Detailed Action
This action is in response to the amendment filed on 01/29/2026 for application 17/454,086, in which:
Claims 1 and 15 are independent claims.
Claims 2, 12, 16 and 20 are cancelled.
Claims 23 and 24 are newly added.
Claims 1, 6-7, 11, 15, and 18, are currently amended.
Claims 1, 3-11, 13-15, 17-19, and 21-24 are currently pending.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 01/29/2026 and 05/04/2026 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
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 01/29/2026 have been fully considered but they are not persuasive.
Regarding the 35 U.S.C. § 101 Rejections:
Applicant's arguments regarding the 35 U.S.C. § 101 rejections of the previous office action have been fully considered, but are unpersuasive.
Applicant asserts (Page 9), that the amended claims satisfy the Alice/Mayo § 101 eligibility analysis. Thus, the claims recite a specific technological improvement to provide for training an anomaly prediction model for improved accuracy in detecting potential issues in real-time systems, such as double sided grinders used in wafer processing systems.
Examiner respectfully disagrees. The rejection follows the steps of the analysis as laid out in the MPEP which was followed for the previous and current examination (see MPEP 2106). Thus, the office action does not fail to establish a proper and well-supported prima facie case as the claims are explained to be not patentable via the Patent Subject Matter Eligibility steps within MPEP 2106. The claims do not integrate the judicial exception into a practical application nor amount to significantly more. The claim is not patent eligible. Although the Claims are interpreted in light of the specification, limitations from the specification are not read into the Claims.
MPEP 2106.05(a) recites:
After the examiner has consulted the specification and determined that the disclosed invention improves technology, the claim must be evaluated to ensure the claim itself reflects the disclosed improvement in technology … the claim must include the components or steps of the invention that provide the improvement described in the specification
…
It is important to note, the judicial exception alone cannot provide the improvement. The improvement can be provided by one or more additional elements. See the discussion of Diamond v. Diehr, 450 U.S. 175, 187 and 191-92, 209 USPQ 1, 10 (1981)) in subsection II, below.
Applicant fails to show how any alleged technical improvement would be provided by anything more than the judicial exception on its own. Additionally, applicant fails to show how the claim includes components or steps that would provide the alleged improvement described in the specification or by the cited case law. By MPEP 2106.05(f)(1), "the claim recites only the idea of a solution or outcome, i.e. the claim fails to recite details of how a solution to a problem is accomplished". Moreover, the examiner maintains that the Claim does not impose any meaningful limits on the judicial exceptions. As noted in the rejection, the Claim does not include additional elements that are sufficient to amount to an integration of the identified abstract idea into a practical application, thus the claim is directed to an abstract idea.
Applicant asserts (Pages 9-14),that similar to Carmody and Desjardins the present claims recite specific improvements related to machine learning. The pending claims recite prediction models that are specifically trained to recognize patterns and to be able to apply those patterns when executing models to improve efficiency of that process, and to reduce processing resources while improving accuracy. Furthermore, the present claims describe using the system in real-time. The claims recite additional elements that integrate the judicial exceptions into a practical application and directed to improvement to technology/technical field. The claimed invention at issue in Ex Parte Desjardins related to training machine learning models. More specifically, as described in the Specification, the invention related to an improvement in efficiency of machine-learning models through training and training. Amended claim 1 describes training an anomaly prediction model with a first plurality of training datasets and a second plurality of training datasets consisting of sensor readings for a first duration of time, where the first plurality of training datasets includes anomaly data and the second plurality of training datasets includes non-anomaly data. This two-dataset methodology provides for an improvement with regards to the operation and accuracy of the anomaly prediction model. Furthermore, the calibration of the datasets allows for more accuracy in determining the lead-up to potential anomalies and detect them. The accompanying Specification of the application as filed supports such improvement in at least paragraphs [0020]-[0027]. Thus, the amended claims are patent eligible under 101.
Examiner respectfully disagrees. For the reasons given below and in the 35 U.S.C. § 101 rejections, the claims are directed to an abstract idea (Step 2A Prong 1) and do not integrate the abstract ideas into a practical application (Step 2A Prong 2). The pending claims recite abstract ideas that fall in at least one of the permissible groups, and noted within the office action below in more details. The independent claims fail to recite the steps that achieve the improvement. The independent claims are no more detailed than to train a prediction model based off datasets including anomalous and non-anomalous data to generate prediction results by comparing random probability distribution curves to associated time slide windows to analyze the data within the respective dataset with no steps on how to achieve an improvement by the determination as there is no particular solution to a particular problem; thus, the Claims are not a technical solution to a technical problem. The rejections have been updated below, the rejection to all Claims (including Claim 1, analogous independent Claims, and all dependent Claims) are maintained and updated as necessitated by Claim amendments.
Applicant asserts (Pages 15-16), that the technical improvements within the pending claim recitations disclose a process of adaptively training a model in a manner based on historical data with leadup data to ensure that the time windows being compared are similar. This allows for real-time analysis. Thus, the clear technical improvements of increased efficiency and accuracy of the machine-learning (ML) model is described in claim 1, like in Enfish. Thus, the i) overall technical field of software inventions and ii) claimed technological improvements in the present application and claims are both very similar to Enfish.
Examiner respectfully disagrees. As noted above, the amended independent claims are no more detailed than to train a prediction model based off datasets including anomalous and non-anomalous data to generate prediction results by comparing random probability distribution curves to associated time slide windows to analyze the data within the respective dataset with no steps on how to achieve an improvement by the determination as there is no particular solution to a particular problem; thus, the Claims are not a technical solution to a technical problem. For the reasons given below, noted above, and in the 35 U.S.C. § 101 rejections, the claims are directed to an abstract idea (Step 2A Prong 1) and do not integrate the abstract ideas into a practical application (Step 2A Prong 2). The pending claims recite abstract ideas that fall in at least one of the permissible groups, and noted within the office action below in more details. The independent claims fail to recite the steps that achieve the alleged improvement.
Applicant asserts (Page 15), that Examiner alleges that certain claim steps are certain methods of organizing human activity (Office Action, p. 4). However, Applicant submits that these steps do not fall into any of the certain methods of organizing human activity as enumerated in the MPEP 2106.04(a)(2)(11).
Examiner respectfully disagrees. The examiner agrees that the amended claim limitations do not include abstract ideas that fall within organizing human activity; however, they fall within the "mental processes" and/or "mathematical concepts" groups of abstract ideas (shown within Step 2A Prong 1). Thus, the recited abstract ideas are “the basic tools of scientific and technological work” (see MPEP 2106.04). The claim is not directed to significantly more than the judicial exception due to the pending Claims not including additional elements that contribute to an “inventive concept”. In terms of the limitations being evaluated individually and in combination, the specific details are discussed below within the Examiner’s Responses and within the 35 USC § 101 Rejections and the rejection follows the steps of the analysis laid out in the MPEP.
Applicant asserts (Pages 15-17), that the present claims are not directed to merely mental processes and/or mathematical concepts, but instead to a specific, computer-implemented technological process that trains and executes various machine-learning models using real-time sensor data, with technical improvements very similar to those in Desjardins and Enfish. The amended recitations of independent claims 1 and 15 constitute a specific implementation of machine learning, not a generic mental processes or mathematical concepts. Examiner alleges that certain claim steps are mental processes, but Applicant submits that these steps cannot practically be performed mentally. Performing real-time dataset analysis using "a plurality of additional real-time datasets" involves processing massive amounts of sensor data points that cannot be mentally processed, compared, or analyzed by a human. A human cannot mentally align and compare a plurality of datasets, with and without anomalies, to determine if real-time sensor data does or does not potentially include anomaly data as recited in the claims. Similarly, continually performing this analysis requires computational processing to organize and perform and the analysis and provide results. The claims do not explicitly recite any mathematical concepts, formulas, or algorithms. At best, the claims recite limitations that may be based on or involve mathematical concepts, but no mathematical concepts are recited in the claims themselves. Applicant points the Examiner to the recent August 4 memo from Charles Kim, reminding Examiners that a limitation that merely involves math is not an abstract idea. Thus, the rejection should be withdrawn. Applicant also notes that the present claims describe improvements to a process of adaptively training a model in a manner based on historical data with the lead-up data to ensure that the time windows being compared are similar. This allows for real-time analysis. Thus, the clear technical improvements of increased efficiency and accuracy of the machine-Iearning (ML) model is described in claim 1, like in Desjardins describing Enfish. Applicant respectfully submits that the present claims are not directed to merely mental processes and/or mathematical concepts, but instead to a specific, computer implemented technological process that trains and executes various machine-learning models using real-time sensor data, with technical improvements very similar to those in Desjardins and Enfish. Thus, taken together, the present claims integrate any alleged judicial exception into a practical application under Prong Two of the Step 2A analysis and Applicant respectfully requests that the Examiner withdraw the rejection
Examiner respectfully disagrees. The claims recites abstract ideas a-h; where the abstract ideas are mathematical relationships between variables using formulas/equations, or evaluations/judgements that can be performed in the human mind (or by a human using pen and paper). The noted features/limitations within the newly amended claims contain additional elements and abstract ideas but the additional elements are unable to integrate the judicial exception. Currently, the three types of additional elements fall within MPEP 2106.05 (f), (g), and (h) for Claim 1 (shown in Step 2A Prong 2 (a-e). The pending claims recite abstract ideas that fall in at least one of the permissible groups, and noted within the office action below in more details. The independent claims fail to recite the steps that achieve the improvement. The limitations are unable to provide improvement as they are currently being evaluated as either abstract idea(s) or additional elements that fall within MPEP 2106.05. The claims are directed towards the improvement of an abstract idea. Improvements to an abstract idea are still considered to an abstract idea. Additionally, the Claims does not reflect any improvement in the functioning of a computer or hardware processor rather the additional elements merely use a generic computer component to perform the abstract idea or restricting the abstract idea to a particular technological environment or insignificant extra solution activities. Therefore, the claims do not integrate the judicial exception into a practical application nor amount to significantly more. The claim is not patent eligible. Although the Claims are interpreted in light of the specification, limitations from the specification are not read into the Claims.
Applicant asserts (Pages 17-18), it is not necessary to consider Step 2B of the Alice/Mayo framework because the claims are eligible under either prong of Step 2A. Nevertheless, Applicant submits that the claims contain an inventive concept that amounts to significantly more than any alleged judicial exception. Accordingly, the claims satisfy the requirements of Step 2B and are patent-eligible subject matter under 35 U.S.C. § 101.
Examiner respectfully disagrees. As noted above, the claims are not eligible under Step 2A Prong 1 and 2; thus, Step 2B needs to be considered. The amended independent claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements recited, alone or in combination, do not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding the 35 U.S.C. § 103 Rejections:
Applicant's arguments regarding the 35 U.S.C. § 103 rejections of the previous office action have been fully considered, but are unpersuasive.
Applicant asserts (Pages 18-20), that Bowers describes a system for predictive analytics. However, as acknowledged at page 31 of the Office Action, Bowers does not describe or suggest generating time slide windows for each dataset, as required by amended claim 1. Li does not rectify the deficiencies of Bowers. Li describes generating a time slide window for each real-time dataset. However, Li does not describe or suggest comparing anomaly detection random probability distribution curves generated using a plurality of historical datasets with time slide window random probability distribution curves for each time slide window generated for each real-time dataset of the plurality of real-time datasets to determine if the corresponding time slide window includes anomaly data, as is required by amended claim 1. Sultani does not rectify the deficiencies of Bowers. Sultani describes having two sets of training data, but does not describe or suggest comparing anomaly detection random probability distribution curves generated using a plurality of historical datasets with time slide window random probability distribution curves for each time slide window generated for each real-time dataset of the plurality of real-time datasets to determine if the corresponding time slide window includes anomaly data, as is required by amended claim 1. Thus, no combination of Bowers, Li, and Sultani describes or suggests a computer device as is recited in amended claim 1. More specifically, no combination of Bowers, Li, and Sultani describes or suggests a computer device programmed to: compare anomaly detection random probability distribution curves generated using a plurality of historical datasets with time slide window random probability distribution curves for each time slide window generated for each real-time dataset of the plurality of real-time datasets to determine if the corresponding time slide window includes anomaly data, as is required by amended claim 1. Accordingly, Applicant submits that claim 1 is patentable over Bowers in view of Li and further in view Sultani. Applicant notes that amended independent claims (and dependent claims due to dependency) are patentable over any combination from Bowers/Li/Sultani. The Applicant respectfully requests that the present rejections, at least for the reasons set forth above, that the Section 103 rejections should be withdrawn. Applicant does not concede the propriety of any of the rejections of the dependent claims, but in view of the remarks above, it is unnecessary to discuss each dependent claim rejection
Examiner respectfully disagrees. Applicant’s arguments with respect to the amended independent claims where the newly amended limitation recites generating a time slide window random probability distribution curve for each time slide window has been considered but is moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument. Applicant’s arguments regarding the other independent and dependent claims rely upon the same assertions as with respect to Claim 1, and are thus likewise unpersuasive. Therefore, for the reasons given above and in the rejections below, the rejection to all Claims (including Claim 1, similar independent claims, and all dependent Claims) are maintained. More specific details are discussed below within the 35 USC § 103 Rejections.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1, 3-11, 13-15, 17-19, and 21-24 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding Claim 1:
Subject Matter Eligibility Analysis Step 1:
Claim 1 recites a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 1 further recites the device comprising of:
calibrate the plurality of real-time datasets so that each real-time dataset describes sensor readings for the first duration of time (a mathematical relationship between variables and/or numbers using a mathematical formula/equations)
generate a time slide window for each real-time dataset of the plurality of real-time datasets by combining a real-time dataset with a plurality of additional real-time datasets of the plurality of real-time datasets that occur prior to the real-time dataset for a second duration of time (a mathematical relationship between variables and/or numbers using a mathematical formula/equations)
… determine if the real-time dataset includes anomaly data … (a human being can mentally apply evaluation to make a determination that a specific dataset includes anomaly data)
generating one or more anomaly detection random probability distribution curve using a plurality of historical datasets (a mathematical relationship between variables and/or numbers using a mathematical formula/equations)
generating a time slide window random probability distribution curve for each time slide window generated for each real-time dataset of the plurality of real-time datasets (a mathematical relationship between variables and/or numbers using a mathematical formula/equations)
comparing the one or more anomaly detection random probability distribution curves to each of the plurality of time slide window random probability distribution curves to determine if the corresponding time slide window includes anomaly data (a human being can mentally apply evaluation to compare a specific distribution curve to each time slide window to make a judgement/determination if the time slide window includes anomaly data)
generating prediction results based on the comparison (a human being can mentally apply evaluation to generate prediction results based on the comparison)
analyzing the plurality of additional real-time datasets of the plurality of real-time datasets if the real-time dataset includes anomaly data (a human being can mentally apply evaluation to analyze the datasets for the inclusion of anomaly data)
Claim 1 thus recites an abstract idea (that falls into the “mental processes” and/or “mathematical concepts” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the additional elements recited consists of:
at least one processor in communication with at least one memory device, wherein the at least one processor is programmed to (to perform a mental process and the performance of an abstract idea on a computer is no more than instructions to “apply it” on a computer, by MPEP 2106.05(f))
train an anomaly prediction model with a first plurality of training datasets and a second plurality of training datasets consisting of sensor readings for a first duration of time (to perform a mental process and the performance of an abstract idea on a computer is no more than instructions to “apply it” on a computer, by MPEP 2106.05(f))
wherein the first plurality of training datasets includes anomaly data and the second plurality of training datasets includes non-anomaly data (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h))
receive a plurality of real-time datasets from one or more sensors associated with a tool to be analyzed (which is insignificant extra-solution activity of data gathering, by MPEP 2106.05(g))
execute the anomaly prediction model on each real-time dataset of the plurality of real-time datasets … by performing the steps of: (to perform a mental process and the performance of an abstract idea on a computer is no more than instructions to “apply it” on a computer, by MPEP 2106.05(f))
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements recited, alone or in combination, do not provide significantly more than the abstract idea itself. Additional elements a, b, and e are merely applying the abstract idea on a computer (MPEP 2106.05(f)) which cannot provide significantly more. Additional element c is only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which cannot provide significantly more. Additional element d falls within MPEP 2106.05(d) as well-understood, routine and conventional activities of receiving or transmitting data over a network (MPEP 2106.05(d)(II): buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014)). Thus, the claim is subject-matter ineligible.
Regarding Claim 3:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 3 recites the device of Claim 1. Claim 1 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 3 further recites the device to compare the determination if the real time dataset includes anomaly data to the determination if the corresponding time slide window includes anomaly data (a human being can mentally apply evaluation to compare the determined dataset and corresponding time slide window to review the inclusion of anomalous data). Claim 3 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This claim does not recite any further additional elements. Therefore, the claim as a whole is rejected for the same reasons set forth in Claim 1. The new judicial exception recited within Claim 3 is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 4:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 4 recites the device of Claim 3. Claim 3 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 4 further recites the device to generate prediction results based on the comparison of the two determinations (a human being can mentally apply evaluation to generate prediction results based on the comparison of two determinations). Claim 4 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This claim does not recite any further additional elements. Therefore, the claim as a whole is rejected for the same reasons set forth in Claim 1. The new judicial exception recited within Claim 4 is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 5:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 5 recites the device of Claim 1. Claim 1 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 1’s rejection is incorporated within this rejection. Claim 5 does not recite any additional abstract ideas and only inherits the abstract ideas from Claim 1. Claim 5 thus recites an abstract (that falls into the “mental processes” and/or “mathematical concepts” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the new sole additional element recited consists of train the anomaly prediction model with supervised learning (to perform a mental process and the performance of an abstract idea on a computer is no more than instructions to “apply it” on a computer, by MPEP 2106.05(f)).
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the new sole additional element recited, alone or in combination, does not provide significantly more than the abstract idea itself. The additional element is merely applying the abstract idea on a computer (MPEP 2106.05(f)) which cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claim 6:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 6 recites the device of Claim 5. Claim 5 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 6 further recites the device to generate the plurality of training datasets by removing datasets with anomalies from the plurality of historical datasets where the anomaly is not associated with the tool being measured and by removing noisy (a human being can mentally apply evaluation to generate training datasets by removing specific datasets with specific constraints). Claim 6 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This claim does not recite any further additional elements. Therefore, the claim as a whole is rejected for the same reasons set forth in Claim 5. The new judicial exception recited within Claim 6 is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 7:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 7 recites the device of Claim 6. Claim 6 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 7 further recites the device comprising of:
classify the plurality of raw datasets as either non-anomaly or anomaly (a human being can make a judgement to classify data based on an attribute such as anomalous data detection)
for each anomaly dataset, determine if an observed anomaly is associated with a tool being observed and another source (a human being can mentally apply evaluation to determine if an anomaly is associated with a tool being observed and another source)
if the observed anomaly is associated with another source, remove the corresponding anomaly dataset (a human being can make a judgement to remove data based on a correspondence)
align the remaining plurality of anomaly datasets to match a same time period (a human being can make a judgement to align anomaly datasets to match a specific time period)
clean any noisy datasets (a human being can make a judgement to clean noisy datasets)
generate the second plurality of training datasets using a plurality of non-anomaly datasets (a human being can mentally apply evaluation to generate training datasets using non-anomalous datasets)
generate the first plurality of training datasets using the remaining plurality of anomaly datasets (a human being can mentally apply evaluation to generate training datasets using the remaining anomalous datasets)
Claim 7 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the new sole additional element recited consists of extract a plurality of raw datasets (which is insignificant extra-solution activity of data gathering, by MPEP 2106.05(g)).
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the new sole additional element recited, alone or in combination, does not provide significantly more than the abstract idea itself. The additional element falls within MPEP 2106.05(d) as well-understood, routine and conventional activities of receiving or transmitting data over a network (MPEP 2106.05(d)(II): buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014)). Thus, the claim is subject-matter ineligible.
Regarding Claim 8:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 8 recites the device of Claim 7. Claim 7 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 8 further recites the device to perform data clustering on the remaining plurality of anomaly datasets to determine one or more relationships with the remaining plurality of anomaly (a human being can mentally apply evaluation to perform data clustering on specific anomaly datasets to make a mental evaluation of one or more relationships with the anomaly). Claim 8 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This claim does not recite any further additional elements. Therefore, the claim as a whole is rejected for the same reasons set forth in Claim 7. The new judicial exception recited within Claim 8 is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not recite any new additional elements that are sufficient to amount to significantly more than the judicial exception. Thus, the claim is subject-matter ineligible.
Regarding Claim 9:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 9 recites the device of Claim 1. Claim 1 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 9 further recites the device to:
align the plurality of real-time datasets (a human being can make a judgement to align real-time datasets)
adjust an amount of time in each of the real-time datasets to be equal to a predetermined amount of time (a human being can make a judgement to adjust an amount of time to be within each real-time dataset)
Claim 9 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 10:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 10 recites the device of Claim 9. Claim 9 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 10 further recites the device to adjust each real-time dataset to include a predetermined amount of time (a human being can make a judgement to adjust each real-time dataset to include a predetermined amount of time). Claim 10 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This claim does not recite any further additional elements. Therefore, the claim as a whole is rejected for the same reasons set forth in Claim 9. The new judicial exception recited within Claim 10 is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 11:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 11 recites the device of Claim 9. Claim 9 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 11 further recites the device to adjust each real-time dataset to include a predetermined number of data points from the one or more sensors (a human being can make a judgement to adjust specific datasets to include a predetermined number of data points). Claim 11 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This claim does not recite any further additional elements. Therefore, the claim as a whole is rejected for the same reasons set forth in Claim 9. The new judicial exception recited within Claim 11 is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 13:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 13 recites the device of Claim 1. Claim 1 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 13 further recites the device to indicate a potential future issue with the tool to be analyzed (a human being can mentally apply evaluation and make a judgement to indicate a future issue with a specific tool). Claim 13 thus recites an abstract (that falls into the “mental processes” and/or “mathematical concepts” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This claim does not recite any further additional elements. Therefore, the claim as a whole is rejected for the same reasons set forth in Claim 1. The new judicial exception recited within Claim 13 is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 14:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 14 recites the device of Claim 1. Claim 1 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 14 further recites the device to raise an alarm when a future issue is detected (a human being can mentally apply evaluation and make a judgement to raise an alarm when a specific issue is detected). Claim 14 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This claim does not recite any further additional elements. Therefore, the claim as a whole is rejected for the same reasons set forth in Claim 1. The new judicial exception recited within Claim 14 is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claims 15 and 17-19:
Claims 15 and 17-19 incorporate substantively all the limitations of Claims 1, [3+4], and [5+7], in a method (thus, a process) and further recites no new limitations; thus, Claims 15 and 17-19 are rejected for reasons set forth in the rejections of Claims 1, [3+4], and [5+7], respectively.
Regarding Claim 21:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 21 recites the device of Claim 1. Claim 1 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 21 does not recite any additional abstract ideas and only inherits the abstract ideas from Claim 1. Claim 21 thus recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the new sole additional element recited consists of wherein the first duration of time is one minute and the second duration of time is thirty minutes (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h)).
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the new sole additional element recited, alone or in combination, does not provide significantly more than the abstract idea itself. The additional element is only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claim 22:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 22 recites the device of Claim 1. Claim 1 is a device, thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 22 does not recite any additional abstract ideas and only inherits the abstract ideas from Claim 1. Claim 22 thus recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the new sole additional element recited consists of wherein the tool is a double sided grinder (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h)).
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the new sole additional element recited, alone or in combination, does not provide significantly more than the abstract idea itself. The additional element is only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claims 23-24:
Claims 23-24 incorporate substantively all the limitations of Claims 21-22, in a method (thus, a process) and further recites no new limitations; thus, Claims 23-24 are rejected for reasons set forth in the rejections of Claims 21-22, respectively.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1, 3-5, 9-11, 13-15, 17 and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Bowers et al., US-2017/0351241 A1, in view of Li et al., “No Pane, No Gain: Efficient Evaluation of Sliding-Window Aggregates over Data Streams”, in view of Sultani et al., “Real-world Anomaly Detection in Surveillance Videos”, in view of Maurya et al., “Contrastive Structured Anomaly Detection for Gaussian Graphical Models”.
Regarding Claim 1:
Bowers teaches:
A computer device comprising at least one processor in communication with at least one memory device, wherein the at least one processor is programmed to:
(Bowers, Fig. 1; Fig. 3; Page 7, Column 1, [0068], “Referring initially to FIG. 1, illustrated is a block diagram of an embodiment of a communication system 100. The communication system 100 includes an AOS architect subsystem 110 (including a processor ("PR") and memory ("M")) and an AOS edge subsystem 120 (including a processor ("PR") and memory ("M")). For an exemplary discussion of a processor and memory see FIG. 3”; Page 8, Column 2, “The programs stored in the memories may include program instructions or computer program code that, when executed by an associated processor, enable the respective communication device to perform its intended tasks”)
train an anomaly prediction model with a first plurality of training datasets and a second plurality of training datasets consisting of sensor readings for a first duration of time …
(Bowers, Table 1; Figure 4-6; Page 1, Column 2, [0006], “… model typically employs both a set of data to teach the model the characteristics of the system (the training set), and a separate set of data to verify the accuracy of the model (the test set)”; Page 10, Column 1, [0095], “As data is accumulated from the sensors 506 and other data sources, it is stored within the input database 527. At regular intervals … the input database 527 extracts the data values that have been accumulated between the previous time of execution and the current time of execution”; Page 8, Column 2, “The cause and effect analysis module 415 subscribes through a standardized data interface to data sets from the sensor database 410. At prescribed execution intervals … This process of applying stochastic standards reduces each data set to a known format … within the current sensor data interval”. Table 1 shows the node library for the Cause and Effect Analysis model building within Edge Processing (the examiner interprets the Cause an Effect Analysis model to be an anomaly prediction model which is used to predict anomalies within the predictive analysis). Table 1 also notes the expression ‘Trailing Calculation’ which “(c)alculates the sum or average of node results accumulated over a given time period”; where the given time period defines how much historical data the anomaly prediction model will parse to train the model and thus is considered a first duration of time as a given time period is a duration of time. Figure 4 shows the sensor data which is used to train the model and depicted within Figure 6. Figure 5 shows the Input Database (527) and Models (539) being stored within the computer system. The anomaly prediction model is trained with two separate sets of data for training (training set and testing set (which is used to validate the training with known answers)); thus, a first plurality of datasets and a second plurality of datasets which both are consisting of sensor data. Thus, the system is programmed to train an anomaly prediction model with a first plurality of datasets and a second plurality of datasets consisting of sensor readings for a first duration of time).
receive a plurality of real-time datasets from one or more sensors associated with a tool to be analyzed;
(Bowers, Fig. 1; Fig 4; Fig. 5; Fig. 12; Page 7, Column 1, [0069], “The AOS edge subsystem 120 accepts dynamic inputs in two principal forms, namely, dynamic input updates 130 and sensors 140 … The sensors 140 represents data coming from sensors in real-time or near real-time and processed by the AOS edge subsystem 120 in real or near real-time”; Page 13, Column 2, [0127], “With respect to FIG. 12, system operating conditions 1210 are the sensor and other input values that represent the real-time or near real-time operational conditions of the system … or system component”. Fig. 1 & 5 shows the systems receiving dynamic inputs for updates and sensors in real-time; Fig. 4 shows the Aggregate Database hosting the plurality of real-time datasets from the sensors. Bowers notes the sensors and other input values within the system is for a system or system component that is being analyzed for prediction. The examiner interprets tool as a device to carry out a function).
calibrate the plurality of real-time datasets so that each real-time dataset describes sensor readings for the first duration of time;
(Bowers, Page 9, Column 2, [0092], “A data shovel 512 is a component that provides an interface between different types of sensors 506 and formatting of data into an acceptable standard”; Page 12, [0115], “Turning now to FIGS. 8 and 9, illustrated are graphical representations … of processing input data … the stream of data 810 is segmented at regular time intervals for execution. All incoming values between a previous execution time 830 and a current execution time 840 are batched together and then summarized statistically as a data population 850”; Page 14, [0131], “A segment depicts a time history (generally designated 1410) of the system performance metric (see also distribution of results 920 of FIG. 9 wherein calculation results are analyzed, stored, and visualized as distributions with associated percentile values)”; Page 4, [0049], “The edge processing provides … an ability to apply data fill techniques to align various data inputs appearing at various rates …”. The real-time datasets are calibrated (interpreted by the examiner as datasets being aligned/adjusted/formatted) through the system by segmenting and fill techniques. The incoming data for the real-time datasets are segmented and batched together from previous -> current execution time as a population/interval/window of data which is interpreted as the current evaluation interval/window; thus, the first duration of time).
… execute the anomaly prediction model on each real-time dataset of the plurality of real-time datasets to determine if the real-time dataset includes anomaly data by performing the steps of:
(Bowers, Fig. 4; Fig. 25; Page 4, Column 2, [0049], “The edge processing provides many advantages including an ability to accept synchronous and asynchronous data, an ability to apply data fill techniques to align various data inputs appearing at various rates, an ability to analyze and trap data at the data input for anomalies and also trap other relevant data occurring at the same or similar time, an ability to accept data and inputs from disparate sources including manually entered data or decisions or status, an ability to accept input data in sub second increments, an ability to process data in a stochastic manner using Monte Carlo processing, an ability to compress data and present it stochastically, an ability to process data and note anomalies during stochastic processing and to present them for output in various ways including simple green-yellow-red alerts, and an ability to process multiple sensor data and combine them algorithmically to generate results that can then be analyzed”. Fig.4 shows the Edge Gateway which handles edge processing and is able to process/analyze the data from the Sensor Data Interface Module to determine if the real-time datasets contain anomalies and trap. This is done on each real-time dataset via the AOS Edge system’s Cause and Effect module which uses models and provides alerts (Fig. 25: depiction of the information for a cause and effect analysis));
generating one or more anomaly detection random probability distribution curves using a plurality of historical datasets;
(Bowers, Fig. 10; Page 12, Column 2, [0118], “During a second stage 1030, a stochastic summary is provided for each sensor. The summary provides a distribution with low bound ("LB," e.g., a minimum), 10th percentile, median (50 percentile), 90th percentile, and high bound ("HB," e.g., a maximum) values employable to produce an asymmetric Gaussian distribution”; Page 4, Column 2, [0049], “The edge processing provides many advantages including an ability to ability to apply data fill techniques to align various data inputs appearing at various rates, an ability to analyze and trap data at the data input for anomalies and also trap other relevant data occurring at the same or similar time … accept input data in sub second increments, an ability to process data in a stochastic manner using Monte Carlo processing, an ability to compress data and present it stochastically, an ability to process data and note anomalies during stochastic processing and to present them for…” The examiner interprets a random probability distribution as any distribution that describes the probabilities of possible outcomes of a random variable which includes a Gaussian distribution. A stochastic summary is also interpreted as a random probability distribution as both describe outcomes governed by randomness/chance. The edge processing also employs Monte Carlo Processing which utilizes random probability distributions to generate random samples which are used to approximate the distribution of a target function; thus, utilizing the random distribution curves for anomaly detection as the edge processing analyzes/traps anomalous data)
…
analyzing the plurality of additional real-time datasets of the plurality of real-time datasets if the real-time dataset includes anomaly data.
(Bowers, Page 4, [0049], “… edge processing provides … an ability to analyze and trap data at the data input for anomalies and also trap other relevant data occurring at the same or similar time …”. The edge processing system, when determining anomaly data within the real-time dataset, is trapping the anomalous data and also trapping other relevant data occurring at the same or similar time (thus, real-time datasets). The evaluation analysis in Bower’s system innately compares all real-time datasets with their associated time slide windows to analyze and trap anomalous and relevant data).
While Bowers teaches the device to receive and calibrate real-time datasets associated with sensors and generation of time slide windows/random distribution curves. Bowers fails to teach specific training datasets within anomalous data/non-anomalous data and generation of time slide windows via specific combination:
… wherein the first plurality of training datasets includes anomaly data and the second plurality of training datasets includes non-anomaly data;
…
generate a time slide window for each real-time dataset of the plurality of real-time datasets by combining a real-time dataset with a plurality of additional real-time datasets of the plurality of real-time datasets that occur prior to the real-time dataset for a second duration of time;
…
generating a time slide window random probability distribution curve for each time slide window generated for each real-time dataset of the plurality of real-time datasets;
comparing the one or more anomaly detection random probability distribution curves to each of the plurality of time slide window random probability distribution curves to determine if the corresponding time slide window includes anomaly data; and generate prediction results based on the comparison;
However, Li does teach:
generate a time slide window for each real-time dataset of the plurality of real-time datasets by combining a real-time dataset with a plurality of additional real-time datasets of the plurality of real-time datasets that occur prior to the real-time dataset for a second duration of time;
(Li, Page 39, Figure 1, Column 1, Paragraph 3, “Query 1: "Find the maximum bid price for the past 4 minutes and update the result every 1 minute."
SELECT max (bid-price)
FROM bids[WATTR timestamp
RANGE 4 minutes
SLIDE 1 minute]
In the query above, we introduce a window specification with three parameters: RANGE specifies the window size, SLIDE specifies how the window moves, and WATTR specifies the windowing attribute on which that the RANGE and SLIDE parameters are defined. The window specification of Query 1 breaks the bid stream into overlapping 4-minute sub-streams that start every minute, with respect to the timestamp attribute. These overlapping substreams are called sliding windows”; Page 41, Figure 2. Li teaches sliding-window aggregates to manage data streams. Query 1 shows an example of a sliding-window aggregation query, where the sliding windows are generated for each real-time dataset (by timestamped attributes (WATTR) of RANGE/SLIDE as noted by Li) where aggregation occurs with using panes (Figure 1 depicts a logical representation of the segmentation of the panes and generating sliding windows). Figure 2 shows the use of panes to evaluate Query 1 where the first duration of time is the pane size which is 1 minute (SLIDE) and the second duration of time is the prior data to be aggregate within the window (RANGE (interval of evaluation)). Thus, Li teaches generating a time slide window for each real-time dataset (where the time slide window size = RANGE = 4 total minutes; where the segmentation/sliding is based on PANE = 1 minute)… by combining a real-time dataset with a plurality of additional real-time (aggregation of the SLIDE size panes within the RANGE)… datasets that occur prior to the real-time dataset for a second duration of time (where the RANGE is considered the second duration of time as it includes 4 minutes of prior data)).
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to utilize the device of Bowers to receive and calibrate real-time datasets with the aggregation of prior datasets and time sliding windows in Li. One having ordinary skill in the art would have been motivated to implement this change before the effective filing date of the claimed invention, as this leads to the data to be managed sequentially, low computation costs, handle multiple sliding windows, and monitored in a timely fashion (Li, Page 44, Column 1, Paragraph 5, “In this paper, we presented a technique called panes, which reduces both the space and computation cost of evaluating sliding-window queries by sub-aggregating and sharing computation. We discussed using panes to exploit data reduction and computation sharing among multiple window-aggregate computation within a single query. We believe that panes can be extended to improve execution of multiple sliding-window queries over the same stream by sharing panes”).
Bowers and Li both do not explicitly disclose:
… wherein the first plurality of training datasets includes anomaly data and the second plurality of training datasets includes non-anomaly data;
…
generating a time slide window random probability distribution curve for each time slide window generated for each real-time dataset of the plurality of real-time datasets;
comparing the one or more anomaly detection random probability distribution curves to each of the plurality of time slide window random probability distribution curves to determine if the corresponding time slide window includes anomaly data; and generate prediction results based on the comparison;
However, Sultani teaches:
… wherein the first plurality of training datasets includes anomaly data and the second plurality of training datasets includes non-anomaly data;
(Sultani, Page 6846, Column 2, Paragraph 1, “We propose a deep learning approach to detect real-world anomalies … We attempt to exploit both normal and anomalous videos …”. The data within Sultani is video data; however, the training is explicitly taught to be done with both normal (includes non-anomaly data) and anomalous data (includes anomaly data) for anomalous activity recognition).
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to utilize the device of Bowers/Li with the explicit teaching on including anomaly data in the first training set and non-anomaly data in the second training set. One having ordinary skill in the art would have been motivated to implement this change before the effective filing date of the claimed invention, as this leads to avoiding labor-intensive temporal annotations of anomalous segments, performance, and utilizing useful datasets for detection and prediction (Sultani, Page 6846, Column 2, Paragraph 1, “… using only normal data alone may not be optimal for anomaly detection. We attempt to exploit both normal and anomalous videos. To avoid labor-intensive temporal annotations of anomalous segments in training videos, we learn a general model of anomaly detection using deep MIL framework with weakly labeled data …The experimental results on this dataset show that our proposed anomaly detection approach performs significantly better than baseline methods. Furthermore, we demonstrate the usefulness of our dataset for the task of anomalous activity recognition”).
However, Bowers/Li/Sultani do not explicitly disclose:
generating a time slide window random probability distribution curve for each time slide window generated for each real-time dataset of the plurality of real-time datasets;
comparing the one or more anomaly detection random probability distribution curves to each of the plurality of time slide window random probability distribution curves to determine if the corresponding time slide window includes anomaly data; and generate prediction results based on the comparison;
Nevertheless, Maurya teaches:
generating a time slide window random probability distribution curve for each time slide window generated for each real-time dataset of the plurality of real-time datasets;
(Maurya, Page 736, Column 2, Paragraph 3, “We detect sudden structural changes that may occur in Gaussian graphical models (GGMs). In the data modeling phase, we use background data from the past that has been identified to not contain any anomalies, to learn a GGM that describes the structural relationships between the random variables in the background data-generating process. In the anomaly monitoring phase, we move a sliding window over newly arriving data and perform detection within each window. Given a set of new observations of the random variables within a window i.e. foreground datapoints, we intend to learn the minimum structural changes in the background graphical model that can explain the new set of observations. Hence, we call our algorithm Contrastive Structured Anomaly Detection (CSAD) for Gaussian graphical model”. Maurya teaches generating a time slide window gaussian distributions for random variables within a window (which are known to one skilled in the art as random probability distribution curves as noted previously); thus, each time sliding window generates random probability distribution curves for each interval/window to compare and update the model)
comparing the one or more anomaly detection random probability distribution curves to each of the plurality of time slide window random probability distribution curves to determine if the corresponding time slide window includes anomaly data; and generate prediction results based on the comparison;
(Maurya, Page 736, Column 2, Paragraph 4, “We can accomplish our goal by solving the graphical lasso optimization problem via penalizing the resulting foreground graphical model to be as close as possible in structure to the background graphical model”. Maurya’s CSAD method (Contrastive Structured Anomaly Detection method) compares the background gaussian graphs/curves and the foreground gaussian graphs/curves to determine if any anomaly data is within the time sliding window and generates prediction results by updating the graphical model; thus, comparing the one or more anomaly detection random probability distribution curves to each of the plurality of time slide window random probability distribution curves).
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to utilize the device of Bowers/Li/Sultani with the explicit teaching of generating time slide window random distribution curves for specific windows and the comparisons to generate predictions. One having ordinary skill in the art would have been motivated to implement this change before the effective filing date of the claimed invention, as this leads to enhanced precision and recall of structural changes, generalization to identify model misfits, detecting vulnerabilities, and more (Maurya, Page 739, Column 1, Paragraph 3, “We proposed a method to detect structural changes in GGMs given datapoints from the background GGM and a structurally different foreground GGM. We evaluated the method on a network of 100 nodes and found promising improvements on both precision and recall of structural changes. One direction of future work is to test for statistical significance of detected changes using a statistical test or a scoring mechanism [11]. Another useful direction of generalization is to identify model misfit of a single GGM and use a mixture of GGMs to model background data …”; Page 736, Column 2, Paragraph 1, “Real-time analysis for structural anomaly detection in GGMs is useful in a wide variety of applications such as detecting organizational process disruption, detecting vulnerabilities, and modeling gene regulation anomalies that cause diseases”).
Regarding Claim 3:
Bowers/Li/Sultani/Maurya teach the device of Claim 1. Bowers further teaches:
compare the determination if the real-time dataset includes anomaly data to the determination if the corresponding time slide window includes anomaly data
(Bowers, Table 8; Fig. 25; Page 4, Column 2, [0049], “The edge processing provides many advantages including an ability to accept synchronous and asynchronous data, an ability to apply data fill techniques to align various data inputs appearing at various rates, an ability to analyze and trap data at the data input for anomalies and also trap other relevant data occurring at the same or similar time, an ability to accept data and inputs from disparate sources including manually entered data or decisions or status, an ability to accept input data in sub second increments, an ability to process data in a stochastic manner using Monte Carlo processing, an ability to compress data and present it stochastically, an ability to process data and note anomalies during stochastic processing and to present them for output in various ways including simple green-yellow-red alerts, and an ability to process multiple sensor data and combine them algorithmically to generate results that can then be analyzed”. The edge processing system, when determining anomaly data within the real-time dataset, is trapping the anomalous data and also trapping other relevant data occurring at the same or similar time. The evaluation analysis in Bower’s system innately compares all real-time datasets with their associated time slide windows (shown in Table 8: Evaluation Timestamps and Thresholds (which are defined by the model)) as the datasets are reviewed with the thresholds shown in Figure 8/9 and processed via Monte Carlo processing).
Regarding Claim 4:
Bowers/Li/Sultani/Maurya teach the device of Claim 3. Bowers further teaches:
generate prediction results based on the comparison of the two determinations.
(Bowers, Table 8; Fig. 11; Fig. 25-26; Page 10, Column 1, [0098], “As results are accumulated within the result database 548, an evaluation analysis subsystem 551 performs trend analysis on the time history of system performance characteristics generated within the models 539. Results from the evaluation analysis subsystem 551 may be employed in generating predictive analysis and prescriptive analysis”; Page 4, Column 2, [0049], “The edge processing provides many advantages including an … ability to process data and note anomalies during stochastic processing and to present them for output in various ways including simple green-yellow-red alerts, and an ability to process multiple sensor data and combine them algorithmically to generate results that can then be analyzed”; Page 18, Column 1, [0164 & 0165], “[0164] The predictive and prescriptive analytics may be performed on at least a component of the system based on the sensor data. The predictive analytics may include an estimate of an operation ( e.g., a health and remaining life) of a component of the system associated with the sensor. The prescriptive analytics may include a future action on a component of the system associated with the sensor. The prescriptive analytics may be based on a threshold performance of a component of the system associated with the sensor. [0165] At a step or module 2640, the method includes providing results and notifications based on the predictive and prescriptive analytics (e.g., from the edge subsystem 120)”. The AOS system hosts an evaluation analysis subsystem which employs predictive/prescriptive analysis (shown in Fig. 26) and provides alerts when thresholds are surpassed (shown in Fig. 25). The prediction results are based on the determination of the anomaly data set and the corresponding time slide window as the evaluation taught by Bowers always evaluates the real-time datasets with the corresponding time slide window (with or without anomaly data) as the probability distribution curves are used for the thresholds that are shown in Table 8. Edge processing processes the data and notes anomalies which is then analyzed by the predictive/prescriptive analytics).
Regarding Claim 5:
Bowers/Li/Sultani/Maurya teach the device of Claim 1. Bowers further teaches:
train the anomaly prediction model using a plurality of training datasets
(Bowers, Table 1; Page 10, Column 1, [0095], “An input database 527 provides a dynamic local cache of data to be used by the system environment in the execution of the system model. As data is accumulated from the sensors 506 and other data sources, it is stored within the input database 527. At regular intervals specified during installation or as a configurable setting, the input database 527 extracts the data values that have been accumulated between the previous time of execution and the current time of execution.”; Page 12, Column 2, [0119], “During a third stage 1060, stochastic model inputs are provided for each sensor. The inputs may be randomly generated trial values from each stochastic summary. Each set of trial values is then ready for further calculation within an AOS model. It should be noted that the number of trials for a model evaluation is a known constant, but can be adjusted during the model development and configuration phase. As the model is evaluated, the values for trial N are selected from each sensor data input and other input data nodes. Calculations then flow through interim calculations and arrive at overall system performance calculations. This process is repeated for all trials (in the exemplary embodiment 1000 trials)”. Table 1 shows the node library for the Cause and Effect Analysis model building within Edge Processing (the examiner interprets the Cause an Effect Analysis model to be an anomaly prediction model which is used to predict anomalies within the predictive analysis). Data is accumulated in the input database, which is partitioned by previous execution time, and the runtime subsystem executes the model on the dataset to continue training the AOS model. Within the Cause and Effect Analysis model, the trial values are used to train the model and will continue to adjust the model with more data that is accumulated (plurality of training datasets) from the sensors).
Regarding Claim 9:
Bowers/Li/Sultani/Maurya teach the device of Claim 1. Bowers further teaches:
align the plurality of real-time datasets; and
(Bowers, Page 4, Column 2, [0049], “The edge processing provides many advantages including an ability to accept synchronous and asynchronous data, an ability to apply data fill techniques to align various data inputs appearing at various rates …”. Edge processing teaches alignment of the real-time datasets via data fill within the first stage)
adjust an amount of time in each of the real-time datasets to be equal to a predetermined amount of time (Bowers, Page 12, Column 1, [0115], “All incoming values between a previous execution time 830 and a current execution time 840 are batched together and then summarized statistically as a data population 850. Within the data population 850, different approaches may be taken to characterize the sensor output, depending on the use of the sensor input within the cause and effect model”; Page 5, Column 2, [0058], “The prediction of system performance at the current time is an outcome of the cause and effect analysis and predictive analysis. Forecasting future system performance from current system performance includes the application of trend analysis, the details of which depend on a time interval between data collection periods, characteristic time period of the system, and a time interval between analysis executions”. The incoming values from the sensors (real-time datasets) are batched together based off a predetermined amount of time (between previous execution time and current execution time) and adjusted when compared within the anomaly prediction model (cause and effect model) based on time intervals for analysis).
Regarding Claim 10:
Bowers/Li/Sultani/Maurya teach the device of Claim 9. Bowers further teaches:
adjust each real-time dataset to include a predetermined amount of time.
(See the rejection of Claim 9, which explains to adjust an amount of time in each of the real-time datasets to be equal to a predetermined amount of time. A dataset to be equal to a predetermined amount of time also inherently includes a predetermined amount of time).
Regarding Claim 11:
Bowers/Li/Sultani/Maurya teach the device of Claim 9. Bowers further teaches:
adjust to include a predetermined number of data points from the one or more sensors (Bowers, Page 4, Column 1, [0045], “For instance, a stochastic diagnostic filter may deem the system to be out of bounds if there is greater than a five percent chance the system value is beyond a control threshold. By declaring a system state when a meaningful number of data points are observed or calculated, false positives from small fluctuations in incoming data can be reduced”; Page 21, Column 2, [0184], “Using M-of-N as a filtering process allows systems and processes to be evaluated at an aggregate level, while still alarming if a defined portion of the calculations are within one or more thresholds (e.g., trigger an alarm state if 90% of temperature readings are between 200 and 300 degrees)”. While using stochastic filters, system (tool) states may require a threshold for the amount of datapoints that are recorded (to reduce false positives) before consideration. The example provided by Bowers, raises an alarm state once 90% of the temperature readings (predetermined number of datapoints (recordings from the sensors)) are between 200 and 300 degrees (recording specific data points)).
Regarding Claim 13:
Bowers/Li/Sultani/Maurya teach the device of Claim 1. Bowers further teaches:
prediction results indicate a potential future issue with the tool to be analyzed.
(Bowers, Fig. 2; Table 8; Page 7, Column 2, [0073], “A prescriptive analysis subsystem 230 is a computer-driven analysis model that provides prescriptive action to be taken with respect to the system. In order to provide the prescriptive action, a quantitative predictive analysis is performed by a predictive analysis subsystem 220 of an operational status of the system at the present time, as well as at some future time”; Page 22, Column 1, [0186], “Intelligent predictor rules allow an AOS model to estimate the future point in time in which a calculated or sensor node will intersect a defined threshold … A historical trend is used to forecast forward in time to determine the date-time at which the defined threshold value will be crossed.”. Fig. 2 shows a block diagram for process subsystems including Predictive Analysis (230). Predictive Analysis occurs after the System Operational Status (where real-time characteristics are monitored via sensors) and is able to indicate/predict a potential future issue with the tool (system or system component). Table 8 describes the outputs generated by intelligent predictor rule and includes a field named Prediction Timestamp (Date-time at which the Intelligent Predictor will intercept the defined threshold value; may be NULL if current trend will not intercept)).
Regarding Claim 14:
Bowers/Li/Sultani/Maurya teach the device of Claim 1. Bowers further teaches:
raise an alarm when a future issue is detected.
(Bowers, Table 8 & 9; Fig. 25 & 26; Page 3, Column 1, [0036], “The system can alert and alarm on M samples of N population ("M-of-N") events where even a small (e.g., one percent) change in a sensor output can be detected and notified. An intelligent predictor is introduced that can be based on Kalman and other filter techniques where a long term event can be predicted well in advance of its occurrence and appropriate action taken”; Page 7, Column 1, [0070] “In one embodiment, the results 150 are primarily model and post processing outputs presented in a database, for example, a message queuing telemetry transport ("MQTT") protocol, or any other suitable data exchange protocol. These can be stored, processed, or analyzed locally or remotely. The notifications 160 can be of many forms, including but limited to, e-mail, text messaging, status changes of the system, and alarms that are generated by one or more triggering events within the AOS model.”. Table 8 illustrates the output/result types by the intelligent predictor rules which trigger Table 9 to provide MQTT notifications. Fig. 25 shows a representation of alert notifications based off model results from sensor values. The intelligent predictor is used to follow trends based off past and current data values and provides timestamps/thresholds on when a future issue is detected or going to occur to then raise alerts/alarms).
Regarding Claims 15, 17, and 19-20:
Claims 15, 17, and 19-20 incorporate substantively all the limitations of Claims 1, [3+4], and [9+10+11] in a method (Bowers, Fig. 26; Fig. 26 shows a method for operating the communication system which provides prediction and prescriptive analytics for a system/system component) and further recites no new limitations; thus, Claims 15 and 17-20 are rejected for reasons set forth in the rejections of Claims 1, [3+4], and [9+10+11], respectively.
Claims 6-8 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Bowers et al., US-2017/0351241 A1, in view of Li et al., “No Pane, No Gain: Efficient Evaluation of Sliding-Window Aggregates over Data Streams”, in view of Sultani et al., “Real-world Anomaly Detection in Surveillance Videos”, in view of Maurya et al., “Contrastive Structured Anomaly Detection for Gaussian Graphical Models”, in view of Gupta et al., US-9699049 B2.
Regarding Claim 6:
Bowers/Li/Sultani/Maurya teach the device of Claim 5. Bowers further teaches:
generate the plurality of training datasets … with anomalies from the plurality of historical dataset where the anomaly is not associated with the tool being measured and … noisy data.
(Bowers, Table 5/6; Page 4, Column 2, [0049], “The edge processing provides many advantages including an ability to accept synchronous and asynchronous data, an ability to apply data fill techniques to align various data inputs appearing at various rates, an ability to analyze and trap data at the data input for anomalies and also trap other relevant data occurring at the same or similar time, an ability to accept data and inputs from disparate sources including manually entered data or decisions or status, an ability to accept input data in sub second increments, an ability to process data in a stochastic manner using Monte Carlo processing, an ability to compress data and present it stochastically, an ability to process data and note anomalies during stochastic processing…”; Page 13, Column 2, [0127], “With respect to FIG. 12, system operating conditions 1210 are the sensor and other input values that represent the real-time or near real-time operational conditions of the system. System expected useful life 1220 is the total usage expected from the system or system component”; Page 16, Column 1, [0146], “The purpose of a stochastic filter is to process underlying noise in the sensor or calculated node to arrive at a more stable estimate of the system”; Page 16, Column 2, [0153], “The benefit of approaches shown in FIGS. 21 and 22 are the ability to drill down to understand the cause of anomalies, changes in system performance, and other variations”; The anomaly prediction model is trained on a plurality of training datasets and within these training datasets anomalies are trapped and noisy data is filtered while being stochastically processed (which generates the plurality of training datasets for training the prediction model). The datasets that are being processed are for a system or system component that is being analyzed for prediction (the examiner interprets tool as a device to carry out a function; thus, the system or system component is analogous with tool). Figures 21 & 22 are representing the stochastically causal influencers to analyze the anomalous data for the tool that is being represented via sensor data to find the underlying issues. Table 5/6 represent the “Black Swan” (extreme outliers) that affect the system (tool) adversely as the anomalous data does not associate with the system (tool). Both anomalous datasets and noisy datasets are able to be processed within Bower’s teachings via stochastic filtering and used to train the model).
While Bowers teaches the device to generate a plurality of training datasets with anomalies that are not associated with a tool being measured and noisy data. Bowers fails to teach:
removing datasets
However, Gupta does teach:
removing datasets
(Gupta, Fig. 5; Column 4, Lines 36-38, “A data collection module collects data of metrics periodically, e.g., at every minute interval. There are two major types of nodes, enterprise nodes, and data nodes”; Column 8, Lines 41-43, “If an anomaly score is received above the hard threshold for a node, which means the node is anomalous, then the node is removed from the cluster”; Column 5, Lines 48-52, “Theoretically the probability value for anomalous data node should exhibit a very small number, while the normal data nodes should have higher values. The same pattern may be observed in this dataset”. Fig. 5 shows the Feedback Policy Module which Gupta teaches to analyzes the anomalous data from a node and removed once it surpasses a threshold. The examiner interprets anomalous data nodes as a dataset with anomalous data; thus, Gupta teaches removing anomalous datasets).
Thus, incorporating the removing of datasets of Gupta with the generation of training datasets with anomalous/noisy data of Bowers would teach the device to generate the plurality of training datasets by removing datasets with anomalies from the plurality of historical datasets where the anomaly is not associated with the tool being measured and by removing noisy data.
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to utilize the device generating a plurality of training datasets with anomalies/noisy data of Bowers with the removal of datasets in Gupta to be able to generate a plurality of training datasets by removing anomalous/noisy datasets. One having ordinary skill in the art would have been motivated to implement this change before the effective filing date of the claimed invention, as this leads to the model being more precise by improving performance results by reducing false positives and allows for a chance to review structural integrity of the model by reviewing influencers for anomalies (Gupta, Column 6, Lines 48-50, “As the data node is anomalous, the factorized value should be very small and there must be features with significant smaller individual probabilities than others, and those contribute to low final probability. The final probabilities of these features would lie around the boundary of the distribution curve. These features with lower individual probabilities are selected and sorted in ascending order. Based on experimentation, the top 5 ranked features may be viewed as a good indication of root cause for the anomaly, and these may be called "influencers"”; Column 15, Lines 1-7, “This computing of anomalies may include building a model using a trading data set using Multivariate Gaussian Distribution, applying a Matthews Correlation coefficient as a threshold to reduce false positives, applying a half total error rate as a threshold to reduce false positives, and/or defining a function to calculate an anomaly score of data nodes”).
Regarding Claim 7:
Bowers/Li/Sultani/Maurya/Gupta teach the device of Claim 6; where using a plurality of non-anomaly and anomaly dataset is taught within Li. Bowers further teaches:
extract a plurality of raw datasets;
(Bowers, Fig. 10; Page 10, Column 1, [0095], “An input database 527 provides a dynamic local cache of data to be used by the system environment in the execution of the system model. As data is accumulated from the sensors 506 and other data sources, it is stored within the input database 527. At regular intervals specified during installation or as a configurable setting, the input database 527 extracts the data values that have been accumulated between the previous time of execution and the current time of execution”. The input database contains the raw datasets and continues to accumulate data as sensors collect information. With the cause and effect analysis, raw datasets are then extracted in a first stage and then a stochastic summary is created shown in Fig. 10); clean any noisy datasets (Bowers, Fig. 19; Page 3, Column 2, [0041], “The cause and effect approach to analysis provides significant benefits over purely statistical methods, exemplified by machine learning approaches or purely deterministic approaches that assume the availability of clean reliable data”; Page 16, Column 1, [0146], “The purpose of a stochastic filter is to process underlying noise in the sensor or calculated node to arrive at a more stable estimate of the system”. Fig. 19 (1910) shows a filtered distribution provided by a stochastic variation of the time varying history by filtering. The examiner interprets clean as a process of updating or removing. Stochastic filters are applied to all datasets, including noisy datasets, which cleans the dataset via filtering the underlying noise. The generation of the plurality of training datasets. The prediction model is trained on a plurality of training datasets and within these training datasets anomalies are trapped and noisy data is filtered while being stochastically processed (which generates the plurality of training datasets for training the prediction model)).
generate the second plurality of training datasets using a plurality of non-anomaly datasets; and
(Bowers, Page 12, Column 2, [0119], “During a third stage 1060, stochastic model inputs are provided for each sensor. The inputs may be randomly generated trial values from each stochastic summary. Each set of trial values is then ready for further calculation within an AOS model. … Calculations then flow through interim calculations and arrive at overall system performance calculations. This process is repeated for all trials (in the exemplary embodiment 1000 trials)”; Page 4, Column 2, [0049], “The edge processing provides … an ability to analyze and trap data at the data input for anomalies and also trap other relevant data occurring at the same or similar time … an ability to process data in a stochastic manner using Monte Carlo processing, an ability to compress data and present it stochastically, an ability to process data and note anomalies during stochastic processing …”. The datasets that are accumulated and even randomly generated (from stochastic summaries) are used to train the AOS model).
generate the first plurality of training datasets using the remaining plurality of datasets;
(Bowers, Page 12, Column 2, [0119], “During a third stage 1060, stochastic model inputs are provided for each sensor. The inputs may be randomly generated trial values from each stochastic summary. Each set of trial values is then ready for further calculation within an AOS model. … Calculations then flow through interim calculations and arrive at overall system performance calculations. This process is repeated for all trials (in the exemplary embodiment 1000 trials)”; Page 4, Column 2, [0049], “The edge processing provides … an ability to analyze and trap data at the data input for anomalies and also trap other relevant data occurring at the same or similar time … an ability to process data in a stochastic manner using Monte Carlo processing, an ability to compress data and present it stochastically, an ability to process data and note anomalies during stochastic processing …”. The model continues to train based on continuous sensor data when evaluation occurs using the remaining datasets (that are not trapped due to anomalies)).
Bowers/Li/Sultani/Maurya fail to teach:
classify the plurality of raw datasets as either non-anomaly or anomaly; for each anomaly dataset, determine if an observed anomaly is associated with a tool being observed and another source; if the observe anomaly is associated with another source, remove the corresponding anomaly dataset; align the remaining plurality of datasets to match time period;
However, Gupta does teach:
classify the plurality of raw datasets as either non-anomaly or anomaly;
(Gupta, Fig. 3; Fig. 4; Column 5: Lines 63-67 & Column 6: 1-5, “Cut-off Threshold to Differentiate Normal and Abnormal: It is useful to find out which data nodes in the cluster fall within an abnormal region from the Gaussian Distribution Function, so that they can be marked as anomalous nodes. In an example embodiment the probability for a data point can be very small, even for a normal data node, and potentially be classified as abnormal, which may indicate false-positives. Thus, it may be helpful to find a threshold that can help reduce the false-positives even more and also maximize true anomalies or true-positives”; Column 14, Lines 33-38 & 42-47, “The large-scale training system 1300 includes a data collector 1302, which collects various pieces of information about data, including, for example, a list of timestamps, a list of metrics, node names, colocation names, application names, and whether the data is normal or abnormal … Thus, the multinode hadoop cluster 1308 may contain normal CSV data and normal/abnormal CSV data. A predictive model 1310 may then operate on the data in the multi-node Hadoop cluster 1308 and compute a threshold 1312 to determine if data is normal or abnormal.”. Fig 3 & 4 show the data classified within two classifications: Normal/Abnormal Data. The raw datasets (nodes) are classified by each data point being compared to thresholds to determine if the data is either normal or abnormal).
for each anomaly dataset, determine if an observed anomaly is associated with a tool being observed and another source; if the observe anomaly is associated with another source, remove the corresponding anomaly dataset;
(Gupta, Fig. 5: 504; Column 6, Lines 23-29, “Anomaly Score Calculation: Since the probability of an anomaly may be a very small number, a function to calculate anomaly score of data nodes, which is between 0 and 10, may be defined. The higher score denotes a node is anomalous. Two thresholds may be picked: one is soft threshold and the other one is hard threshold. The thresholds can be determined from the training dataset”; Column 6, Lines 30-33 & 54-56 & 63-67, “Determining Features that Contributes to Anomaly: Once it is detected that a data node is anomalous, it may be helpful to find out the features that contribute to the anomaly … Based on experimentation, the top 5 ranked features may be viewed as a good indication of root cause for the anomaly, and these may be called "influencers" … Once the anomaly score is calculated and the list of influencers is obtained as described above, this information can be passed to the resource manager 104. Referring back to FIG. 1, the resource manager 104 may include a scheduler 110 and a YARN scheduler 112”; Column 8, Lines 41-46, “If an anomaly score is received above the hard threshold for a node, which means the node is anomalous, then the node is removed from the cluster. In this case, the feedback policy module will generate a plan for removing the node 45 from the cluster and provides this plan to action executor, which is described later”. Data nodes (datasets) are classified as abnormal/normal based on predefined thresholds and anomalous data is reviewed within the data nodes via anomaly score calculations. Anomaly Score Calculations are used to determine if a node is anomalous and once a node is determined to be removed it will be sent to the Action Executor/Scheduler as depicted in Fig. 5. The examiner interprets an anomaly score for a dataset (node) being higher than a hard threshold to be from another source as the data does not pertain to the current dataset (node/cluster)).
align the remaining plurality of anomaly datasets to match a same time period;
(Gupta, Column 5, Lines 9-17, “Since the machine-learning algorithm does not allow empty data points, data can be copied from the latest available time window. The rationale behind doing a copy is that timeseries data do not drastically change within short durations. For some metrics, data for the whole five-hour period of time are empty. In that case, those empty data may be filled out with a constant value, for example with zeros, so that it is ignored at a later time during training phase”. The machine algorithm does not allow empty data points as Gupta teaches filling out empty data with a constant value which keeps all datasets aligned even after removing an anomalous data set).
Thus, incorporating the extraction/generation/cleaning of datasets of Bowers with the classification/removal/alignment of datasets of Gupta would teach:
extract a plurality of raw datasets; classify the plurality of raw datasets as either non-anomaly or anomaly; for each anomaly dataset, determine if an observed anomaly is associated with a tool being observed and another source; if the observed anomaly is associated with another source, remove the corresponding anomaly dataset; align the remaining plurality of anomaly datasets to match time period; clean any noisy datasets; generate the second plurality of training datasets using a plurality of non-anomaly datasets; generate the first plurality of training datasets using the remaining plurality of datasets.
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to extract/clean datasets and the generation of the plurality of training datasets taught by Bowers in combination with the classification/removal and alignment taught by Gupta. One having ordinary skill in the art would have been motivated to implement this change before the effective filing date of the claimed invention, as this leads to the model being more precise by improving performance results by reducing false positives and allows for a chance to review structural integrity of the model by reviewing influencers for anomalies (Gupta, Column 6, Lines 48-50, “As the data node is anomalous, the factorized value should be very small and there must be features with significant smaller individual probabilities than others, and those contribute to low final probability. The final probabilities of these features would lie around the boundary of the distribution curve. These features with lower individual probabilities are selected and sorted in ascending order. Based on experimentation, the top 5 ranked features may be viewed as a good indication of root cause for the anomaly, and these may be called "influencers"”; Column 15, Lines 1-7, “This computing of anomalies may include building a model using a trading data set using Multivariate Gaussian Distribution, applying a Matthews Correlation coefficient as a threshold to reduce false positives, applying a half total error rate as a threshold to reduce false positives, and/or defining a function to calculate an anomaly score of data nodes”).
Regarding Claim 8:
Bowers/Li/Sultani/Maurya/Gupta teach the device of Claim 7.
While Bowers teaches:
… to determine one or more relationships with the remaining plurality of anomaly datasets.
(Bowers, Fig. 4; Fig. 25; Page 3, Column 1, [0037], “The system contemplates a cause and effect approach to analysis. The cause and effect approach includes a quantified series of linked algorithmic connections between measured data (the cause) and overall system performance (the effect)... The connections between sources of measured data and system performance may be introduced by system specifications, physics-based relationships, statistical analysis of historical data, definition by system subject matter experts, or other suitable means”. The cause and effect analysis determines relationships with the datasets within the Sensor Database (shown in Fig. 4). The examiner interprets relationships as connected concepts (thus connections between the measured data (sensor data) and the system/system component (tool being observed)). Fig. 25 depicts a presentation of information based on the cause and effect analysis (which determines relationships via cause and effect modeling to determine relationships between the measured data and system performance to produce Alert Notifications)).
Bowers/Li/Sultani/Maurya fail to teach to do so by clustering:
perform data clustering on the remaining plurality of anomaly datasets …
However, Gupta teaches the device to perform clustering on the datasets to make the determination:
perform data clustering on the remaining plurality of anomaly datasets to determine one or more relationships with the remaining plurality of anomaly datasets.
(Gupta, Fig. 13; Column 4, Lines 63-66, “Generally machine-learning models may need data in tabular format. Each row denotes data point for a specific timestamp and each column indicates a feature or Hadoop metric”; Column 5, Lines 18-28, “Standard deviation can be computed for each column (feature). If the standard deviation is less than a pre-defined threshold, generally a very small number, the column can be removed as input from the computation of the model. After these low-variance columns are removed, Principal Component Analysis (PCA) can be applied on it for feature reduction. In this case, some of the features for data points are highly correlated. For example, hadoop.app.cpuUser and hadoop.app.cpuNice are related with each other. By applying PCA, the system can find a linear combination of correlated features”; Column 14, Lines 42-49, “Thus, the multinode hadoop cluster 1308 may contain normal CSV data and normal/abnormal CSV data. A predictive model 1310 may then operate on the data in the multi-node Hadoop cluster 1308 and compute a threshold 1312 to determine if data is normal or abnormal. In an example embodiment, the predictive model may be saved in JSON format in a text file in one server or the HDFS”; Column 15, Lines 8-16, “Finally, at operation 1414, the prediction outcomes and metrics are used to move or reduce workloads from problematic clusters of nodes in the network. This may include 10 detecting that a data node is anomalous and, in response to the detection that the data node is anomalous, locating one or more features contributing to the anomaly. The locating may include deducing one or more features contributing to the anomaly using a single-variate Gaussian Distribution 15 Function.” Fig. 13 depicts the performing of clusters via Multi-node Hadoop Clusters. Data/Hadoop Nodes (datasets) can be removed from the clusters that were previously created. The system taught by Gupta also teaches using the Hadoop Cluster to correlate features (the examiner interprets as determining relationships) with the remaining data nodes (datasets)).
Thus, incorporating the determination of relationships of Bowers with the clustering of datasets within Gupta would teach perform data clustering on the remaining plurality of anomaly datasets to determine one or more relationships with the remaining plurality of anomaly datasets.
It would have been obvious to determine relationships, using the clustering method of Gupta, in the Bowers/Gupta combination. One having ordinary skill in the art would have been motivated to implement this change before the effective filing date of the claimed invention, as Hadoop clusters are able to analyze large amounts of data, early detection of anomalies, availability, responsiveness, and scheduling/workload optimizations (Gupta, Column 1: Lines 33-59, “A Hadoop cluster is a special type of computational cluster designed specifically for storing and analyzing large amounts of unstructured data in a distributed computing environment … Hadoop is widely deployed in industry and academia. In academia, Scientists have enabled Hadoop clusters to generate and analyze data at a larger Scale than was ever possible before. Today, scientists in a variety of disciplines such as earthquake simulations, bioinformatics, climate science, and astrophysics, are able to simulate and experiment on petabytes of data. Around the Hadoop ecosystem, a number of programming paradigms, applications, and services have evolved lately, including stream processing and analytics and real time graph processing to name a few. Many users running applications on Hadoop often times have hard deadlines on the finish time of their applications. Therefore, the Hadoop cluster providers strive for availability and responsiveness”).
Regarding Claim 18:
Claim 18 incorporates substantively all the limitations of Claims [5+7] in a method (Bowers, Fig. 26; Fig. 26 shows a method for operating the communication system which provides prediction and prescriptive analytics for a system/system component) and further recites no new limitations; thus, Claim 18 is rejected for reasons set forth in the rejections of Claims [5+7].
Claims 21 and 23 are rejected under 35 U.S.C. 103 as being unpatentable over Bowers et al., US-2017/0351241 A1, in view of Li et al., “No Pane, No Gain: Efficient Evaluation of Sliding-Window Aggregates over Data Streams”, in view of Sultani et al., “Real-world Anomaly Detection in Surveillance Videos”, in view of Maurya et al., “Contrastive Structured Anomaly Detection for Gaussian Graphical Models”, in view of Huertas et al., US-10,636,282 B2.
Regarding Claim 21:
Bowers/Li/Sultani/Maurya teach Claim 1 and also teaches the first duration + second duration of time within Claim 1 but does not explicitly disclose the specific intervals of time:
the first duration of time … and the second duration of time …
However, Huertas explicitly teaches:
is one minute and is thirty minutes for the first duration and second duration of time, respectively (Huertas, Column 10, Lines 39-42, “The configured processor retrieves first sensor data indexed over some period of … (one minute …)”; Column 12, Lines 35-37, “… the configured processor retrieves all first sensor data from the storage resource 105 indexed for 30 minutes prior to the peer threat alarm time”. The indexing/segmenting to be done over some period of time and notes one minute intervals; thus, the first duration is a one minute duration of time. Huertas also teaches a prior duration of time being a 30 minute look-back from the incident; thus, the second duration is a thirty minute duration of time).
Thus, incorporating the first and second durations of time of Bowers with the specific intervals (one minute and thirty minutes, respectively) of Huertas would teach the first duration of time is one minute and the second duration of time is thirty minutes.
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to utilize the durations from the Bowers device in combination with the specific durations of time taught by Huerta’s device. One having ordinary skill in the art would have been motivated to implement this change before the effective filing date of the claimed invention, as this leads to the model using consistent/fixed/indexed sensor data and prior historical data for alert detection and specific intervals for durations of time and can be utilized for controlled SaaS where users can connect to the cloud infrastructure and not underlying architectures and only user-specific application configuration settings where the application can be used by consumers for loss prevention in multiple services/functions (Huertas, Column 5, Lines 48-50, “Software as a Service (SaaS): the capability provided to the consumer is to use the provider's applications running on a cloud infrastructure. The applications are accessible from various client devices through a thin client interface such as a web browser (e.g., web-based e-mail). The consumer does not manage or control the underlying cloud infrastructure including network, servers, operating systems, storage, or even individual application capabilities, with the possible exception of limited user-specific application configuration settings”; Column 1, Lines 11-24, “Security alarm systems are generally designed to detect the occurrence of a condition that presents a risk of loss to a protected domain … Security alarm systems are used in residential, commercial, industrial, and governmental organization properties for protection … may also provide life safety functions, such as fire detection and suppression services, flood warning and prevention, and severe weather warnings and associated loss prevention actions …”).
Regarding Claim 23:
Claim 23 incorporates substantively all the limitations of Claims 21 in a method (Bowers, Fig. 26; Fig. 26 shows a method for operating the communication system which provides prediction and prescriptive analytics for a system/system component) and further recites no new limitations; thus, Claim 23 is rejected for reasons set forth in the rejections of Claims 21.
Claims 22 and 24 are rejected under 35 U.S.C. 103 as being unpatentable over Bowers et al., US-2017/0351241 A1, in view of Li et al., “No Pane, No Gain: Efficient Evaluation of Sliding-Window Aggregates over Data Streams”, in view of Sultani et al., “Real-world Anomaly Detection in Surveillance Videos”, in view of Maurya et al., “Contrastive Structured Anomaly Detection for Gaussian Graphical Models”, in view of Bhagvat et al., US 8,712,575 B2.
Regarding Claim 22:
Bowers/Li/Sultani/Maurya teach Claim 1 and also teaches the tool within Claim 1 but does not explicitly disclose the tool being a double sided grinder.
However, Bhagvat explicitly teaches:
… is a double sided grinder (Bhagvat, Column 4, Lines 18-, “One aspect is a method of processing a semiconductor wafer using a double side grinder of the type that holds the wafer between a pair of grinding wheels and a pair of hydrostatic pads having a hydrostatic pressure therein”. Bhagvat utilizes a system and method to determine and predict a grinding stage for a double sided grinder tool).
Thus, incorporating the tool of Bowers with the specific tool (double sided grinder) of Bhagvat would teach the tool is a double sided grinder.
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to utilize the tool from Bowers in combination with the specific tool (double sided grinder) taught by Bhagvat. One having ordinary skill in the art would have been motivated to implement this change before the effective filing date of the claimed invention, as this leads to specific tool analysis where prediction can be done on a double sided grinder to improve nanotopology of the wafers as the grinder is used on the wafers (Bhagvat, Abstract, “Systems and methods are disclosed for modulating the hydro static pressure in a double side wafer grinder having a pair of grinding wheels. … Pattern detection software is used to predict a grinding stage based on the measured electrical current. The hydrostatic pressure is changed by flow control valves at each stage to change the clamping pressure applied to the wafer and to thereby improve nanotopology in the processed wafer”).
Regarding Claim 24:
Claim 24 incorporates substantively all the limitations of Claims 22 in a method (Bowers, Fig. 26; Fig. 26 shows a method for operating the communication system which provides prediction and prescriptive analytics for a system/system component) and further recites no new limitations; thus, Claim 24 is rejected for reasons set forth in the rejections of Claims 22.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 nonprovisional extension fee (37 CFR 1.17(a)) 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 IBRAHIM RAHMAN whose telephone number is (703)756-1646. The examiner can normally be reached M-F 8am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Kakali Chaki can be reached at (571) 272-3719. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/I.R./Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights