Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections – 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1: Claims 8-14 are method claims. Claims 1-7, 15-20 are machine/system/product claims. Therefore, claims 1-20 are directed to either a process, machine, manufacture or composition of matter.
With respect to claim 1:
Step 2A – Prong 1:
…
…
and processing the set of training data under the one or more constraints, in a machine learning model, wherein an operation of the machine learning model includes: generating a hierarchical feature graph classifying data points in the set of training data; (mental process – a person can manually generate a hierarchical feature graph classifying data points with the assistance of a pen/paper.)
discovering a plurality of classification rules applicable to the data points using a linear program problem or a quadratic program problem; (mental process – a person can manually discover a plurality of classification rules applicable to the data points using a linear program problem or a quadratic program problem with the assistance of a pen/paper.)
assigning a weighted value to each rule applied to the data points, to generate weighted classification rules; (mental process – a person can manually assign a weighted value to each rule applied to the data points, to generate weighted classification rules with the assistance of a pen/paper.)
assigning a combination of the weighted rules to a plurality of the data points; (mental process – a person can manually assign a combination of the weighted rules to a plurality of the data points with the assistance of a pen/paper.)
and generating interpretable classification trees from the data points based on the combination of weighted rules of respective data points. (mental process – a person can manually generate interpretable classification trees from the data points based on the combination of weighted rules of respective data points with the assistance of a pen/paper.)
Step 2A – Prong 2: This judicial exception is not integrated into a practical application.
A computer program product for generating interpretable weighted classification trees, the computer program product comprising: one or more computer readable storage media, and program instructions collectively stored on the one or more computer readable storage media, the program instructions comprising: (mere instructions to apply the exception using a generic computer component – storage, computer instructions apply exception.)
receiving, by a processor, a set of training data and one or more constraints to be applied to the set of training data; (Adding insignificant extra-solution activity to the judicial exception - see MPEP 2106.05(g)).
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
A computer program product for generating interpretable weighted classification trees, the computer program product comprising: one or more computer readable storage media, and program instructions collectively stored on the one or more computer readable storage media, the program instructions comprising: (mere instructions to apply the exception using a generic computer component – storage, computer instructions apply exception.)
receiving, by a processor, a set of training data and one or more constraints to be applied to the set of training data; (MPEP 2106.05(d)(II) indicate that merely “Receiving or transmitting data over a network, e.g., using the Internet to gather data” is a well‐understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim – the training data and one or more constraints is merely received by the processor.) Thereby, a conclusion that the claimed distribute step is well-understood, routine, conventional activity is supported under Berkheimer.)
With respect to claim 2:
Step 2A – Prong 1:
The computer program product of claim 1, wherein the program instructions further comprise: in response to receiving the set of training data, sorting, … a list of features associated with the set of training data in an order of importance based on a score derived from a black-box prediction model; (mental process – a person can manually sort a list of features associated with the set of training data in an order of importance based on a score derived from a black-box prediction model with the assistance of a pen/paper.)
given a feature f, creating a node for each distinct feature value, Lf, wherein: a last node Lfl denotes a SKIP node, a path passes thru the SKIP node, the feature f is not part of a rule, and numerical features, and discretized and cumulative bins are stored symbolically; (mental process – a person can recognize that the last node Lfl denotes a SKIP node, a path passes thru the SKIP node, the feature f is not part of a rule, and numerical features, and discretized and cumulative bins are stored symbolically.)
creating a source node o and a sink node s; (mental process – a person can manually create a source node o and a sink node s with the assistance of a pen/paper.)
generating a node set FROM = {o}, wherein a feature indexf= 0; (mental process – a person can manually generate a node set FROM = {o}, with the assistance of a pen/paper.)
and connecting the node set FROM to one or more TO nodes by directed arcs, and adding to an arc set A: {ni,f} X {njf+1}, for all i, j, and iff= N-1, TO node = SINK s, and iff< N,f=f+1, else STOP and Return the arc set A; (mental process – a person can manually connecting the node set FROM to one or more TO nodes by directed arcs, and adding to an arc set A: {ni,f} X {njf+1}, for all i, j, and iff= N-1, TO node = SINK s, and iff< N,f=f+1, else STOP and Return the arc set A with the assistance of a pen/paper.)
wherein a total number of nodes = Ef L f+2, Arcs = E f(Lf * Lf+1); and wherein total feasible rules = prod{L_f} (mental process – a person can recognize that the total number of nodes = Ef L f+2, Arcs = E f(Lf * Lf+1); and wherein total feasible rules = prod{L_f}.)
Step 2A – Prong 2: This judicial exception is not integrated into a practical application.
… by a computer, … (mere instructions to apply the exception using a generic computer component – computer applies exception.)
No additional elements recited.
With respect to claim 3:
Step 2A – Prong 1:
The computer program product of claim 2, wherein the program instructions further comprise: generating an incidence matrix aij = 1 if features in rule j are present in sample i; (mental process – a person can manually generating an incidence matrix aij = 1 if features in rule j are present in sample i with the assistance of a pen/paper.)
and including in the generation of the hierarchical feature graph, a coverage penalty c, to force each observation to be covered by a classification rule; (mental process – a person can manually including in the generation of the hierarchical feature graph, a coverage penalty c, to force each observation to be covered by a classification rule with the assistance of a pen/paper.)
wherein I(i) is a true label class of sample i; (mental process – a person can recognize that I(i) is a true label class of sample i)
wherein K is a corresponding proportion of class I(i) in rule j; (mental process – a person can recognize that K is a corresponding proportion of class I(i) in rule j)
and wherein a Gini index of rule j, gj is measuring impurity and purity of a rule. (mental process – a person can recognize Gini index of rule j, gj is measuring impurity and purity of a rule)
With respect to claim 4:
Step 2A – Prong 1:
The computer program product of claim 3, wherein beta=1 corresponds to the linear programming problem and beta=2 corresponds to the quadratic programming problem. (mental process – a person can recognize that beta=1 corresponds to the linear programming problem and beta=2 corresponds to the quadratic programming problem.)
With respect to claim 5:
Step 2A – Prong 1:
The computer program product of claim 1, wherein the program instructions further comprise generating optimal multiway split regression trees including coefficients. (mental process – a person can recognize that program instructions further comprise generating optimal multiway split regression trees including coefficients.)
With respect to claim 6:
Step 2A – Prong 1:
The computer program product of claim 1, wherein the program instructions further comprise factoring in one or more inter-rule and intra-rule constraints in generating the hierarchical feature graph. (mental process – a person can recognize that the program instructions further comprise factoring in one or more inter-rule and intra-rule constraints in generating the hierarchical feature graph.)
With respect to claim 7:
Step 2A – Prong 1:
The computer program product of claim 1, wherein a total value of weighted values assigned to rules applied to a data point equals 1. (mental process – a person can recognize that the total value of weighted values assigned to rules applied to a data point equals 1.)
Claims 8-14 are rejected on the same grounds under 35 U.S.C. 101 as claims 1-7 as they are substantially similar, respectively. Mutatis mutandis.
Claim 15 is substantially similar to claim 1, but has the following additional elements:
With respect to claim 15:
Step 2A – Prong 2:
A computing device for generating an interpretable predictive model, comprising: a processor; a memory coupled to the processor, the memory storing instructions configured to cause the processor to perform acts comprising: (mere instructions to apply the exception using a generic computer component – processor, memory apply exception.)
With respect to claim 16:
Step 2A – Prong 1:
The computing device of claim 15, wherein the instructions cause the processor to perform an additional act comprising assigning an error value threshold to the linear program problem or to the quadratic program problem. (mental process – a person can recognize that the instructions cause the processor to perform an additional act comprising assigning an error value threshold to the linear program problem or to the quadratic program problem.)
Claim 17 is rejected on the same grounds under 35 U.S.C. 101 as claim 6 as they are substantially similar. Mutatis mutandis.
With respect to claim 18:
Step 2A – Prong 1:
The computing device of claim 15, wherein the instructions cause the processor to perform an additional act comprising including a purity value threshold for discovering classification rules in the linear program problem or in the quadratic program problem. (mental process – a person can recognize that the instructions cause the processor to perform an additional act comprising including a purity value threshold for discovering classification rules in the linear program problem or in the quadratic program problem.)
With respect to claim 19:
Step 2A – Prong 1:
The computing device of claim 15, wherein the instructions cause the processor to perform an additional act comprising optimizing the linear program problem or the quadratic program problem to discover classification rules prioritizing one of a misclassification error value or a purity value for rule discovery. (mental process – a person can recognize that the instructions cause the processor to perform an additional act comprising optimizing the linear program problem or the quadratic program problem to discover classification rules prioritizing one of a misclassification error value or a purity value for rule discovery.)
Claim 20 is rejected on the same grounds under 35 U.S.C. 101 as claim 7 as they are substantially similar. Mutatis mutandis.
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.
Claims 1, 5, 7-8, 12, 14-16, 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Shabtai et al. (US20220230070A1) hereinafter known as Shabtai in view of Benard et al. (“SIRUS: Stable and Interpretable RUle Set for Classification”) hereinafter known as Benard in view of Lumadjeng et al. (“Rule Generation for Classification: Scalability, Interpretability, and Fairness”) hereinafter known as Lumadjeng.
Regarding independent claim 1, Shabtai teaches:
A computer program product for generating interpretable weighted classification trees, the computer program product comprising: one or more computer readable storage media, and program instructions collectively stored on the one or more computer readable storage media, the program instructions comprising: (Shabtai [0183]: “The translated code includes, among other things, operation codes (opcodes) which are computer instructions that defines the operations to be performed” Shabtai teaches instructions that define the operations to be performed, which are in the form of storable opcodes.)
receiving, by a processor, a set of training data and one or more constraints to be applied to the set of training data; (Shabtai [0033]: “defining a cost function that represents the acceptable cost while considering the constrains; providing a plurality of analysis and processing modules in a computational environment, for processing data associated with the computational environment and returning results” Shabtai teaches processing data to be trained and defining constraints with respect to a cost function.)
Shabtai does not explicitly teach:
and processing the set of training data under the one or more constraints, in a machine learning model, wherein an operation of the machine learning model includes: generating a hierarchical feature graph classifying data points in the set of training data;
discovering a plurality of classification rules applicable to the data points …
… using a linear program problem or a quadratic program problem;
assigning a weighted value to each rule applied to the data points, to generate weighted classification rules;
assigning a combination of the weighted rules to a plurality of the data points;
and generating interpretable classification trees from the data points based on the combination of weighted rules of respective data points.
However, Benard teaches:
and processing the set of training data under the one or more constraints, in a machine learning model, wherein an operation of the machine learning model includes: generating a hierarchical feature graph classifying data points in the set of training data; (Benard [Page 7, Paragraph 1]: “a path describes the sequence of splits to go from the root of the tree to a specific (inner or terminal) node” Benard teaches that the decision tree is hierarchical and contains features (see Figure 1 right above this excerpt) as there is a variable index j_k used at each split (see equation below this excerpt).)
discovering a plurality of classification rules applicable to the data points … (Benard [Page 8, Paragraph 1]: “an elementary rule … returns the empirical probability that the output Y is of class 1 conditional on whether the query point x belongs to H_n(P) or not” Benard teaches discovering rule that gives the probability of whether a data point belongs to the rule’s hyperrectangle or not.)
…
assigning a weighted value to each rule applied to the data points, to generate weighted classification rules; (Benard [Page 8, Paragraph 1]: “an elementary rule … returns the empirical probability that the output Y is of class 1 conditional on whether the query point x belongs to H_n(P) or not” Benard teaches discovering rule that gives the probability of whether a data point belongs to the rule’s hyperrectangle or not. Benard [Page 10, Eq. 3.3]: Benard teaches assigning every rule a weighted classification rule that is inversely proportional to the number of selected rules.)
assigning a combination of the weighted rules to a plurality of the data points; (Benard [Page 9, last paragraph]: “the resulting small set of rules … is combined to form a simple, compact, and stable rule classification model. We simply average the set of elementary rules … that have been selected in the first steps of SIRUS” Benard teaches evaluating the selected rules for each point and combining them by averaging.)
and generating interpretable classification trees from the data points based on the combination of weighted rules of respective data points. (Benard [Page 7, Paragraph 1]: “a path describes the sequence of splits to go from the root of the tree to a specific (inner or terminal) node. Since a hyperrectangle is associated to each node, a rule can be defined as a piecewise constant estimate with this hyperrectangle as support” Benard teaches using a path to sequence through the tree via splits to go from the root to a specific node using rules.)
Shabtai and Benard are in the same field of endeavor as the present invention, as the references are directed to policy implementation with constraints, which in the latter takes place in the form of rules in tree structure. It would have been obvious, before the effective filing date of the claimed invention, to a person of ordinary skill in the art, to combine constraining data and evaluating it based on a score as taught in Shabtai with implementing a hierarchical decision tree for the rules as taught in Benard. Benard provides this additional functionality. As such, it would have been obvious to one of ordinary skill in the art to modify the teachings of Shabtai to include teachings of Benard because the combination would allow for the weighted combination of rules by traversing the tree. This has the potential benefit of speeding up finding rules, as traversing a tree may be more efficient than traditional mixing optimization methods.
Shabtai and Benard do not explicitly teach:
… using a linear program problem or a quadratic program problem;
However, Lumadjeng teaches:
… using a linear program problem or a quadratic program problem; (Lumadjeng [Page 5, Section 3.1]: “we are ready to present the LP model of our master problem … measure of classification accuracy of rule j for sample i given that sample is covered by the rule, and the coefficients … stand for the costs of rules” Lumadjeng formulates a LP problem and generates candidate rules from the leaves of a weighted decision tree.)
Lumadjeng is in the same field as the present invention, since it is directed to new rule-based optimization method for classification with constraints. It would have been obvious, before the effective filing date of the claimed invention, to a person of ordinary skill in the art, to combine the weighted combination of rules by traversing graphs as taught in Shabtai as modified by Benard with using a linear programming problem to optimize the rule as taught in Lumadjeng. Lumadjeng provides this additional functionality. As such, it would have been obvious to one of ordinary skill in the art to modify the teachings of Shabtai as modified by Benard to include teachings of Lumadjeng because the combination would allow for certain values to be minimized via linear programming. This has the potential benefit of constraining the rules effectively such that undesired values are optimally minimized, to the furthest extent possible.
Regarding dependent claim 5, Shabtai, Benard, and Lumadjeng teach:
The computer program product of claim 1, wherein the program instructions further comprise generating optimal multiway split regression trees including coefficients. (Lumadjeng [Page 8, Paragraph 3]: “proxy PSP, which is based on growing a decision trees with sample weights that correspond to the dual optimal variables. After constructing the tree, we visit each leaf of the tree and check whether the resulting rule has a negative reduced cost” Lumadjeng [Page 5, 3.2]: “exponentially many rules due to all possible combinations of different levels of the considered features” Lumadjeng teaches generating rules from decision-tree leaves. Since the candidate rule space is exponential large as all possible combinations are considered, the branches that are alternatives each the multiway split.)
The reasons to combine are substantially similar to those of claim 1.
Regarding dependent claim 7, Shabtai, Benard, and Lumadjeng teach:
The computer program product of claim 1, wherein a total value of weighted values assigned to rules applied to a data point equals 1. (Benard [Page 9, last paragraph]: “the resulting small set of rules … is combined to form a simple, compact, and stable rule classification model. We simply average the set of elementary rules … that have been selected in the first steps of SIRUS” Benard teaches evaluating the selected rules for each point and combining them by averaging. Benard [Page 10, Equation 3.3]: Benard teaches that each of the rules gets a coefficient of 1 divided by the total number of rules, showing that the total of all coefficients equals 1.)
The reasons to combine are substantially similar to those of claim 1.
Claim 8 is rejected on the same grounds under 35 U.S.C. 103 as claim 1 as they are substantially similar. Mutatis mutandis.
Claims 12, 14 are rejected on the same grounds under 35 U.S.C. 103 as claims 5, 7 as they are substantially similar, respectively. Mutatis mutandis.
Claim 15 is substantially similar to claim 1, but has the following additional elements:
Regarding independent claim 15, Shabtai, Benard, and Lumadjeng teach:
A computing device for generating an interpretable predictive model, comprising: a processor; a memory coupled to the processor, the memory storing instructions configured to cause the processor to perform acts comprising: (Shabtai [0089]: “at least one processor and associated memory” Shabtai teaches a processor and memory that can execute and store instructions.)
The reasons to combine are substantially similar to those of claim 1.
Regarding dependent claim 16, Shabtai, Benard, and Lumadjeng teach:
The computing device of claim 15, wherein the instructions cause the processor to perform an additional act comprising assigning an error value threshold to the linear program problem or to the quadratic program problem. (Shabtai [0138]: “Incorrect classification is defined as a confidence threshold above 0.5 for a benign file or one that is equal or smaller than 0.5 for a malicious file.” Shabtai teaches a error threshold of 0.5 where above that means that benign and below means a malicious file.)
The reasons to combine are substantially similar to those of claim 1.
Regarding dependent claim 19, Shabtai, Benard, and Lumadjeng teach:
The computing device of claim 15, wherein the instructions cause the processor to perform an additional act comprising optimizing the linear program problem or the quadratic program problem to discover classification rules prioritizing one of a misclassification error value or a purity value for rule discovery. (Lumadjeng [Page 5, Equation 3]: Lumadjeng teaches a LP that minimizes classification loss and allows for additional constraints.
The reasons to combine are substantially similar to those of claim 1.
Claim 20 is rejected on the same grounds under 35 U.S.C. 103 as claim 7 as they are substantially similar. Mutatis mutandis.
Claims 2-4, 6, 9-11, 13, 17-18 are rejected under 35 U.S.C. 103 as being unpatentable over Shabtai in view of Benard in view of Lumadjeng in view of Aghaei et al. (“Strong Optimal Classification Trees”) hereinafter known as Aghaei.
Regarding dependent claim 2, Shabtai, Benard, and Lumadjeng teach:
The computer program product of claim 1, wherein the program instructions further comprise: in response to receiving the set of training data, sorting, by a computer, a list of features associated with the set of training data in an order of importance based on a score derived from a black-box prediction model; (Shabtai [0133]: “it selects the 300 features with the highest DF values. Using the selected features, a Random Forest classifier with 100 trees was trained.” Shabtai teaches that the 300 features with the highest DF values were selected, showing that there is a sorting based on a score to use to train the random forest.)
…
… a last node Lfl denotes a SKIP node, a path passes thru the SKIP node, the feature f is not part of a rule, … (Benard [Page 7, Paragraph 1]: “a path describes the sequence of splits to go from the root of the tree to a specific (inner or terminal) node. Since a hyperrectangle is associated to each node, a rule can be defined as a piecewise constant estimate with this hyperrectangle as support” Benard teaches using a path to sequence through the tree via splits to go from the root to a specific node using rules. The rule/path contains only the coordinates actually split on, which makes the unused alternative as effectively a skip node.)
Shabtai, Benard, and Lumadjeng do not explicitly teach:
given a feature f, creating a node for each distinct feature value, Lf, wherein: … and numerical features, and discretized and cumulative bins are stored symbolically;
…
creating a source node o and a sink node s;
generating a node set FROM = {o}, wherein a feature indexf= 0;
and connecting the node set FROM to one or more TO nodes by directed arcs, and adding to an arc set A: {ni,f} X {njf+1}, for all i, j, and iff= N-1, TO node = SINK s, and iff< N,f=f+1, else STOP and Return the arc set A;
wherein a total number of nodes = Ef L f+2, Arcs = E f(Lf * Lf+1); and wherein total feasible rules = prod{L_f}
However, Aghaei teaches:
given a feature f, creating a node for each distinct feature value, Lf, wherein: … and numerical features, and discretized and cumulative bins are stored symbolically; (Aghaei [Page 9, Remark 1]: “this formulation can also be applied to datasets involving categorical or integer features … for each level of the feature, we create a new binary column” Aghaei creates a binary column for each categorical level and cumulative binary columns for integer values. Each encoded level/value is analogous to a feature-layer node as the features are discrete and cumulative bins are stored as a symbol (0 or 1).)
…
creating a source node o and a sink node s; (Aghaei [Page 6, last paragraph]: “directed acyclic graph by augmenting it with a single source node s that is connected to the root node (node 1) of the tree and a single sink node t connected to all leaf nodes of the tree” Aghaei teaches converting the decision tree to a direct acyclic graph with a single source and a single sink.)
generating a node set FROM = {o}, wherein a feature indexf= 0; (Aghaei [Page 6, last paragraph]: “directed acyclic graph by augmenting it with a single source node s that is connected to the root node (node 1) of the tree and a single sink node t connected to all leaf nodes of the tree” Aghaei initializes flow at the source and traverses feature tests toward the sink, while implementing the routine current-node set and feature-layer index.)
and connecting the node set FROM to one or more TO nodes by directed arcs, and adding to an arc set A: {ni,f} X {njf+1}, for all i, j, and iff= N-1, TO node = SINK s, and iff< N,f=f+1, else STOP and Return the arc set A; (Aghaei [Page 10, Paragraph 1]: “each datapoint in problem (1) is allotted one unit of flow which it attempts to guide from the source node to the sink node” Aghaei [Page 11, Paragraph 3]: “This path is connected to the sink via an arc of capacity 1 if and only if the datapoint is correctly classified” Aghaei teaches directed arcs from a node to the next level and defines the flow from source to sink. Aghaei [Page 21, Paragraph 2]: “Each sink node tk, k ∈K, is connected to all nodes n” Aghaei teaches that the complete flow graph connects every tree node to every class-specific rank, providing a fully connected bipartite portion of the graph.)
wherein a total number of nodes = Ef L f+2, Arcs = E f(Lf * Lf+1); and wherein total feasible rules = prod{L_f} (Aghaei [Page 9, Remark 1]: “this formulation can also be applied to datasets involving categorical or integer features … for each level of the feature, we create a new binary column” Aghaei creates a binary column for each categorical level and cumulative binary columns for integer values. Since one encoded node is provided for each feature value and there are two additional nodes for the source and sink, the sum is the feature node count plus two. Note that in fully connected layers the condition for the number of arcs holds as a consequence of above.)
Aghaei is in the same field as the present invention, since it is directed to learning optimal binary classification trees. It would have been obvious, before the effective filing date of the claimed invention, to a person of ordinary skill in the art, to combine weighted combination of rules by traversing graphs as taught in Shabtai as modified by Benard as modified by Lumadjeng with categorizing feature values as nodes and cumulatively storing these, as well as directing the flow from source to sink in the graph, as taught in Aghaei. Aghaei provides this additional functionality. As such, it would have been obvious to one of ordinary skill in the art to modify the teachings of Shabtai as modified by Benard with Lumadjeng to include teachings of Aghaei because the combination would allow for the graph to avoid cycles as it is directed and goes from source to sink. This has the potential benefit of increasing the efficiency of the optimization of rules constraints without resulting in cycles.
Regarding dependent claim 3, Shabtai, Benard, and Lumadjeng teach:
The computer program product of claim 2, wherein the program instructions further comprise: generating an incidence matrix aij = 1 if features in rule j are present in sample i; (Lumadjeng [Page 5, Paragraph 3]: “aij in {0, 1} indicates whether rule j in J covers sample i in I or not; i.e., aij = 1 or aij = 0” Lumadjeng defines a_{ij} as a binary variable to indicate whether rule j covers sample i.)
and including in the generation of the hierarchical feature graph, a coverage penalty c, to force each observation to be covered by a classification rule; (Lumadjeng [Page 5, Paragraph 4]: “we take the weighted combination of the rules for covering the samples” Lumadjeng teaches accounting for the rules coverage of the samples. Lumadjeng [Page 5, Equation 3]: Lumadjeng teaches LP that minimizes the classification margin v_i, showing that there is a penalty for lack of coverage of the rule.)
wherein I(i) is a true label class of sample i; (Lumadjeng [Page 5, Equation 3]: Lumadjeng teaches using y_i as the class vector for sample i.)
wherein K is a corresponding proportion of class I(i) in rule j; (Benard [Page 8, Paragraph 2]: “empirical probability that the output Y is of class 1 conditional on whether the query point x belongs to ^H n(P) or not” Benard teaches that each rule outputs an empirical probability that Y a particular class amongst observations in the rule’s hyperrectangle.)
and wherein a Gini index of rule j, gj is measuring impurity and purity of a rule. (Aghaei [Page 2, Last paragraph]: “uses the Gini Index to decide on the splitting” Aghaei teaches using the Gini index to decide on the pure/impure split of a rule.)
The reasons to combine are substantially similar to those of claim 2.
Regarding dependent claim 4, Shabtai, Benard, and Lumadjeng teach:
The computer program product of claim 3, wherein beta=1 corresponds to the linear programming problem and beta=2 corresponds to the quadratic programming problem. (Lumadjeng [Page 21, Paragraph 2]: “With a piecewise-linear loss function (e.g., mean absolute deviation), our linear programming model can also be reformulated for regression problems” Lumadjeng teaches changing the loss function to be a LP problem or a regression/quadratic programming problem as needed.
The reasons to combine are substantially similar to those of claim 2.
Regarding dependent claim 6, Shabtai, Benard, and Lumadjeng teach:
The computer program product of claim 1, wherein the program instructions further comprise factoring in one or more inter-rule and intra-rule constraints in generating the hierarchical feature graph. (Aghaei [Page 1, Abstract]: “flow-based MIO formulation for learning optimal binary classification trees. Our formulation can accommodate side constraints to enable the design of interpretable and fair decision trees.” Aghaei teaches side constraints of the rules which allow interpretable and fair decision trees.)
The reasons to combine are substantially similar to those of claim 2.
Claims 9-11, 13 are rejected on the same grounds under 35 U.S.C. 103 as claims 2-4, 6 as they are substantially similar, respectively. Mutatis mutandis.
Claim 17 is rejected on the same grounds under 35 U.S.C. 103 as claim 6 as they are substantially similar. Mutatis mutandis.
Regarding dependent claim 18, Shabtai, Benard, and Lumadjeng teach:
The computing device of claim 15, wherein the instructions cause the processor to perform an additional act comprising including a purity value threshold for discovering classification rules in the linear program problem or in the quadratic program problem. (Aghaei [Page 2, Last paragraph]: “uses the Gini Index to decide on the splitting” Aghaei teaches using the Gini index to decide on the pure/impure split of a rule.)
The reasons to combine are substantially similar to those of claim 2.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KYU HYUNG HAN whose telephone number is (703) 756-5529. The examiner can normally be reached on MF 9-5.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Alexey Shmatov can be reached on (571) 270-3428. 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.
/Kyu Hyung Han/
Examiner
Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123
Prosecution Insights