational graph comprises a plurality of nodes connected by edges; and generating the graph structure for training the graph neural network by removing edges from the computational graph.
Shen, et al, WO 2024/144808 A1, “Machine-Learning for Content Interaction” discloses a system that can generate content recommendations and facilitate interactions using machine-learning by receiving entity data and interaction data associated with a target entity and generating at least a first graph structure and a second graph structure.
Allowable Subject Matter
Claim 6 is objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BART RYLANDER whose telephone number is (571)272-8359. The examiner can normally be reached Monday - Thursday 8:00 to 5:30.
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, Miranda Huang can be reached at 571-270-7092. 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.
/Bart I Rylander/Examiner, Art Unit 2124
Prosecution Insights