DETAILED ACTION
This action is responsive to the communication filed on 09/22/2025. Claims 1-8, 10-15 and 17-20 are pending in the case. Claims 18-20 are new. Claims 1, 10 and 11 are independent claims. Claims 1, 5, 10, 11 and 15 are amended.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
Applicant's arguments filed 09/22/2025 have been fully considered but they are not persuasive.
With respect to the 35 U.S.C. 112(b) rejection of claims 5 and 17:
Applicant does not dispute with any substantive argument the rejection. Applicant amended claim 5 such that it does not recite “uncertain”. Claim 17 is not amended, so the rejection is maintained.
Applicant points out claims 7 and 4 recite uncertainty and is not indefinite. Examiner points out these claims were not rejected under 112(b).
With respect to the 35 U.S.C. 101 rejections:
Applicant argues the claims are similar to example 35 and disclose improvements to functioning of the computer because they are non-conventional and non-generic.
Examiner disagrees, no details about how the model functions is described, only that certain selected data, the recited abstract idea, is used by the model in a generic way. In contrast Example 35 recites specific additional limitations describing the functioning of the computer system in a specific way.
Applicant cites several portions of the specification to suggest manually adding rules increase fraud detection and therefore improve over existing systems as these rules cannot be determined by the model to suggest the non-conventional operation and that this is similar to example 35.
Examiner disagrees. The cited section in fact supports that the functions of the model cannot actually be improved, rather the goal of fraud detection is achieved via manual specification of rules. To the extent the invention is not well understood routine and conventional, the claims do not recite any additional elements that suggest “manually adding rules”, rather the means for adding rules are a result of the abstract idea alone and as such do not make the claims eligible. As previously addressed, the claims are similar to example 35 in so far as they are used to verify identity. Example 35 recites specific computer functional limitations, while the claims do not.
Examiner highlights that the grounds for a claim providing a practical application or significantly more over the recited judicial exception are specific additional elements which amount to significantly more than the abstract idea. Combining both a generic existing model and an abstract idea does not explain the specific of the functioning of the model such that it is an additional element which is non generic or non-conventional and as such significantly more.
With respect to the prior art rejections:
Applicant asserts Sugumaran does not teach “classify the data input into the machine learning model based on a predetermined rule”. Applicant substantiates this by arguing that the decision tree is learned from the data. Therefore, it is not a predetermined rule as the recited predetermined rule is not originally learned.
Examiner disagrees. First nothing in the claim suggests the predetermined rule cannot be created by learning from data generally, only that the rule is a rule that can be determined to be learned based on statistical information as claimed.
The rejection makes clear that first if-then rules are generated by the AVN and SVM, subsequent to that rules are learned by the fuzzy inference engine (i.e the machine learning model) based on statistics. Nothing in the claim suggests that predetermined rules cannot be generated via ANN and SVM methods. The inference engine does not generate the predetermined rules, but rather built/learned from the generated if-then rules. Certain separate other machine learning methods, the AVN and SVM, generate rules which applicant appears to misconstrue as the claimed machine learning model.
Applicant continues noting the predetermined rules are generated outside the machine learning model and that the reference’s rules are generated within the model and also that rules are discarded rather than "adhered to"
Examiner disagrees. As mentioned previously, the rules in the art are generated outside the claimed machine learning model, corresponding to the inference engine. Further, choosing to discard rules implies that those rules that are not discarded are maintained, or “learned” as claimed.
Applicant disputes the other claim rejections for the same reasons which have been addressed already. Further the new claims, 19 and 20, have been addressed in the updated rejection.
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 1-8, 10-15 and 17-20 are rejected under 35 U.S.C. 101 because the claims are directed to an abstract idea without significantly more.
Regarding Claim 1
Under step 1, the claim is directed to A learning system, which is directed to a machine, one of the statutory categories.
Under Step 2A Prong 1, the claim recites the following limitations which are considered mental evaluations “classify the same data input into the machine learning model based on a predetermined rule to obtain a second classification… wherein the predetermined rule is a rule for classifying whether the user is an unauthorized user and the second classification indicates whether the user is classified as unauthorized;… restrict the user's access to the service when either the first classification indicates the user is unauthorized or the second classification indicates the user is unauthorized;
…determine whether the predetermined rule is to be learned based on statistical information relating to data satisfying the predetermined rule;” These limitations are abstract ideas because they are abstract decisions about certain data. Limiting the data or rules to describe certain specific data as claimed does not suggest decisions about the data or rules is not a mental evaluation. Similarly, restricting user access when an indication is met amounts again to a decision made in the mind, particularly because the specification provides no detail about how restriction is performed but that restriction is a consequence of administrative review (para 0038), which is itself a mental process.
Therefore, the claim recites an abstract idea
Furthermore under step 2A Prong 2 and 2B the claims recite the additional element(s) that are mere instructions to implement an abstract idea on a computer, or merely uses a computer or computer machinery as a tool to perform an abstract idea. (“comprising at least one processor… based on a machine learning model;… input data into a machine learning model to obtain a first classification of the data;…and execute training processing of the predetermined rule that is determined to be learned. ”) See MPEP 2106.05(f). Examiner notes that applying a machine learning model and performing training processing describes computer machinery at a high degree of generality and does not proport any improvements to the functioning of such machinery. In addition, the limitations “wherein the data is behavior data relating to behavior of a user on a service,wherein the machine learning model is a machine learning model for classifying whether the user is an unauthorized user and the first classification indicates whether the user is classified as unauthorized;” is generally linking the use of the judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h). Accordingly, the recited additional elements, when taken alone or in combination, do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea, nor do they amount to significantly more than the judicial exception because they do not impose any meaningful limits on practicing the abstract idea.
Regarding Claim 2 , 3, 5, 6 and 12
The claims are directed to a machine. Each of the limitations described in these claims, under Step 2A Prong 1, only serve to describe the abstract idea addressed in the independent claim.
Furthermore under step 2A Prong 2 and 2B, the claim(s) do not recite additional elements to consider other than those considered in the independent claim. Accordingly, the recited additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea, nor do they amount to significantly more than the judicial exception because they do not impose any meaningful limits on practicing the abstract idea.
Regarding Claim 4
The claim is directed to a machine. The claim recites the following limitations “evaluates an uncertainty of the rule in the machine learning model”. Under Step 2A Prong 1, these limitations correspond to a mental step.
Furthermore under step 2A Prong 2 and 2B the claims recite the additional element(s) that are mere instructions to implement an abstract idea on a computer, or merely uses a computer or computer machinery as a tool to perform an abstract idea. (“the at least one processor…. executes the training processing based on an evaluation result.”) See MPEP 2106.05(f). Examiner notes that performing training processing describes computer machinery at a high degree of generality and does not proport any improvements to the functioning of such machinery. Accordingly, the recited additional elements, when taken alone or in combination, do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea, nor do they amount to significantly more than the judicial exception because they do not impose any meaningful limits on practicing the abstract idea.
Regarding Claim 7
The claim is directed to a machine. The claim recites the following limitations “evaluates an uncertainty of a predetermined rule in a machine learning model in which the predetermined rule has been learned by the training processing”. Under Step 2A Prong 1, these limitations correspond to a mental step.
Furthermore under step 2A Prong 2 and 2B the claims recite the additional element(s) that are mere instructions to implement an abstract idea on a computer, or merely uses a computer or computer machinery as a tool to perform an abstract idea. (“and the at least one processor switches to a machine learning model in which a predetermined rule has been learned by the training processing based on the evaluation result”) See MPEP 2106.05(f). Accordingly, the recited additional elements, when taken alone or in combination, do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea, nor do they amount to significantly more than the judicial exception because they do not impose any meaningful limits on practicing the abstract idea.
Regarding Claim 8
The claim is directed to a machine. The claim recites the following limitations “determines whether the predetermined rule, which is designated as a rule to be learned among the plurality of rules, is to be learned based on statistical information relating to data satisfying the predetermined rule designated as a rule to be learned… performs classification relating to the data based on each of a plurality of rules ”. Under Step 2A Prong 1, these limitations correspond to a mental step.
Furthermore, under step 2A Prong 2 the claims recite the additional element(s) “the at least one processor receives a designation of the predetermined rule to be learned among the plurality of rules” that amounts to adding insignificant extra-solution activity to the judicial exception and mere data gathering. See MPEP 2106.05(g). Accordingly, the additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Further, the additional elements, “the at least one processor receives a designation of the predetermined rule to be learned among the plurality of rules” are insignificant extra-solution activities that are considered well-understood, routine, conventional activities, for the following reasons. Examiner notes that the additional element amounts to receiving or transmitting data over a network (MPEP 2106.05(d)(II)(i). According to MPEP 2106.05(d)(II)(i), “The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner”. As such, the insignificant extra-solution activities are considered well-understood, routine, conventional activities. Therefore, the claim is not patent eligible
Regarding Claim 10
The claim is directed to a process, one of the four statutory categories. Each of the limitations in the body of the claim are evaluated similarly as rejected claim 1.
Regarding Claim 11
The claim is directed to an article of manufacture, one of the four statutory categories. Each of the limitations in the body of the claim are evaluated similarly as rejected claim 1.
Regarding Claim 13
The claim is directed to a machine. The claim recites the following limitations “calculate a kurtosis of the data; determine if the kurtosis is equal to or more than a threshold value; wherein if the kurtosis is equal to or more than a threshold value, determine that the distribution of data is the predetermined distribution; and if the kurtosis is less than a threshold value, determine that the distribution of data is not the predetermined distribution.”. Under Step 2A Prong 1, these limitations correspond to a mental step.
Furthermore under step 2A Prong 2 and 2B, the claim(s) do not recite additional elements to consider other than those considered in the independent claim. Accordingly, the recited additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea, nor do they amount to significantly more than the judicial exception because they do not impose any meaningful limits on practicing the abstract idea.
Regarding Claim 14
The claim is directed to a machine. The claim recites limitation which Under Step 2A Prong 1, only serve to describe the abstract idea addressed in the independent claim.
Furthermore, under step 2A Prong 2 the claims recite the additional element(s) “system deletes an old machine learning model from a server and stores the machine learning model in the server” that amounts to adding insignificant extra-solution activity to the judicial exception and mere data gathering. See MPEP 2106.05(g). Accordingly, the additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Further, the additional elements, “system deletes an old machine learning model from a server and stores the machine learning model in the server” are insignificant extra-solution activities that are considered well-understood, routine, conventional activities, for the following reasons. Examiner notes that the additional element amounts to storing and retrieving information in memory (MPEP 2106.05(d)(II)(iv). According to MPEP 2106.05(d)(II)(i), “The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner”. As such, the insignificant extra-solution activities are considered well-understood, routine, conventional activities. Therefore, the claim is not patent eligible
Regarding Claim 15
The claim is directed to a machine. The claim recites the following limitations “the system generates the behavior data when the user accesses a server; wherein the user behavior data comprises a user ID, a user name, an IP address, an access location, and an access time” Under Step 2A Prong 1, these limitations correspond to a mental step.
Furthermore under step 2A Prong 2 and 2B, the claim(s) do not recite additional elements to consider other than those considered in the independent claim. Accordingly, the recited additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea, nor do they amount to significantly more than the judicial exception because they do not impose any meaningful limits on practicing the abstract idea.
Regarding Claim 17
The claim is directed to a machine. The claim recites the following limitations “wherein the system determines a learning coefficient based on the evaluated uncertainty of the predetermined rule; wherein the system lowers the learning coefficient if the predetermined rule is not evaluated as uncertain; and wherein the lower the learning coefficient of the predetermined rule, the lower the influence the predetermined rule will have on training”. Under Step 2A Prong 1, these limitations correspond to a mental step.
Furthermore under step 2A Prong 2 and 2B, the claim(s) do not recite additional elements to consider other than those considered in the independent claim. Accordingly, the recited additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea, nor do they amount to significantly more than the judicial exception because they do not impose any meaningful limits on practicing the abstract idea.
Regarding Claim 18
The claim is directed to a machine. The claim recites the following limitations “remove the predetermined rule from the plurality of rules after the machine learning model learns the predetermined rule such that the predetermined rule will not be used to obtain the second classification for a subsequent second classification” Under Step 2A Prong 1, these limitations correspond to a mental step.
Furthermore under step 2A Prong 2 and 2B, the claim(s) do not recite additional elements to consider other than those considered in the independent claim. Accordingly, the recited additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea, nor do they amount to significantly more than the judicial exception because they do not impose any meaningful limits on practicing the abstract idea.
Regarding Claim 19
The claim is directed to a machine. The claim recites the following limitations “and wherein the predetermined rule is generated in a viewpoint that is not learned by the machine learning model” Under Step 2A Prong 1, these limitations correspond to a mental step.
Furthermore under step 2A Prong 2 and 2B the claims recite the additional element(s) “the data comprises a plurality of features, wherein the machine learning model can learn any of the plurality of features as viewpoints;” is generally linking the use of the judicial exception to a particular technological environment or field of use. Examiner points of the machine learning models by definition are capable of learning viewpoints of features. As such, the limitation does little more than link the judicial exception to the machine learning field. see MPEP 2106.05(h). Accordingly, the recited additional elements, when taken alone or in combination, do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea, nor do they amount to significantly more than the judicial exception because they do not impose any meaningful limits on practicing the abstract idea.
Regarding Claim 20
The claim is directed to a machine. The claim does not recite any additional judicial exceptions beyond those already identified in the parent claim.
Furthermore under step 2A Prong 2 and 2B the claims recite the additional element(s) “wherein the each of the plurality of features comprise an access time zone, an access frequency, an access location, or a username;” is generally linking the use of the judicial exception to a particular technological environment or field of use. see MPEP 2106.05(h). Accordingly, the recited additional elements, when taken alone or in combination, do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea, nor do they amount to significantly more than the judicial exception because they do not impose any meaningful limits on practicing the abstract idea.
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
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
The term “uncertain” in claim 17 is a relative term which renders the claim indefinite. The term is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. For the purpose of compact prosecution, the term in the context of the claim is understood to describe not using the predetermined rule based on the certainty of the rule.
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.
Claim(s) 1-8, 10-13, 15 and 17, 19-20 rejected under 35 U.S.C. § 103 as being unpatentable over Sugumaran et al. “Automatic rule learning using decision tree for fuzzy classifier in fault diagnosis of roller bearing, further in view of Koottayi et al. US PG Pub US 20180288063 A1.
Regarding claim 1
Sugumaran teaches, A learning system comprising at least one processor configured to:
input data into a machine learning model to obtain a first classification of the data; (Section 5.3 “After defining membership functions and generating the ‘if-then’ rules, the next step is to build the fuzzy inference engine. The fuzzy toolbox available in MATLAB 6.5 was used for building fuzzy inference engine. Each rule was taken at a time and using membership functions and fuzzy operators the rules were entered” Section 5 pg 7 “Fuzzy inference provides a precise approach for dealing with uncertainty. Fuzzy inference is a method that interprets the values in the input vector and, based on some set of rules, assigns values to the output vector” the inference engine is the machine learning model which obtains a first classification as a output vector.)classify the same data input into the machine learning model based on a predetermined rule to obtain a second classification of the data; (Section 5.2 pg 9 “ANN and SVM are the techniques used to generate set of ‘if-then’ rules, which forms the knowledge base for the fuzzy classifier [18,19]. In this study, decision tree is used for that purpose. Decision tree shows the relation between features and the condition of the bearing” the decision tree is the predetermined rule set which obtains a classification features into a condition of the bearing based on the decision tree rules.) determine whether the predetermined rule is to be learned based on statistical information relating to data satisfying the predetermined rule; ( Section 4 “A standard tree induced with ID3 (induction decision) consists of a number of branches, one root, a number of nodes and a number of leaves… The occurrence of an attribute in a tree provides the information about the importance of the associated attribute… At each decision node in the decision tree, one can select the most useful feature for classification using appropriate estimation criteria. The criterion used to identify the best feature invokes the concepts of entropy reduction and information gain” the tree which is used to build/learn the fuzzy inference engine is based on statistics of the data satisfying a rule. For example, when the Kurtosis statistic of the data is greater than 13.47 the bearing is classified as IORF. Which is a rule learned by the inference engine.
PNG
media_image1.png
373
778
media_image1.png
Greyscale
)
and execute training processing of the predetermined rule that is determined to be learned (Section 5 “Rules are the inputs for building a fuzzy inference engine. All rules are evaluated in parallel, and the order of the rules is unimportant.” The rules learned or used to build the inference engine amounts to executing training processing.)
Sugumaran does not teach, wherein the data is behavior data relating to behavior of a user, the machine learning model is a machine learning model for classifying whether the user is an unauthorized user, and the predetermined rule is a rule for classifying whether the user is an unauthorized user. … wherein the predetermined rule is a rule for classifying whether the user is an unauthorized user and the second classification indicates whether the user is classified as unauthorized… restrict the user’s access to the service when either the first classification indicates the user is unauthorized or the second classification indicates the user is unauthorized
Koottayi when addressing combining rule mining with neural network teaches, wherein the data is behavior data relating to behavior of a user, (para 0095 “After a period of time, the system may learn and adapt from the default or static policies and the behavior models, and determine that based on user behavior pattern, the system should not trigger an anomalous activity” the model learns models based on user behavior patterns or data.) the machine learning model is a machine learning model for classifying whether the user is an unauthorized user, and the rule is a rule for classifying whether the user is an unauthorized user. (para 0088 “With the system of FIG. 1 deployed, the access management and threat detection system… can make informed decisions based on: (i) rules within static and dynamic enforcement policies, and (ii) behavior models, about whether an end user is authenticated and/or authorized to access resources on a target system” para 0101 “As shown in FIG. 3, a threat detection component (e.g., threat detection component 165) may comprise a rules engine 305 having a machine learning component 310” the machine learning component as well as the rules classifying the user as authorized or not.) wherein the predetermined rule is a rule for classifying whether the user is an unauthorized user and the second classification indicates whether the user is classified as unauthorized… restrict the user’s access to the service when either the first classification indicates the user is unauthorized or the second classification indicates the user is unauthorized (para 0062 “For example, certain embodiments may enforce compliance or block [restrict] users to unsanctioned applications, perform adaptive authorization, challenge users based on policy and behavior models” i.e challenging based on policy or predetermined rule, necessarily involves a determination based on policy, ie classification, that the user should be challenged, ie unauthorized)
Accordingly, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify data rule selection of Sugumaran to operate on data related to intrusion detection as disclosed by Koottayi. One would have been motivated to make such a combination because as noted by Koottayi “Manual or home-grown analysis with user interfaces is cost prohibitive as there is an extremely high volume of user activity (e.g., billions of events per day). Moreover, it is difficult to have automated pattern detection for a large number of users that would be highly accurate and thereby be capable of preventing those unauthorized users from gaining access to services… To address these problems, various embodiments provide techniques… for analyzing security events using dynamic policies and behavior models” (Koottayi para 0060-0061)
Regarding claim 2
Sugumaran/Koottayi teaches claim 1
Sugumaran further teaches, wherein the statistical information indicates a number of data satisfying the predetermined rule, if the number of data satisfying predetermined the rule is less than a threshold value, the at least one processor does not determine that the predetermined rule is to be learned (Section 4.2 “Input to the algorithm is the set of features [data] described in Section 3;… The first number in the parenthesis indicates the number of data points that can be classified using that feature set. The second number indicates the number of samples against this action. If the first number is very small compared to the total number of samples, then the corresponding features can be considered as outliers and hence ignored.” i.e less than a threshold value the rule is not learned as the feature is ignored.) if the number of data satisfying the predetermined rule is equal to or more than the threshold value, the at least one processor determines that the predetermined rule is to be learned, and wherein the statistical information is one of a skewness, moment, or kurtosis of the data. (Section 4.2 “Input to the algorithm is the set of features [data] described in Section 3;… The first number in the parenthesis indicates the number of data points that can be classified using that feature set. The second number indicates the number of samples against this action. If the first number is very small compared to the total number of samples, then the corresponding features can be considered as outliers and hence ignored…Features that have less discriminating capability can be consciously discarded by deciding on the threshold…The features that dominate generally represent the bearing condition descriptors. Referring to Fig. 4, one can identify three such most dominant features, (a) minimum value (b) standard error (c) kurtosis.” i.e conversely, the feature being not ignored indicates that the comparison is above or equal to a threshold.)
Regarding claim 3
Sugumaran/Koottayi teaches claim 1
Sugumaran further teaches, wherein the statistical information indicates a distribution of data satisfying the predetermined rule, if the distribution of the data satisfying the predetermined rule is not a predetermined distribution, the at least one processor does not determine that the predetermined rule is to be learned, and if the distribution of the data satisfying the predetermined rule is the predetermined distribution, the at least one processor determines that the predetermined rule is to be learned. (Section 4.2 “Input to the algorithm is the set of features [data] described in Section 3;…The first number in the parenthesis indicates the number of data points that can be classified using that feature set. The second number indicates the number of samples against this action. If the first number is very small compared to the total number of samples, then the corresponding features can be considered as outliers and hence ignored…Features that have less discriminating capability can be consciously discarded by deciding on the threshold” the number of samples which satisfy a deciding threshold defines a distribution of data. Furthermore, the decision tree defines a membership function for determining which rules are learned. Section 5.1 “A membership function (MF) is a curve that defines how each point in the input space is mapped to a membership value (or degree of membership) between 0 and 1. The decision tree for the selected three features is shown in Fig. 5”)
Regarding claim 4
Sugumaran/Koottayi teaches claim 1
Sugumaran further teaches, wherein the at least one processor evaluates an uncertainty of the predetermined rule in the machine learning model, and the at least one processor executes the training processing based on an evaluation result. (Section 4.2 “The first number in the parenthesis indicates the number of data points that can be classified using that feature set. The second number indicates the number of samples against this action. If the first number is very small compared to the total number of samples, then the corresponding features can be considered as outliers and hence ignored. The other features appear in the nodes of decision tree in descending order of importance” the evaluation of the outliers and the uncertainty of the data correspond to the uncertainty of the rules defined by the decision tree. As previously noted, the training process is based on the decision tree constructed.)
Regarding claim 5
Sugumaran/Koottayi teaches claim 4
Sugumaran further teaches, wherein the at least one processor excludes the predetermined rule as so that the predetermined rule is not used based on the evaluation result. (Section 4.2 “The first number in the parenthesis indicates the number of data points that can be classified using that feature set. The second number indicates the number of samples against this action. If the first number is very small compared to the total number of samples, then the corresponding features can be considered as outliers and hence ignored. The other features appear in the nodes of decision tree in descending order of importance” the resulting decision tree is an evaluation result of the certainty of the rules. Rules of high importance are certain at higher levels of the tree and thus not used at lower levels of the tree.)
Regarding claim 6
Sugumaran/Koottayi teaches claim 1
Sugumaran further teaches, wherein the at least one processor excludes the predetermined rule learned by the training processing so that the predetermined rule is not used. (Section 4.2 “The first number in the parenthesis indicates the number of data points that can be classified using that feature set. The second number indicates the number of samples against this action. If the first number is very small compared to the total number of samples, then the corresponding features can be considered as outliers and hence ignored.”)
Regarding claim 7
Sugumaran/Koottayi teaches claim 1
Sugumaran further teaches, wherein the at least one processor evaluates an uncertainty of the predetermined rule in a machine learning model in which the predetermined rule has been learned by the training processing; (Section 5.2 “In this study, decision tree is used for that purpose. Decision tree shows the relation between features and the condition of the bearing. Tracing a branch from the root node leads to a condition of the bearing” Section 5.3 “A number in the parenthesis at the end of each rule indicates the relative importance of the rule in ‘0–1’ scale.” the decision tree machine learning model which learns a set of rules and their corresponding certainty captured by the rules scale value)
and the at least one processor switches to a machine learning model in which the predetermined rule has been learned by the training processing based on the evaluation result. (Section 5.3 “The rules were obtained from a training data set (150 trials in each condition). For testing the built model a portion of the data (100 trials in each condition) called testing data was kept aside. Using the testing data, the fuzzy inference engine was evaluated and its performance was presented as confusion matrix in Fig. 10” the Fuzzy inference model is the model switched to in the evaluation stage, which has learned the rules according to training processing.)
Regarding claim 8
Sugumaran/Koottayi teaches claim 1
Sugumaran further teaches, wherein the at least one processor performs classification relating to the data based on each of a plurality of rules, (See figure 1 pg 2
PNG
media_image2.png
603
557
media_image2.png
Greyscale
the fuzzy system classifies the data as a fuzzy output.)
the at least one processor receives a designation of the predetermined rule to be learned among the plurality of rules, and the at least one processor determines whether the predetermined rule, which is designated as a rule to be learned among the plurality of rules, (as shown in the figure the rules to be learned are a result of the rule generations.) is to be learned based on statistical information relating to data satisfying the predetermined rule designated as a rule to be learned. (Section 4.2 “The algorithm has been applied to the problem under discussion for feature selection… The level of contribution is not same and all eight features are not equally important. The level of contribution by each individual feature is given by a statistical measure within the parenthesis in the decision tree. Fig. 4. The first number in the parenthesis indicates the number of data points that can be classified using that feature set” the rule generation as shown in the flowchart is based on the feature selection which is performed based on the statistical information of the data satisfying the rule.)
Regarding claim 10
Independent Claim 10 is rejected for the reasons set forth in the rejection of claim 1
Regarding claim 11
Independent Claim 11 is rejected for the reasons set forth in the rejection of claim 1
Further,
Sugumaran teaches, A non-transitory computer-readable information storage medium for storing a program for causing a computer to: (Section 5.3 “The fuzzy toolbox available in MATLAB 6.5 was used for building fuzzy inference engine”)
Regarding claim 12
Sugumaran/Koottayi teaches claim 1
Sugumaran further teaches, wherein the predetermined rule has not been learned by the machine learning model (Section 4.2 “The first number in the parenthesis indicates the number of data points that can be classified using that feature set. The second number indicates the number of samples against this action. If the first number is very small compared to the total number of samples, then the corresponding features can be considered as outliers and hence ignored.”)
Regarding claim 13
Sugumaran/Koottayi teaches claim 3
Sugumaran further teaches, wherein the at least one processor is further configured to: calculate a kurtosis of the data; (Section 3 pg 4-5 “We selected a fairly wide set of these parameters as the basis for our study. They are mean, standard error, median, standard deviation, sample variance, kurtosis, skewness, range, minimum, maximum, and sum. These features were extracted from vibration signals. The statistical features are explained below…. Kurtosis indicates the flatness or the spikiness of the signal. Its value is very low for good bearings and high for faulty bearings due to the spiky nature of the signal” Section 5.2 “Rule 3. if (min is minmin) and (SE is se3) and (Kur is kur1) then (condition is IRF) (0.9) where, kur—kurtosis” this rule decision is based on the calculation of kurtosis.) determine if the kurtosis is equal to or more than a threshold value; wherein if the kurtosis is equal to or more than a threshold value, determine that the distribution of data is the predetermined distribution; and if the kurtosis is less than a threshold value, determine that the distribution of data is not the predetermined distribution. (Section 5.1 pg 7 “In the present study, trapezoidal membership function is used. The selection of this membership function is to some extent arbitrary. However, the following points were considered while selecting membership function. Referring to Figs. 4 and 7, as indicated by decision tree, if standard error is less than 0.253 then the condition is good; otherwise ORF. The threshold value (0.253) is defined based on the representative training dataset.” And figure 5
PNG
media_image3.png
402
751
media_image3.png
Greyscale
” as described the decision tree determines membership of statistical data to a defined membership function. The std_error is used as an example, however the tree also defined a membership distribution for kurtosis. Data is classified as belonging do a predetermined distribution based on the threshold value.)
Regarding claim 15
Sugumaran/Koottayi teaches claim 9
Koottayi further teaches, system generates the behavior data when the user accesses a server; wherein the user behavior data comprises a user ID, a user name, an IP address, an access location, and an access time. (para 0100 “the information associated with an event may be collected by the collector 165 from one or more agents including the one or more agents 145, one or more proxies 150, one or more access managers 155, and/or one or more Webgates 160, and include, for example, the client IP address from where the request originated, device information, user information, resource being requested, time of request, and so on…In certain embodiments, the information is additionally collected from third party agents including GPS applications within the client device… the collector 165 may be configured to organize the information related to an access request received from a user into various categories such as the client context 205 (user-agent identifier, IP address, host name, GPS location, etc.)”
Regarding claim 17
Sugumaran/Koottayi teaches claim 4
Further Sugumaran teaches, wherein the system determines a learning coefficient based on the evaluated uncertainty of the predetermined rule (Section 5.3 “A number in the parenthesis at the end of each rule indicates the relative importance of the rule in ‘0–1’ scale. Referring to Fig. 4, the relative importance of each rule was assigned. The shortest leaf node from the root (The leaf node which appears at the top, just below the root node) forms strongest rule. Hence, the scale value of 1 for rules 1 and 2. The next leaf node (one level down) forms the basis for next set of rules which are next best rule. Hence scale value of (0.9) and so on. The same idea is extended” The importance is a measure of the certainty of the features on the rules. The importance is the learning coefficient.) wherein the system lowers the learning coefficient if the predetermined rule is not evaluated as uncertain; and wherein the lower the learning coefficient of the predetermined rule, the lower the influence the predetermined rule will have on training. ( Section 5.3 “The number of data points which supports/is against the rule is given in leaf node. It acts as a key factor in determining the scale value” Section 4.2 “If the first number is very small compared to the total number of samples, then the corresponding features can be considered as outliers and hence ignored” a lower scale value/learning coefficient is a result of a rule not being uncertain, while a higher scale value has a higher influence on the trained inference engine. as previously noted, those which have a low amount of data which supports the rule are removed entirely from the training set.)
Regarding claim 19
Sugumaran/Koottayi teaches claim 1
Further Sugumaran teaches, wherein the data comprises a plurality of features, wherein the machine learning model can learn any of the plurality of features as viewpoints; and wherein the predetermined rule is generated in a viewpoint that is not learned by the machine learning model ( Section 5 “The rules were obtained from a training data set (150 trials in each condition). For testing the built model a portion of the data (100 trials in each condition) called testing data was kept aside. Using the testing data, the fuzzy inference engine was evaluated and its performance was presented as confusion matrix in Fig. 10. The diagonal elements in the confusion matrix (Fig. 10)
PNG
media_image4.png
243
410
media_image4.png
Greyscale
” the diagonal elements represented features which are learned as viewpoints, the rules that are not learned are embodied by the incorrectly classified of diagonal elements.)
Regarding claim 20
Sugumaran/Koottayi teaches claim 19
Koottayi further teaches, wherein the each of the plurality of features comprise an access time zone, an access frequency, an access location, or a username (para 0100 “the information associated with an event may be collected by the collector 165 from one or more agents including the one or more agents 145, one or more proxies 150, one or more access managers 155, and/or one or more Webgates 160, and include, for example, the client IP address from where the request originated, device information, user information, resource being requested, time of request, and so on…In certain embodiments, the information is additionally collected from third party agents including GPS applications within the client device… the collector 165 may be configured to organize the information related to an access request received from a user into various categories such as the client context 205 (user-agent identifier, IP address, host name, GPS location, etc.)”
Claim(s) 14 is rejected under 35 U.S.C. § 103 as being unpatentable over Sugumaran/Koottayi, further in view of Yang et al. US PG Pub 20190311246 A1.
Regarding claim 14
Sugumaran/Koottayi teaches claim 7
Sugumaran/Koottayi does not teach, wherein the system deletes an old machine learning model from a server and stores the machine learning model in the server.
Yang when addressing storage of machine learning models teaches, wherein the system deletes an old machine learning model from a server and stores the machine learning model in the server. (para. 0035 “Anytime a new AI model is generated by any of the processing devices on the network and subsequently verified by the network, the chain may be updated to add a new block that includes the new performance value associated with the new AI model… For example, because ultimately only the best AI model is of interest, as more new verified AI models are generated, the network may detach (or remove) some old AI models and keep only more recent AI models”)
Accordingly, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to machine learning system of Sugumaran to store and update machine learning models as described by Yang. One would have been motivated to make such a combination because as noted by Yang, selection of the higher performing model results in the best or better performance “A higher performance may refer to a higher recognition rate or a lower error (or loss) rate. In some scenarios, the processing device may be configured to mine the best (or better) AI model that may result in the best (or better) performance.” (Yang para 0020)
Claim(s) 18 is rejected under 35 U.S.C. § 103 as being unpatentable over Sugumaran/Koottayi, further in view of Alvi “Turning a Blind Eye: Explicit Removal of Biases and Variation from Deep Neural Network Embeddings
Regarding claim 18
Sugumaran/Koottayi teaches claim 8
Sugumaran/Koottayi does not explicitly teach, to remove the predetermined rule from the plurality of rules after the machine learning model learns the predetermined rule such that the predetermined rule will not be used to obtain the second classification for a subsequent second classification
Alvi when addressing rule or bias removal teaches, to remove the predetermined rule from the plurality of rules after the machine learning model learns the predetermined rule such that the predetermined rule will not be used to obtain the second classification for a subsequent second classification (pg 9-10 “T-SNE visualizations of these feature embeddings are shown in figure 5. The feature representation of the baseline network that…was trained to classify age is clearly separable by gender, demonstrating that the bias in the training data was learned. After unlearning gender, the feature representation is no longer separable by gender, demonstrating that this bias has been removed”
PNG
media_image5.png
231
524
media_image5.png
Greyscale
first a predetermined bias or rule is learned by the model shown in figure 5(a) after it is learned it is removed or “unlearned” such that the predetermined rule will not be used to obtain the second classification for a subsequent second classification.)
Accordingly, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to machine learning system of Sugumaran/Koottayi to remove bias or unlearn rules of the machine learning system as described by Alvi. One would have been motivated to make such a combination because as noted by Alvi “our algorithm allows networks trained on biased data to generalize better to unbiased settings, by removing each known bias from the feature representation of the network. This is a significant step towards trusting that a network definitely isn’t basing its decisions on the wrong reasons… it is of great importance (Alvi pg 14)
Conclusion
Prior art:
Ribeiro “An Association Rule-Based Method to Support Medical Image Diagnosis With Efficiency” describes using features including statistical features like Kurtosis to determine an association rule set
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JOHNATHAN R GERMICK whose telephone number is (571)272-8363. The examiner can normally be reached M-F 7:30-4:30.
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, Kakali Chaki can be reached on 571-272-3719. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/J.R.G./
Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122