DETAILED ACTION
This Office Action is in response to communications filed on January 29, 2026 for Application No. 18/116,772, in which claims 1-20 are presented for examination. The amendments filed on January 29, 2026 have been entered, where claims 1-3, 6, and 9-19 are amended.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to abstract ideas without significantly more.
Regarding Claim 1:
Step 1: Claim 1 is a process claim. Therefore, Claims 1-8 are directed to a statutory category of eligible subject matter.
Step 2A Prong 1: If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the "Mental Processes" grouping of abstract ideas. Here, steps of the claimed method are mental processes. Specifically, the claim recites
“A method comprising: determining reference distribution data associated with a feature used . . . wherein the reference distribution data comprises bin ranges determined based on a distribution of training data values for the feature” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed or known data, where the opinion includes a determination of ranges with reference to other known or observed information, which may be aided by pen and paper);
“sort at least one of input data or predictive data into bins corresponding to the bin ranges to generate current distribution data” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed data, using a sorting mechanism, which may be aided by pen and paper); and
“based on the current distribution data, performance of a corrective action” (mental process - amounts to exercising judgment to determine changes to observed data, based on a determination that correction is required, which may be aided by pen and paper)
Step 2A Prong 2: This judicial exception is not integrated into a practical application.
The claim recites the additional elements:
“to train a machine learning model . . . to generate a trained machine learning model . . . to enable the edge device to . . . causing . . .” (amounts to mere instructions to apply the judicial exception on generic and unspecialized computer components, which do not impose any meaningful limits on practicing the abstract idea);
“providing the reference distribution data to an edge device . . . receiving the current distribution data” (providing and receiving data amounts to insignificant extra-solution activity because the transmission of data is incidental to the claimed subject matter); and
“associated with substrate processing equipment . . . associated with the substrate processing equipment . . . associated with the feature from the edge device responsive to the using of the trained machine learning model at the edge device . . . associated with the substrate processing equipment” (amounts to merely generally linking the use of the judicial exception to a particular technological environment or field of use, which do not impose any meaningful limits on practicing the abstract idea).
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The claim recites the additional elements:
“to train a machine learning model . . . to generate a trained machine learning model . . . to enable the edge device to . . . causing . . .” (mere instructions to apply the exception using generic computer components does not provide an inventive concept);
“providing the reference distribution data to an edge device . . . receiving the current distribution data” (transmitting data is well‐understood, routine, and conventional, see generally Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362; see also buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014); therefore the limitation, which is recited with a high level of generality, remains insignificant extra-solution activity even upon reconsideration); and
“associated with substrate processing equipment . . . associated with the substrate processing equipment . . . associated with the feature from the edge device responsive to the using of the trained machine learning model at the edge device . . . associated with the substrate processing equipment” (merely generally linking the use of the judicial exception to a particular technological environment or field of use does not provide an inventive concept).
For the reasons above, Claim 1 is rejected as being directed to an abstract idea without significantly more. This rejection applies equally to dependent claims 2-8. The additional limitations of the dependent claims are addressed below.
Regarding Claim 2:
Step 2A Prong 1: See the rejection of Claim 1 above, which Claim 2 depends on.
Step 2A Prong 2: This judicial exception is not integrated into a practical application.
The claim recites the additional elements:
“wherein the corrective action comprises at least one of providing an alert” (providing data amounts to insignificant extra-solution activity because the transmission of data is incidental to the claimed subject matter);
“or retraining the trained machine learning model” (amounts to mere instructions to apply the judicial exception on generic and unspecialized computer components, which do not impose any meaningful limits on practicing the abstract idea); and
“associated with the substrate processing equipment” (amounts to merely generally linking the use of the judicial exception to a particular technological environment or field of use, which do not impose any meaningful limits on practicing the abstract idea).
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The claim recites the additional elements:
“wherein the corrective action comprises at least one of providing an alert” (transmitting data is well‐understood, routine, and conventional, see generally Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362; see also buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014); therefore the limitation, which is recited with a high level of generality, remains insignificant extra-solution activity even upon reconsideration);
“or retraining the trained machine learning model” (mere instructions to apply the exception using generic computer components does not provide an inventive concept); and
“associated with the substrate processing equipment” (merely generally linking the use of the judicial exception to a particular technological environment or field of use does not provide an inventive concept).
Accordingly, Claim 2 is rejected as being directed to an abstract idea without significantly more.
Regarding Claim 3:
Step 2A Prong 1: See the rejection of Claim 1 above, which Claim 3 depends on. Here, the claim recites additional elements that are mental processes. Specifically, the claim recites
“wherein the reference distribution data comprises the bin ranges” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed or known data, in a specific bin range format, which may be aided by pen and paper).
Step 2A Prong 2: This judicial exception is not integrated into a practical application.
The claim recites the additional elements:
“of training data associated with the feature” (amounts to merely generally linking the use of the judicial exception to a particular technological environment or field of use, which do not impose any meaningful limits on practicing the abstract idea) and
“the training data being used to train the machine learning model” (amounts to mere instructions to apply the judicial exception on generic and unspecialized computer components, which do not impose any meaningful limits on practicing the abstract idea).
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The claim recites the additional elements:
“of training data associated with the feature” (merely generally linking the use of the judicial exception to a particular technological environment or field of use does not provide an inventive concept) and
“the training data being used to train the machine learning model” (mere instructions to apply the exception using generic computer components does not provide an inventive concept).
Accordingly, Claim 3 is rejected as being directed to an abstract idea without significantly more.
Regarding Claim 4:
Step 2A Prong 1: See the rejection of Claim 3 above, which Claim 4 depends on. Here, the claim recites additional elements that are mental processes. Specifically, the claim recites
“identifying the training data . . . ; and sorting the training data into bins, wherein the determining of the reference distribution data comprises determining the bin ranges of the training data sorted into the bins” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed data, using a specific sorting method, which may be aided by pen and paper).
Step 2A Prong 2: This judicial exception is not integrated into a practical application.
The claim recites the additional elements:
“associated with the feature” (in the event that identifying specifically of data associated with the feature requires a particular technological environment, amounts to merely generally linking the use of the judicial exception to a particular technological environment or field of use, which do not impose any meaningful limits on practicing the abstract idea) and
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The claim recites the additional elements:
“associated with the feature” (merely generally linking the use of the judicial exception to a particular technological environment or field of use does not provide an inventive concept).
Accordingly, Claim 4 is rejected as being directed to an abstract idea without significantly more.
Regarding Claim 5:
Step 2A Prong 1: See the rejection of Claim 1 above, which Claim 5 depends on. Here, the claim recites additional elements that are mental processes. Specifically, the claim recites
“determining a difference between the current distribution data and the reference distribution data; and determining that the difference between the current distribution data and the reference distribution data meets a threshold value, wherein the causing of the corrective action is responsive to the difference meeting the threshold value” (mental process – amounts to exercising judgment to compare and evaluate observed or known information to determine whether to take an action, which may be aided by pen and paper).
Step 2A Prong 2 & Step 2B: There are no elements left for consideration of implementation within a practical application or for consideration of significantly more.
Accordingly, Claim 5 is rejected as being directed to an abstract idea without significantly more.
Regarding Claim 6:
Step 2A Prong 1: See the rejection of Claim 4 above, which Claim 6 depends on. Here, the claim recites additional elements that are mental processes. Specifically, the claim recites
“wherein the current distribution data comprises counts per bin histogram data” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed data, using a sorting mechanism, which may be aided by pen and paper).
Step 2A Prong 2: This judicial exception is not integrated into a practical application.
The claim recites the additional elements:
“associated with the at least one of the input data or the predictive data” (amounts to merely generally linking the use of the judicial exception to a particular technological environment or field of use, which do not impose any meaningful limits on practicing the abstract idea).
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The claim recites the additional elements:
“associated with the at least one of the input data or the predictive data” (merely generally linking the use of the judicial exception to a particular technological environment or field of use does not provide an inventive concept).
Accordingly, Claim 6 is rejected as being directed to an abstract idea without significantly more.
Regarding Claim 7:
Step 2A Prong 1: See the rejection of Claim 1 above, which Claim 7 depends on. Here, the claim recites additional elements that are mental processes. Specifically, the claim recites
“performing, based on the reference distribution data and the current distribution data, comparison of distributions” (mental process – amounts to exercising judgement to evaluate observed or known data, which may be aided by pen and paper).
Step 2A Prong 2 & Step 2B: There are no elements left for consideration of implementation within a practical application or for consideration of significantly more.
Accordingly, Claim 7 is rejected as being directed to an abstract idea without significantly more.
Regarding Claim 8:
Step 2A Prong 1: See the rejection of Claim 1 above, which Claim 8 depends on. Here, the claim recites additional elements that are mental processes. Specifically, the claim recites
“wherein the reference distribution data and the current distribution data are histogram data” (mental process – amounts to exercising judgement to organize data in the form of a histogram, which may be aided by pen and paper).
Step 2A Prong 2 & Step 2B: There are no elements left for consideration of implementation within a practical application or for consideration of significantly more.
Accordingly, Claim 8 is rejected as being directed to an abstract idea without significantly more.
Regarding Claim 9:
Step 1: Claim 9 is a process claim. Therefore, Claims 9-13 are directed to a statutory category of eligible subject matter.
Step 2A Prong 1: If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the "Mental Processes" grouping of abstract ideas. Here, steps of the claimed method are mental processes. Specifically, the claim recites
“A method comprising: . . . wherein the reference distribution data comprises bin ranges determined based on a distribution of training data values for the feature . . . determining current distribution data associated with the feature responsive to the using” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed or known data, where the opinion includes a determination of ranges with reference to other known or observed information, which may be aided by pen and paper);
“wherein the determining of the current distribution data comprises sorting at least one of the input data or predictive data into bins corresponding to the bin ranges to determine the current distribution data” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed data, using a sorting mechanism, which may be aided by pen and paper); and
“performance of a corrective action” (mental process - amounts to exercising judgment to determine changes to observed data, based on a determination that correction is required, which may be aided by pen and paper).
Step 2A Prong 2: This judicial exception is not integrated into a practical application.
The claim recites the additional elements:
“receiving, from a server device, reference distribution data . . . providing the current distribution data to the server device” (providing and receiving data amounts to insignificant extra-solution activity because the transmission of data is incidental to the claimed subject matter);
“associated with a feature used to train a machine learning model associated with substrate processing equipment to generate a trained machine learning model . . . data associated with the substrate processing equipment . . . associated with the substrate processing equipment” (amounts to merely generally linking the use of the judicial exception to a particular technological environment or field of use, which do not impose any meaningful limits on practicing the abstract idea); and
“using the trained machine learning model based on input . . . of the trained machine learning model; and . . . to cause” (amounts to mere instructions to apply the judicial exception on generic and unspecialized computer components, which do not impose any meaningful limits on practicing the abstract idea).
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The claim recites the additional elements:
“receiving, from a server device, reference distribution data . . . providing the current distribution data to the server device” (transmitting data is well‐understood, routine, and conventional, see generally Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362; see also buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014); therefore the limitation, which is recited with a high level of generality, remains insignificant extra-solution activity even upon reconsideration);
“associated with a feature used to train a machine learning model associated with substrate processing equipment to generate a trained machine learning model . . . data associated with the substrate processing equipment . . . associated with the substrate processing equipment” (merely generally linking the use of the judicial exception to a particular technological environment or field of use does not provide an inventive concept); and
“using the trained machine learning model based on input . . . of the trained machine learning model; and . . . to cause” (mere instructions to apply the exception using generic computer components does not provide an inventive concept.
For the reasons above, Claim 9 is rejected as being directed to an abstract idea without significantly more. This rejection applies equally to dependent claims 10-13. The additional limitations of the dependent claims are addressed below.
Regarding Claim 10, the claim recites limitations that are all substantially the same as limitations of Claim 2. The claim is also directed to performing mental processes without integration into a practical component or significantly more.
Accordingly, Claim 10 is rejected under the same rationale.
Regarding Claim 11, the claim recites limitations that are all substantially the same as limitations of Claim 3. The claim is also directed to performing mental processes without integration into a practical component or significantly more.
Accordingly, Claim 11 is rejected under the same rationale.
Regarding Claim 12:
Step 2A Prong 1: See the rejection of Claim 11 above, which Claim 12 depends on. Here, the claim recites additional elements that are mental processes. Specifically, the claim recites
“wherein the current distribution data comprises the bin ranges of the at least one” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed data, which may be aided by pen and paper).
Step 2A Prong 2: This judicial exception is not integrated into a practical application.
The claim recites the additional elements:
“of the input data or the predictive data associated with the feature” (amounts to merely generally linking the use of the judicial exception to a particular technological environment or field of use, which do not impose any meaningful limits on practicing the abstract idea) and
“the input data being used to generate the predictive data via the machine learning model” (amounts to mere instructions to apply the judicial exception on generic and unspecialized computer components, which do not impose any meaningful limits on practicing the abstract idea).
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The claim recites the additional elements:
“of the input data or the predictive data associated with the feature” (merely generally linking the use of the judicial exception to a particular technological environment or field of use does not provide an inventive concept) and
“the input data being used to generate the predictive data via the machine learning model” (mere instructions to apply the exception using generic computer components does not provide an inventive concept.
Accordingly, Claim 12 is rejected as being directed to an abstract idea without significantly more.
Regarding Claim 13:
Step 2A Prong 1: See the rejection of Claim 12 above, which Claim 13 depends on. Here, the claim recites additional elements that are mental processes. Specifically, the claim recites
“identifying the input data and the predictive data . . . and sorting the at least one of the input data or the predictive data into the bins, wherein the determining of the reference distribution data comprises determining the bin ranges of the input data and the predictive data sorted into the bins” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed data, using a specific sorting method, which may be aided by pen and paper).
Step 2A Prong 2: This judicial exception is not integrated into a practical application.
The claim recites the additional elements:
“associated with the feature” (in the event that identifying specifically of data associated with the feature requires a particular technological environment, amounts to merely generally linking the use of the judicial exception to a particular technological environment or field of use, which do not impose any meaningful limits on practicing the abstract idea) and
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The claim recites the additional elements:
“associated with the feature” (merely generally linking the use of the judicial exception to a particular technological environment or field of use does not provide an inventive concept).
Accordingly, Claim 13 is rejected as being directed to an abstract idea without significantly more.
Regarding Claim 14:
Step 1: Claim 1 is a machine claim. Therefore, Claims 14-20 are directed to a statutory category of eligible subject matter.
Step 2A Prong 1: If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the "Mental Processes" grouping of abstract ideas. Here, steps of the claimed method are mental processes. Specifically, the claim recites
“determining reference distribution data associated with a feature used . . . wherein the reference distribution data comprises bin ranges determined based on a distribution of training data values for the feature” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed or known data, where the opinion includes a determination of ranges with reference to other known or observed information, which may be aided by pen and paper);
“sort at least one of input data or predictive data into bins corresponding to bin ranges to generate current distribution data” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed data, using a sorting mechanism, which may be aided by pen and paper); and
“based on the current distribution data, performance of a corrective action” (mental process - amounts to exercising judgment to determine changes to observed data, based on a determination that correction is required, which may be aided by pen and paper).
Step 2A Prong 2: This judicial exception is not integrated into a practical application.
The claim recites the additional elements:
“A non-transitory machine-readable storage medium storing instructions which, when executed, cause a processing device to perform operations comprising: . . . to train a machine learning model . . . to generate a trained machine learning model . . . to enable the edge device to. . . causing . . .” (amounts to mere instructions to apply the judicial exception on generic and unspecialized computer components, which do not impose any meaningful limits on practicing the abstract idea);
“providing the reference distribution data to an edge device . . . receiving the current distribution data” (providing and receiving data amounts to insignificant extra-solution activity because the transmission of data is incidental to the claimed subject matter); and
“associated with substrate processing equipment . . . associated with the substrate processing equipment . . . associated with the feature from the edge device responsive to the using of the trained machine learning model at the edge device . . . associated with the substrate processing equipment” (amounts to merely generally linking the use of the judicial exception to a particular technological environment or field of use, which do not impose any meaningful limits on practicing the abstract idea).
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The claim recites the additional elements:
“A non-transitory machine-readable storage medium storing instructions which, when executed, cause a processing device to perform operations comprising: . . . to train a machine learning model . . . to generate a trained machine learning model . . . to enable the edge device to. . . causing . . .” (mere instructions to apply the exception using generic computer components does not provide an inventive concept);
“providing the reference distribution data to an edge device . . . receiving the current distribution data” (transmitting data is well‐understood, routine, and conventional, see generally Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362; see also buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014); therefore the limitation, which is recited with a high level of generality, remains insignificant extra-solution activity even upon reconsideration); and
“associated with substrate processing equipment . . . associated with the substrate processing equipment . . . associated with the feature from the edge device responsive to the using of the trained machine learning model at the edge device . . . associated with the substrate processing equipment” (merely generally linking the use of the judicial exception to a particular technological environment or field of use does not provide an inventive concept).
For the reasons above, Claim 14 is rejected as being directed to an abstract idea without significantly more. This rejection applies equally to dependent claims 15-20. The additional limitations of the dependent claims are addressed below.
Regarding Claim 15, the claim recites limitations that are all substantially the same as limitations of Claim 2, in the form of a machine. The claim is also directed to performing mental processes without integration into a practical component or significantly more.
Accordingly, Claim 15 is rejected under the same rationale.
Regarding Claim 16, the claim recites limitations that are all substantially the same as limitations of Claim 3, in the form of a machine. The claim is also directed to performing mental processes without integration into a practical component or significantly more.
Accordingly, Claim 16 is rejected under the same rationale.
Regarding Claim 17, the claim recites limitations that are all substantially the same as limitations of Claim 4, in the form of a machine. The claim is also directed to performing mental processes without integration into a practical component or significantly more.
Accordingly, Claim 17 is rejected under the same rationale.
Regarding Claim 18, the claim recites limitations that are all substantially the same as limitations of Claim 5, in the form of a machine. The claim is also directed to performing mental processes without integration into a practical component or significantly more.
Accordingly, Claim 18 is rejected under the same rationale.
Regarding Claim 19:
Step 2A Prong 1: See the rejection of Claim 17 above, which Claim 19 depends on. Here, the claim recites additional elements that are mental processes. Specifically, the claim recites
“wherein the current distribution data is based on sorting the at least one of the input data or the predictive data into the bins . . . and wherein the current distribution data comprises counts per bin histogram data” (mental process – amounts to exercising judgment to form an opinion on the distribution of observed data, using a sorting mechanism, which may be aided by pen and paper).
Step 2A Prong 2: This judicial exception is not integrated into a practical application.
The claim recites the additional elements:
“via the edge device” (amounts to mere instructions to apply the judicial exception on generic and unspecialized computer components, which do not impose any meaningful limits on practicing the abstract idea).
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The claim recites the additional elements:
“via the edge device” (mere instructions to apply the exception using generic computer components does not provide an inventive concept).
Accordingly, Claim 19 is rejected as being directed to an abstract idea without significantly more.
Regarding Claim 20, the claim recites limitations that are all substantially the same as limitations of Claim 7, in the form of a machine. The claim is also directed to performing mental processes without integration into a practical component or significantly more.
Accordingly, Claim 20 is rejected under the same rationale.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-4, 6-12, 14-17, and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Xu et al. (hereinafter Xu) (“Concept Drift and Covariate Shift Detection Ensemble with Lagged Labels”) in view of Dong et al. (hereinafter Dong) (“ISEE: Industrial Internet of Things perception in solar cell detection based on edge computing”).
Regarding Claim 1, Xu teaches a method comprising (Pg. 1506, Col. 1, Para. 8, “Figure 1 shows the general approach of our methodology . . . to detect both concept drifts and covariate shifts . . . [and] to retrain the model”; see also Pg. 1506, Col. 2, Fig. 1, “Fig. 1. Overall approach. For each batch of incoming data, we calculate six descriptive statistics of time series, then utilize drift detection module for each of them to monitor drift and decide what data used to retrain”):
determining reference distribution data associated with a feature used to train a machine learning model . . . to generate a trained machine learning model (Pg. 1507, Col. 1, Para. 2, “In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch . . . the training dataset (i.e. the dataset that the model is fitted on)”, where the “training set” includes a reference “distribution” of “discretized . . . features”, and is therefore associated with these features, which are used as part of “the training dataset” to train a machine learning “model” to generate a trained, “fitted”, “model”, see generally Pg. 1504, Col. 1, Abstract, “In model serving, having one fixed model during the entire often life-long inference process is usually detrimental to model performance, as data distribution evolves over time, resulting in lack of reliability of the model trained on historical data”, where the “training set” is part of the “historical data” that “the model trained on”; see also Pg. 2, Para. 7, “The proposed method can detect both concept drift and covariate shift; it can determine when to retrain”, where the “proposed method”, which as discussed above includes an analysis of how the “data distribution evolves over time” is used to generate a trained model, “retrain”),
wherein the reference distribution data comprises bin ranges determined based on a distribution of training data values for the feature (Pg. 1507, Col. 1, Para. 2, “Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
where F denotes the set of all features”, where “Xtr” is the reference data, see Pg. 1507, Col. 1, Para. 3, “training features Xtr”, which is formulated as the reference distribution when calculating “the Hellinger distance between the training set and the current batch”, which comprises bins, “
z
”, that must have ranges for the “features” to be “discretized” within the associated bin, “
z
∈
f
”; and where the determination of whether a bin will be included in the reference distribution, and thus have an associated range, is based on whether an associated feature, “f=z” for each of “
z
∈
f
”, is included in the set of features, “
f
∈
F
”, which in turn is determined based on both the distribution of training data values for the feature, “Xtr,f”, and the distribution of current batch data values for the features, “Xn,f”);
providing the reference distribution data . . . (Pg. 1507, Col. 1, Para. 2, “we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
where F denotes the set of all features”, where the reference distribution data, “
PNG
media_image2.png
64
90
media_image2.png
Greyscale
” for “
z
∈
f
”, is provided as a variable in equation “3”; see generally Pg. 1507, Col. 1, Para. 3, “training features Xtr”)
to sort at least one of input data or predictive data into bins corresponding to the bin ranges to generate current distribution data (Pg. 1507, Col. 1, Para. 2, “Hellinger distance is often used to quantify similarity between two distributions . . . Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
”, where the input data “Xn” is identified as being associated with a given feature, “f=z”, associated with a bin corresponding with a nonzero bin range, “
z
∈
f
”, and sorted into the bin, “|Xtr,f=z|”, and therefore the “distribution” of “discretized . . . features” of the “current batch”, the current distribution data, is generated based on the bin ranges “f=z” for the set of features “
z
∈
f
”);
receiving current distribution data associated with the feature from . . . [the deployment environment] responsive to the using of the trained machine learning model at . . . [the deployment environment] (Pg. 1507, Col. 1, Para. 2, “In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch”, where the “current batch” includes a current “distribution” of “discretized . . . features”, and is therefore associated with these features; Pg. 1504, Col. 2, Para. 2, “We consider the model serving process where we continuously receive new batches of data . . . for inference”, and where “receiv[ing] new batches of data” as part of “the model serving process . . . for inference” demonstrates that the “batches” are “receive[d]” from the deployment environment, “model serving process” environment, and responsive to using the trained model “for inference”); and
causing, based on the current distribution data, performance of a corrective action associated with . . . [the trained machine learning model] (Pg. 1507, Col. 1, Para. 2, “Hellinger distance: In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch as . . . [equation 3]. By averaging over all features, we calculate the distance between two datasets or batches. If the distance between the training dataset (i.e. the dataset that the model is fitted on) and the current batch is large, the existing model no longer fits the current distribution and needs to be retrained”, where “
PNG
media_image2.png
64
90
media_image2.png
Greyscale
” for “
z
∈
f
” is the reference distribution data and “
PNG
media_image3.png
68
86
media_image3.png
Greyscale
” for “
z
∈
f
” is the current distribution data, which are used to determine whether the corrective action of “retraining” of the trained model “needs to” occur).
Xu does not explicitly disclose . . . associated with substrate processing equipment . . . to an edge device associated with the substrate processing equipment to enable the edge device . . . the edge device . . . the edge device . . . the substrate processing equipment.
However, Dong teaches [a method comprising] (Pg. 4, Col. 1, Para. 1, “there is an urgent need for a more efficient, accurate, and simple detection method to realize solar panels’ intelligent detection. Therefore, this study used edge equipment to detect the defects of solar panels and formed a complete framework to solve the system theory”)[:]
. . . [a machine learning model] associated with substrate processing equipment (Pg. 5, Col. 2, Para. 3, “This study found that the edge computing unit has a powerful NPU neural unit for AI hardware acceleration. Under the same circumstances, the NPU’s AI computing power consumption is less than 1% of the GPU, significantly saving the hardware cost in a large amount of data processing in the solar photovoltaic industry” and Pg. 4, Col. 1, Para. 2, “In this study, embedded devices are used in the edge to detect the defects of solar panels in real time . . . The overall block diagram of defect detection of solar panels is shown in Figure 4. In the whole system, the PN junction of the solar panel is positively electrified by a DC power supply, which makes the solar panel produce positive bias voltage and emit infrared light. When the electrons are injected into the solar cell and recombined with the existing holes, the energy will release in the form of photons in the local range. The wavelength of the released photons is similar to the infrared wavelength, and then the image of the solar panel is collected in real time by the industrial infrared camera placed in the darkroom”, where the “edge” “embedded devices” are associated with the substrates, such as the contents of the “solar panels”, and substrate processing equipment, such as the “infrared camera”, and the machine learning model is part of “the edge computing unit[‘s] . . . powerful NPU neural unit”, so the model is associated with the substrate processing equipment)
. . . [providing data associated with a feature used to train a machine learning model to generate a trained machine learning model] to an edge device (Pg. 5, Col. 2, Para. 4, “We run the convolutional neural networks (CNNs) in real time on edge computing unit to detect solar panels’ defects to transplant quickly”, where the model data, such as its parameters, must be provided to the “edge computing unit” in order to be “run . . . on” it; see also Pg. 6, Col. 2, Para. 1, “In this article, we design a compiler to map the CNN model to the hardware architecture. Experiments on edge computing units show that the proposed design process can achieve CNN acceleration with high energy efficiency”; Pg. 8, Col. 2, Para. 1-2, “This article collects 300 images of EL . . . the final data set is 30,000, and 24,000 of them act as training sets . . . After 68 epochs of iterative learning, the model’s accuracy rate in the training set is 96:875%”, where the data of the “model”, such as its “parameters”, see Pg. 7, Col. 1, Table 1 and Pg. 6, Col. 2, Para. 4, “Model neural network structure of the convolution layer and pooling layer parameters of the network are shown in Table 1”, are determined during “learning”, which associates the data with the features of the “training sets” because the data is adjusted based on the “training sets”, see Pg. 7, Col. 2, Para. 2, “[data of] each neural network layer changes during training. This is because updating the front layer’s training parameters will lead to the change. . . [data] in the later layer”, to generate the trained model with an “accuracy rate in the training set . . . [of] 96:875%”; see generally Pg. 6, Col. 2, Para. 4, “Convolution layer: the purpose of this layer is to extract the features of the input image”, where the training data must have features in order to be used to generate a model that “extract[s] the features of the input image”)
associated with the substrate processing equipment (Pg. 4, Col. 1, Para. 2, “In this study, embedded devices are used in the edge to detect the defects of solar panels in real time . . . The overall block diagram of defect detection of solar panels is shown in Figure 4. In the whole system, the PN junction of the solar panel is positively electrified by a DC power supply, which makes the solar panel produce positive bias voltage and emit infrared light. When the electrons are injected into the solar cell and recombined with the existing holes, the energy will release in the form of photons in the local range. The wavelength of the released photons is similar to the infrared wavelength, and then the image of the solar panel is collected in real time by the industrial infrared camera placed in the darkroom”, where, as discussed above, the “edge” “embedded devices” are associated with the substrates, such as the contents of the “solar panels”, and substrate processing equipment, such as the “infrared camera”; see also Pg. 5, Col. 1, Fig. 6 and Pg. 2, Col. 1-2, Fig. 1-3)
to enable the edge device to . . . [perform operations associated with the data] (Pg. 5, Col. 2, Para. 4, “We run the convolutional neural networks (CNNs) in real time on edge computing unit to detect solar panels’ defects to transplant quickly”, where the edge device performs actions associated with the model data, “run the convolutional neural networks (CNNs) in real time on edge computing unit”, which as discussed above is the data received by the edge device; see also Pg. 6, Col. 2, Para. 1, “In this article, we design a compiler to map the CNN model to the hardware architecture. Experiments on edge computing units show that the proposed design process can achieve CNN acceleration with high energy efficiency”; Pg. 8, Col. 2, Para. 1-2, “This article collects 300 images of EL . . . the final data set is 30,000, and 24,000 of them act as training sets . . . After 68 epochs of iterative learning, the model’s accuracy rate in the training set is 96:875%”,)
. . . [receiving current data associated with the feature from] the edge device [responsive to the using of the trained machine learning model at] the edge device (Pg. 4, Col. 1, Para. 1, “The edge device collects the solar panel video image through the camera in real time and then uses its own powerful NPU module, combined with the CNN algorithm transplanted to the edge, to detect the defects of solar panels in real time”, where the “edge device” uses the “CNN algorithm transplanted to the edge” to generate current data, “detect the defects of the solar panels”, which must be received to be “detect[ed]” in “real time”; Pg. 8, Col. 2, Para. 1-2, “This article collects 300 images of EL . . . the final data set is 30,000, and 24,000 of them act as training sets . . . After 68 epochs of iterative learning, the model’s accuracy rate in the training set is 96:875%”, where the current data is associated with the feature because it uses the “train[ed]” “model”, which is based on the “training sets”)
. . . [to make a determination, based on the machine learning model output, associated with] the substrate processing equipment (Pg. 4, Col. 1, Para. 1, “The edge device collects the solar panel video image through the camera in real time and then uses its own powerful NPU module, combined with the CNN algorithm transplanted to the edge, to detect the defects of solar panels in real time”, where the machine learning model, “the CNN algorithm transplanted to the edge”, is used to make a determination associated with the substrate equipment, “collects the solar panel video image through the camera in real time” to “detect the defects of solar panels in real time”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the method of causing performance of a corrective action associated with a trained machine learning model based on a comparison between reference distribution data and current distribution data, wherein the reference distribution data comprises bin ranges determined based on distribution of feature values used to train the machine learning model, the reference distribution data is provided for computations, the current distribution data is received from the deployment environment after using the trained model, and, based on the current distribution data, a corrective action associated with the trained machine learning model is performed of Xu with the method of detecting defects in solar panels using an edge device and machine learning model associated with substrate processing equipment, wherein data determined based on a feature used to train a machine learning model is provided to the edge device to enable the edge device to perform operations associated with the data, current data associated with the feature is transmitted from the edge device after using the trained model at the edge device, and the machine learning model is used to make a determination associated with the substrate processing equipment of Dong in order to apply the benefits of dataset drift correction (Xu, Pg. 1504, Col. 1, Abstract, “In model serving, having one fixed model during the entire often life-long inference process is usually detrimental to model performance, as data distribution evolves over time, resulting in lack of reliability of the model trained on historical data. It is important to detect changes and retrain the model in time”; Xu, Pg. 1512, Col. 1, Para. 1, “We present a novel and effective drift detection method in the practical lag of labels setting, and extensive experiments show that our method consistently outperforms the state-of-the-art drift detection methods by a large margin”) to a machine learning application at an edge device, which allows for real time solar panel defect detection (Dong, Pg. 1, Abstract, “ISEE . . . uses the powerful neural network processing unit module of the edge computing unit, combined with the convolutional neural network algorithm transplanted to the edge, to detect the defects of solar panels in real time”), which promotes accuracy (Dong, Pg. 1, Abstract, “ISEE effectively improves the automation degree and identification accuracy in the production and detection process of solar photovoltaic cells and reduces the cost of operation and maintenance. The accuracy rate reaches 93:75%, which has significant theoretical research significance and practical application value”) and cost efficiency by training a model on a cloud server (Dong, Pg. 4, Col. 2, Para. 1, “B/S structure is browser and server structure . . . This architecture dramatically simplifies the client computer load, reduces system maintenance, upgrades cost and workload, and reduces users’ total cost”), while utilizing the computational resources of the edge device (Dong, Pg. 5, Col. 2, Para. 3, “This study found that the edge computing unit has a powerful NPU neural unit for AI hardware acceleration. Under the same circumstances, the NPU’s AI computing power consumption is less than 1% of the GPU, significantly saving the hardware cost in a large amount of data processing in the solar photovoltaic industry”).
Regarding Claim 2, Xu in view of Dong teach the method of claim 1, wherein the corrective action comprises at least one of providing an alert or retraining the trained machine learning model associated with the substrate processing equipment (Xu, Pg. 1507, Col. 1, Para. 2, “Hellinger distance: In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
. By averaging over all features, we calculate the distance between two datasets or batches. If the distance between the training dataset (i.e. the dataset that the model is fitted on) and the current batch is large, the existing model no longer fits the current distribution and needs to be retrained”, where “
PNG
media_image2.png
64
90
media_image2.png
Greyscale
” for “
z
∈
f
” is the reference distribution data and “
PNG
media_image3.png
68
86
media_image3.png
Greyscale
” for “
z
∈
f
” is the current distribution data, which are used to determine whether the corrective action of “retraining” of the trained model “needs to” occur, which, in view of Dong, is associated with the substrate processing equipment, see Dong, Pg. 5, Col. 2, Para. 3, “This study found that the edge computing unit has a powerful NPU neural unit for AI hardware acceleration. Under the same circumstances, the NPU’s AI computing power consumption is less than 1% of the GPU, significantly saving the hardware cost in a large amount of data processing in the solar photovoltaic industry” and Dong, Pg. 4, Col. 1, Para. 2, “In this study, embedded devices are used in the edge to detect the defects of solar panels in real time . . . The overall block diagram of defect detection of solar panels is shown in Figure 4. In the whole system, the PN junction of the solar panel is positively electrified by a DC power supply, which makes the solar panel produce positive bias voltage and emit infrared light. When the electrons are injected into the solar cell and recombined with the existing holes, the energy will release in the form of photons in the local range. The wavelength of the released photons is similar to the infrared wavelength, and then the image of the solar panel is collected in real time by the industrial infrared camera placed in the darkroom”, where the “edge” “embedded devices” are associated with the substrates, such as the contents of the “solar panels”, and substrate processing equipment, such as the “infrared camera”, and the machine learning model is part of “the edge computing unit[‘s] . . . powerful NPU neural unit”, so the model is associated with the substrate processing equipment).
The reasons for obviousness were discussed in regard to the rejection of claim 1 above and remain applicable here.
Regarding Claim 3, Xu in view of Dong teach the method of claim 1, wherein the reference distribution data comprises the bin ranges of training data associated with the feature (Xu, Pg. 1507, Col. 1, Para. 2, “Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
where F denotes the set of all features”, where “
z
∈
f
” are the bins, which must have ranges for the “features” to be “discretized”, and where “Xtr” are the “training features”, see Xu, Pg. 1507, Col. 1, Para. 3, “training features Xtr”, and therefore the bin ranges are associated with the feature because the bin ranges determine where the feature will be “discretized” as and they are both associated with the “training set”),
the training data being used to train the machine learning model (Xu, Pg. 1504, Col. 1, Abstract, “In model serving, having one fixed model during the entire often life-long inference process is usually detrimental to model performance, as data distribution evolves over time, resulting in lack of reliability of the model trained on historical data”, where the “training set”, discussed in regard to equation 3, is part of the “historical data” that “the model trained on”).
Regarding Claim 4, Xu in view of Dong teach the method of claim 3 further comprising: identifying the training data associated with the feature (Xu, Pg. 1507, Col. 1, Para. 3, “training features Xtr”, where any given feature of the “training features Xtr”, can be considered the feature which is identified as part of the training set as condition of the “Hellinger Distance” equation, “f=z”, see Xu, Pg. 1507, Col. 1, Para. 2, equation 3)
and sorting the training data into bins, wherein the determining of the reference distribution data comprises determining the bin ranges of the training data sorted into the bins (Xu, Pg. 1507, Col. 1, Para. 2, “Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
where F denotes the set of all features”, where the “training set” is sorted into bins, “
z
∈
f
”, based on the features, “Xtr, f=z”, where the determined bin ranges, “
z
∈
f
” of “discretized the features”, determines the reference distribution data because “discretiz[ing] the features” of the “training dataset” generates the reference distribution data, “
PNG
media_image3.png
68
86
media_image3.png
Greyscale
” for “
z
∈
f
”).
Regarding Claim 6, Xu in view of Dong teach the method of claim 4, wherein the current distribution data comprises counts per bin histogram data associated with the at least one of the input data or the predictive data (Xu, Pg. 1507, Col. 1, Para. 2, “Hellinger distance: In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
. By averaging over all features, we calculate the distance between two datasets or batches. If the distance between the training dataset (i.e. the dataset that the model is fitted on) and the current batch is large, the existing model no longer fits the current distribution and needs to be retrained”, where “
PNG
media_image3.png
68
86
media_image3.png
Greyscale
” for “
z
∈
f
” is the current distribution data, which is within the broadest reasonable interpretation of histogram data because they include counts “Xn,f=z” sorted into bins “f=z” for “
z
∈
f
”; notably, “Xn” in the denominator is consistent with relative frequency histogram data; Xu, Pg. 1504, Col. 2, Para. 2, “We consider the model serving process where we continuously receive new batches of data (a batch can as well be a single sample corresponding to streaming) for inference”, where the “batches” are input data “for inference”).
Regarding Claim 7, Xu in view of Dong teach the method of claim 1 further comprising performing, based on the reference distribution data and the current distribution data, comparison of distributions (Xu, Pg. 1507, Col. 1, Para. 2, “Hellinger distance: In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image4.png
152
740
media_image4.png
Greyscale
. By averaging over all features, we calculate the distance between two datasets or batches. If the distance between the training dataset (i.e. the dataset that the model is fitted on) and the current batch is large, the existing model no longer fits the current distribution and needs to be retrained”, where “
PNG
media_image2.png
64
90
media_image2.png
Greyscale
” for “
z
∈
f
” is the reference distribution data and “
PNG
media_image3.png
68
86
media_image3.png
Greyscale
” for “
z
∈
f
” is the current distribution data,, which are compared to determine whether the corrective action of “retraining” of the trained model “needs to” occur).
Regarding Claim 8, Xu in view of Dong teach the method of claim 1, wherein the reference distribution data and the current distribution data are histogram data (Xu, Pg. 1507, Col. 1, Para. 2, “Hellinger distance: In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
. By averaging over all features, we calculate the distance between two datasets or batches. If the distance between the training dataset (i.e. the dataset that the model is fitted on) and the current batch is large, the existing model no longer fits the current distribution and needs to be retrained”, where “
PNG
media_image2.png
64
90
media_image2.png
Greyscale
” for “
z
∈
f
” is the reference distribution data and “
PNG
media_image3.png
68
86
media_image3.png
Greyscale
” for “
z
∈
f
” is the current distribution data, both of which are within the broadest reasonable interpretation of histogram data because they include counts “Xn,f=z” sorted into bins “f=z” for “
z
∈
f
”; notably, “Xn” in the denominator is consistent with relative frequency histogram data).
Regarding Claim 9, Xu in view of Dong teach a method comprising: receiving, from a server device (Xu, Pg. 1506, Col. 1, Para. 8, “Figure 1 shows the general approach of our methodology . . . to detect both concept drifts and covariate shifts . . . [and] to retrain the model”, where the “methodology” includes providing data for computations, see Xu, Pg. 1507, Col. 1, Para. 2, “we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
”, where the reference distribution data, “
PNG
media_image2.png
64
90
media_image2.png
Greyscale
” for “
z
∈
f
”, discussed in detail below, is provided as a variable in equation “3”, and where, in view of Dong, the providing is a transmission from a “PC” server device with a “Nvidia GTX1080” “GPU”, see Dong, Pg. 8, Col. 1, Table 2 and Dong, Pg. 8, Col. 2, Para. 2, “This article collects 300 images of EL . . . we need enough GPU memory to store the training data”, and to an edge device, see Dong, Pg. 4, Col. 1, Para. 1, “combined with the CNN algorithm transplanted to the edge”, where data is transmitted from server to “edge” device in order to utilize the computational capacity of the “edge” device, see Dong, Pg. 5, Col. 2, Para. 3, “This study found that the edge computing unit has a powerful NPU neural unit for AI hardware acceleration. Under the same circumstances, the NPU’s AI computing power consumption is less than 1% of the GPU, significantly saving the hardware cost in a large amount of data processing in the solar photovoltaic industry”),
reference distribution data associated with a feature used to train a machine learning model associated with substrate processing equipment to generate a trained machine learning model (Xu, Pg. 1507, Col. 1, Para. 2, “In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch . . . the training dataset (i.e. the dataset that the model is fitted on)”, where the “training set” includes a reference “distribution” of “discretized . . . features”, and is therefore associated with these features, which are used as part of “the training dataset” to train a machine learning “model” to generate a trained, “fitted”, “model”, see generally Xu, Pg. 1504, Col. 1, Abstract, “In model serving, having one fixed model during the entire often life-long inference process is usually detrimental to model performance, as data distribution evolves over time, resulting in lack of reliability of the model trained on historical data”, where the “training set” is part of the “historical data” that “the model trained on”, where, in view of Dong, the machine learning model is associated with substrate processing equipment, see Dong, Pg. 5, Col. 2, Para. 3, “This study found that the edge computing unit has a powerful NPU neural unit for AI hardware acceleration. Under the same circumstances, the NPU’s AI computing power consumption is less than 1% of the GPU, significantly saving the hardware cost in a large amount of data processing in the solar photovoltaic industry” and Dong, Pg. 4, Col. 1, Para. 2, “In this study, embedded devices are used in the edge to detect the defects of solar panels in real time . . . The overall block diagram of defect detection of solar panels is shown in Figure 4. In the whole system, the PN junction of the solar panel is positively electrified by a DC power supply, which makes the solar panel produce positive bias voltage and emit infrared light. When the electrons are injected into the solar cell and recombined with the existing holes, the energy will release in the form of photons in the local range. The wavelength of the released photons is similar to the infrared wavelength, and then the image of the solar panel is collected in real time by the industrial infrared camera placed in the darkroom”, where the “edge” “embedded devices” are associated with the substrates, such as the contents of the “solar panels”, and substrate processing equipment, such as the “infrared camera”, and the machine learning model is part of “the edge computing unit[‘s] . . . powerful NPU neural unit”, so the model is associated with the substrate processing equipment),
wherein the reference distribution data comprises bin ranges determined based on a distribution of training data values for the feature (Xu, Pg. 1507, Col. 1, Para. 2, “Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
where F denotes the set of all features”, where “Xtr” is the reference data, see Xu, Pg. 1507, Col. 1, Para. 3, “training features Xtr”, which is formulated as the reference distribution when calculating “the Hellinger distance between the training set and the current batch”, which comprises bins, “
z
”, that must have ranges for the “features” to be “discretized” within the associated bin, “
z
∈
f
”; and where the determination of whether a bin will be included in the reference distribution, and thus have an associated range, is based on whether an associated feature, “f=z” for each of “
z
∈
f
”, is included in the set of features, “
f
∈
F
”, which in turn is determined based on both the distribution of training data values for the feature, “Xtr,f”, and the distribution of current batch data values for the features, “Xn,f”);
using the trained machine learning model based on input data (Dong, Pg. 5, Col. 2, Para. 4, “We run the convolutional neural networks (CNNs) in real time on edge computing unit to detect solar panels’ defects to transplant quickly”)
associated with substrate processing equipment (Dong, Pg. 4, Col. 1, Para. 2, “In this study, embedded devices are used in the edge to detect the defects of solar panels in real time . . . The overall block diagram of defect detection of solar panels is shown in Figure 4. In the whole system, the PN junction of the solar panel is positively electrified by a DC power supply, which makes the solar panel produce positive bias voltage and emit infrared light. When the electrons are injected into the solar cell and recombined with the existing holes, the energy will release in the form of photons in the local range. The wavelength of the released photons is similar to the infrared wavelength, and then the image of the solar panel is collected in real time by the industrial infrared camera placed in the darkroom”, where the “edge” “embedded devices” are associated with the substrates, such as the contents of the “solar panels”, and substrate processing equipment, such as the “infrared camera”; see also Dong, Pg. 5, Col. 1, Fig. 6 and Dong, Pg. 2, Col. 1-2, Fig. 1-3);
determining current distribution data associated with the feature responsive to the using of the trained machine learning model (Xu, Pg. 1507, Col. 1, Para. 2, “In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch”, where the “current batch” includes a current “distribution” of “discretized . . . features”, and is therefore associated with these features; Xu, Pg. 1504, Col. 2, Para. 2, “We consider the model serving process where we continuously receive new batches of data . . . for inference”, and where “receiv[ing] new batches of data” as part of “the model serving process . . . for inference” demonstrates that the “batches” are “receive[d]” from the deployment environment, “model serving process” environment, and responsive to using the trained model “for inference”),
wherein the determining of the current distribution data comprises sorting at least one of the input data or predictive data into bins corresponding to the bin ranges to determine the current distribution data (Xu, Pg. 1507, Col. 1, Para. 2, “Hellinger distance is often used to quantify similarity between two distributions . . . Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
”, where the input data “Xn” is identified as being associated with a given feature, “f=z”, associated with a bin corresponding with a nonzero bin range, “
z
∈
f
”, and sorted into the bin, “|Xtr,f=z|”, and therefore the “distribution” of “discretized . . . features” of the “current batch”, the current distribution data, is generated based on the bin ranges “f=z” for the set of features “
z
∈
f
”); and
providing the current distribution data to the server device to cause performance of a corrective action associated with the substrate processing equipment (Xu, Pg. 1507, Col. 1, Para. 2, “Hellinger distance: In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch as . . . [equation 3]. By averaging over all features, we calculate the distance between two datasets or batches. If the distance between the training dataset (i.e. the dataset that the model is fitted on) and the current batch is large, the existing model no longer fits the current distribution and needs to be retrained”, where “
PNG
media_image2.png
64
90
media_image2.png
Greyscale
” for “
z
∈
f
” is the reference distribution data and “
PNG
media_image3.png
68
86
media_image3.png
Greyscale
” for “
z
∈
f
” is the current distribution data, which are used to determine whether the corrective action of “retraining” of the trained model “needs to” occur; Xu, Fig. 1, “retrain using batch w to n”, where the “batch” data, which as discussed above includes the current distribution data, must be provided to the training environment to perform the corrective action of “retrain using batch w to n”, and where, in view of Dong, the training environment is the “PC” server with a “Nvidia GTX1080” “GPU”, see Dong, Pg. 8, Col. 1, Table 2 and Dong, Pg. 8, Col. 2, Para. 2, “This article collects 300 images of EL . . . we need enough GPU memory to store the training data”, to utilize its “cost” efficiency, see Dong, Pg. 4, Col. 2, Para. 1, “B/S structure is browser and server structure . . . This architecture dramatically simplifies the client computer load, reduces system maintenance, upgrades cost and workload, and reduces users’ total cost” and the corrective action is associated with the substrate processing equipment, see Dong, Pg. 4, Col. 1, Para. 1, “The edge device collects the solar panel video image through the camera in real time and then uses its own powerful NPU module, combined with the CNN algorithm transplanted to the edge, to detect the defects of solar panels in real time”, where the machine learning model, “the CNN algorithm transplanted to the edge”, which as discussed above is used to determine an action associated with the substrate equipment, “collects the solar panel video image through the camera in real time” to “detect the defects of solar panels in real time”).
The reasons for obviousness were discussed in regard to the rejection of claim 1 above and remain applicable here.
Regarding Claim 10, the additional elements of the dependent claim are substantially the same as limitations of Claim 2, therefore it is rejected under the same rationale.
Regarding Claim 11, the additional elements of the dependent claim are substantially the same as limitations of Claim 3, therefore it is rejected under the same rationale.
Regarding Claim 12, Xu in view of Dong teach the method of claim 11, wherein the current distribution data comprises the bin ranges of the at least one of the input data or the predictive data associated with the feature, the input data being used to generate the predictive data via the machine learning model (Xu, Pg. 1507, Col. 1, Para. 2, “Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
where F denotes the set of all features”, where “
z
∈
f
” are the bins, which must have ranges for the “features” to be “discretized”, and where “Xn” are the current features, and therefore the bin ranges are associated with the feature because the bin ranges determine where the feature will be “discretized” into which bin; Xu, Pg. 1504, Col. 2, Para. 2, “We consider the model serving process where we continuously receive new batches of data (a batch can as well be a single sample corresponding to streaming) for inference”, where the “batches” are input data “for inference”, therefore, the bin ranges of current distribution data are associated with input features).
Regarding Claim 14, Xu in view of Dong teach a non-transitory machine-readable storage medium storing instructions which, when executed, cause a processing device to perform operations comprising (Dong, Pg. 8, Col. 1, Para. 3-4, “The experimental conditions are as follows (Figure 10, Table 2): The main control chip of the ISEE edge computing unit is the rk3399 pro version. Its CPU is an arm 64-bit processor (dual-core cortex-a72 + quad-core cortexa53), and the central frequency is up to 1.8 GHz. NPU neural unit is also built into the chip to support AI hardware acceleration”; Dong, Pg. 8, Col. 1, Table 2, where the non-transitory machine-readable storage medium, “Storage”, must include instructions, executed by the processor, “CPU”, to execute the “programmable” “framework”, see Dong, Pg. 6, Col. 1, Para. 1, “Based on these requirements, this article proposes a programmable and flexible CNN edge computing framework”): . . . .
The reasons for obviousness were discussed in regard to the rejection of claim 1 above and remain applicable here.
The remaining limitations are substantially the same as limitations of Claim 1, therefore it is rejected under the same rationale.
Regarding Claim 15, the additional elements of the dependent claim are substantially the same as limitations of Claim 2, therefore it is rejected under the same rationale.
Regarding Claim 16, the additional elements of the dependent claim are substantially the same as limitations of Claim 3, therefore it is rejected under the same rationale.
Regarding Claim 17, the additional elements of the dependent claim are substantially the same as limitations of Claim 4, therefore it is rejected under the same rationale.
Regarding Claim 19, Xu in view of Dong teach the non-transitory machine-readable storage medium of claim 17, wherein the current distribution data is based on sorting the at least one of the input data or the predictive data into the bins (Xu, Pg. 1507, Col. 1, Para. 2, “Hellinger distance: In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
. By averaging over all features, we calculate the distance between two datasets or batches. If the distance between the training dataset (i.e. the dataset that the model is fitted on) and the current batch is large, the existing model no longer fits the current distribution and needs to be retrained”, where “
PNG
media_image3.png
68
86
media_image3.png
Greyscale
” for “
z
∈
f
” is the current distribution data, and is based on sorting feature counts “Xn,f=z” into bins “f=z” for “
z
∈
f
”; Xu, Pg. 1504, Col. 2, Para. 2, “We consider the model serving process where we continuously receive new batches of data (a batch can as well be a single sample corresponding to streaming) for inference”, where the “batches” are input data “for inference”)
via the edge device (where, in view of Dong, the computations occur after transmission from a “PC” server device with a “Nvidia GTX1080” “GPU”, see Dong, Pg. 8, Col. 1, Table 2 and Dong, Pg. 8, Col. 2, Para. 2, “This article collects 300 images of EL . . . we need enough GPU memory to store the training data”, and to an edge device, see Dong, Pg. 4, Col. 1, Para. 1, “combined with the CNN algorithm transplanted to the edge”, in order to utilize the computational capacity of the “edge” device, see Dong, Pg. 5, Col. 2, Para. 3, “This study found that the edge computing unit has a powerful NPU neural unit for AI hardware acceleration. Under the same circumstances, the NPU’s AI computing power consumption is less than 1% of the GPU, significantly saving the hardware cost in a large amount of data processing in the solar photovoltaic industry”), and
wherein the current distribution data comprises counts per bin histogram data (Xu, Pg. 1507, Col. 1, Para. 2, “Hellinger distance: In statistics and measure theory, the Hellinger distance is often used to quantify similarity between two distributions . . . We utilize the Hellinger distance to construct the distance between two datasets. Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
. By averaging over all features, we calculate the distance between two datasets or batches. If the distance between the training dataset (i.e. the dataset that the model is fitted on) and the current batch is large, the existing model no longer fits the current distribution and needs to be retrained”, where “
PNG
media_image3.png
68
86
media_image3.png
Greyscale
” for “
z
∈
f
” is the current distribution data, which is within the broadest reasonable interpretation of histogram data because they include counts “Xn,f=z” sorted into bins “f=z” for “
z
∈
f
”; notably, “Xn” in the denominator is consistent with relative frequency histogram data; Xu, Pg. 1504, Col. 2, Para. 2, “We consider the model serving process where we continuously receive new batches of data (a batch can as well be a single sample corresponding to streaming) for inference”, where the “batches” are input data “for inference”).
The reasons for obviousness were discussed in regard to the rejection of claim 1 above and remain applicable here.
Regarding Claim 20, the additional elements of the dependent claim are substantially the same as limitations of Claim 7, therefore it is rejected under the same rationale.
Claims 5, 13, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Xu in view of Dong and Benson et al. (hereinafter Benson) (Pat. App. Pub. No. US 2022/0318646).
Regarding Claim 5, Xu in view of Dong teach the method of claim 1 further comprising: determining a difference between the current distribution data and the reference distribution data (Xu, Pg. 1507, Col. 1, Para. 2, “Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
where F denotes the set of all features”, where “
PNG
media_image2.png
64
90
media_image2.png
Greyscale
” for “
z
∈
f
” is the reference distribution data and “
PNG
media_image3.png
68
86
media_image3.png
Greyscale
” for “
z
∈
f
” is the current distribution data, which are subtracted, “-”, to determine the difference);
and determining that [a value calculated using] . . . the current distribution data and the reference distribution data meets a threshold value (Xu, Pg. 1508, Col. 1, Para. 1, “To monitor drift, we apply the following rules . . . If pi + si > pimin + 3 · simin, we report drift and retrain the model using batches in the warning zone”, where the value “> pimin + 3 · simin” is the threshold and the value “pi + si” is calculated using the difference, “Hellinger distance”, see Xu, Pg. 1506-1507, Col. 2-2, Para. 1-6, “A. Descriptive statistics calculation . . . At each time step six scores are calculated, which are either based on only the current batch or based on the current batch as well as the previous batches . . . [including] Hellinger distance . . . B. Drift detection Each descriptive statistic defined in the previous subsection generates a time series over time.. . . to detect drift of a time series . . . we denote the progressive average by pi . . . and standard deviation by si”),
wherein the causing of the corrective action is responsive to . . . [the value] meeting the threshold value (Xu, Pg. 1508, Col. 1, Para. 1, “To monitor drift, we apply the following rules . . . If pi + si > pimin + 3 · simin, we report drift and retrain the model using batches in the warning zone”, where “retrain[ing]” occurs in response to “pi + si” meeting the value “> pimin + 3 · simin”).
Xu in view of Dong do not explicitly disclose . . . the difference between . . . the difference . . . (where the difference between the distributions is not directly compared to the threshold).
However, Benson teaches . . . [determining] the difference between [a new data distribution and the distribution of the model training dataset] . . . [wherein a corrective action is caused based on] the difference [meeting a threshold value] . . . (Para. [0074], “the monitoring module 440 can monitor the new data (e.g., daily control or operational parameters, as input to the trained model to predict) distribution of the unit, and monitor the difference between the new data distribution and the previous input data (on which the model was trained) distribution. If the differences (e.g., the deviation of the inference data from the training data over time) between the previous input data on which the model was trained and the new data that is coming every day exceeds a predetermined threshold, the monitoring module 440 can generate a warning, that the model may need to be updated, retrained, or retuned”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the performance of a corrective action based on a determination that a value generated using the difference between reference distribution data and current distribution data meets a threshold value of Xu in view of Dong with the performance of a corrective action directly based on the difference between a training data distribution and a new input data distribution of Benson in order to detect model drift (Benson, Para. [0073], “Typically model drift may occur when a new failure mode of the TCCS is coming up, or when a fix is in place for an existing climate control failure mode, or there is some problem occurred in the manufacturing process, etc. In such case, the trained model does not match the actual scenario in in the field (i.e., what is predicted does not match what is actually happening in the field), the monitoring module 440 can generate a warning that the model may need to be updated, retrained, or retuned”) using a simplified approach that reduces the necessary computations by solely focusing on the difference between the reference data distribution and the current data distribution (compare Benson, Para. [0074], where the “data distribution[s]” can be “monitor[ed]” using a single “difference”, with Xu, Pg. 1506-1507, Col. 2-2, Para. 1-6, where “six scores are calculated”, some of which are based on “only the current batch” and some of which also include “the previous batches”).
Regarding Claim 13, Xu in view of Dong and Benson teach the method of claim 12, further comprising: identifying the input data and the predictive data associated with the feature (Xu, Pg. 1506, Col. 1, Para. 7, “Data (Xtr; Ytr) are used to train the current classification model clf. We denote an incoming batch as Bn = (Xn; Yn)”, where both the input data, “Xtr” and “Xn”, are identified, and are associated with the feature because they contain the feature, see for Xu, Pg. 1507, Para. 3, “training features Xtr” and Xu, Pg. 1507, Col. 1, para. 2, Equation 3, where both data sets have include “f=z”, of the features “
z
∈
f
”, and where, in view of Benson, predictive data is also identified, see Benson, Para. [0072], “The monitoring module 440 can monitor the number of changes in e.g., the distribution of “unhealthy” prediction using e.g., statistical process control (SPC)”, which is associated with the feature because it is the output from the input data, “Xn”, which as discussed above contains the feature).
and sorting the at least one of the input data or the predictive data into bins, wherein the determining of the reference distribution data comprises determining the bin ranges of the input data and the predictive data sorted into the bins (Xu, Pg. 1507, Col. 1, Para. 2, “Hellinger distance is often used to quantify similarity between two distributions . . . Having discretized the features, we define the Hellinger distance between the training set and the current batch as
PNG
media_image1.png
170
826
media_image1.png
Greyscale
”, where the input data “Xtr” is identified as being associated with a given feature, “f=z”, associated with a bin corresponding with a nonzero bin range, “
z
∈
f
”, and sorted into the bin, “|Xtr,f=z|”, and therefore the “distribution” of “discretized . . . features” of the “training set”, the reference distribution data, are determined based on the ranges “f=z” set of features “
z
∈
f
”; wherein, in this instance, the sorting is only of input data and, therefore, the input and predictive data sorted is only input data).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the identifying of input data associated with a feature for use in drift detection of Xu in view of Dong with the identifying of predictive data associated with a feature for use in drift detection of Benson, in order to perform quality control over both machine learning input and output data, which will increase the effectiveness of the model by increasing the scope of distribution irregulates that can be identified to initiate the retraining process (Benson, Para. [0072], “the monitoring module 440 can monitor or detect a label drift of the model. In an embodiment, the label of predictions made by the machine learning model can be e.g., “healthy” or “unhealthy” for a unit. The monitoring module 440 can monitor the number of changes in e.g., the distribution of “unhealthy” prediction using e.g., statistical process control (SPC). It will be appreciated that SPC can be refer to a method of quality control that uses statistical methods to monitor and control a process . . . For example, the monitoring module 440 can monitor a statistical characteristic (e.g., mean, standard deviation, etc.) of the distribution of the “unhealthy” label predicted by the trained model during a predetermined period of time”).
Regarding Claim 18, the additional elements of the dependent claim are substantially the same as limitations of Claim 5, therefore it is rejected under the same rationale.
Response to Arguments
Applicant's arguments filed on January 29th, 2026 have been fully considered. Each argument is addressed in detail below.
I. Applicant argues the rejections to the claims, under 35 USC § 101, should be withdrawn
(Applicant’s Remarks, 1/29/2026, Pg. 7, Section “Response to Rejections under 35 U.S.C. § 101”).
Specifically, Applicant’s argues the claims, as amended, are directed to patentable subject matter.
Here, a detailed reasoning against the subject matter eligibility of the amended claims is provided above (see above, section “Claim Rejections - 35 USC § 101”). Additionally, Applicant has not made any specific arguments in favor of patentability or pointed out any specific errors in the reasoning relied upon in the 11/04/2025 Office Action, which may be relevant for a determination on the subject matter eligibility of the amended claims (see 37 C.F.R. 1.111(b), “In order to be entitled to reconsideration or further examination, . . . The reply by the applicant or patent owner must . . . specifically points out the supposed errors in the examiner’s action . . . A general allegation that the claims define a patentable invention without specifically pointing out how the language of the claims patentably distinguishes them from the references does not comply with the requirements of this section”).
As a result, the argument is not persuasive.
II. Applicant argues the rejections to the claims, under 35 USC § 103, should be withdrawn
(Applicant’s Remarks, 1/29/2026, Pg. 7-11, Section “Response to Rejections under 35 U.S.C. § 103(a)”).
1) First, Applicant argues “Xu's determining of a distance between a training set and a current batch to determine whether to retrain a model does not teach or suggest” the amended limitations of claim 1 (Pg. 8-9, Para. 6-1).
Specifically, Applicant argues “Xu is silent as to bin ranges determined based on a distribution of training data values for the feature . . . [and] sort[ing] at least one of input data or predictive data into bins corresponding to the bin ranges to generate current distribution data” (Pg. 9, Para. 2-4) (internal quotation marks omitted).
Additionally, Applicant points out that Xu fails to teach or suggest elements of the amended claims relating to recitations of “the edge device” and “substrate processing equipment” (Pg. 9, Para. 2-4).
According to MPEP 2111, “During patent examination, the pending claims must be given their broadest reasonable interpretation consistent with the specification” (internal quotation marks omitted) (see also Phillips v. AWH Corp., 415 F.3d 1303, 1316, 75 USPQ2d 1321, 1329 (Fed. Cir. 2005)).
Additionally, according to MPEP 2145, “One cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references” (see also In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981)).
Furthermore, according to 37 C.F.R. 1.111(b), “In order to be entitled to reconsideration or further examination, . . . The reply by the applicant or patent owner must . . . specifically points out the supposed errors in the examiner’s action . . . A general allegation that the claims define a patentable invention without specifically pointing out how the language of the claims patentably distinguishes them from the references does not comply with the requirements of this section”.
Here, as discussed in detail above, Xu discloses components of a method, which is within the broadest reasonable interpretation of bin ranges determined based on a distribution of training data values for the feature and sorting at least one of input data or predictive data into bins corresponding to the bin ranges to generate current distribution data (see above for details, section “Claim Rejections - 35 USC § 103”). Additionally, Applicant’s assertion that Xu is silent on these limitations does not specifically point out any supposed errors with this interpretation of Xu. Furthermore, while Applicant is correct that Xu fails to teach or suggest elements of the amended claims relating to recitations of “the edge device” and “substrate processing equipment”, Dong is relied upon as part of a combination with Xu to teach these elements. Therefore, arguments against Xu, individually, are not persuasive. Specifically, while Applicant asserts that each reference individually fails to disclose entire limitations of amended claim 1, Applicant has not specifically pointed out any shortcomings in the prior art of record to teach elements which they are actually relied upon to teach. Additionally, Applicant has not argued against the motivation to combine the prior art of record.
As a result, the argument is not persuasive.
2) Second, Applicant argues that Dong fails to remedy the alleged shortcomings of Xu because “Dong's use of an edge device to collect images of a solar panel to determine photovoltaic power generation production defects does not teach or suggest” the amended limitations of claim 1 (Pg. 9, Para. 5).
Specifically, Applicant argues “Dong is silent as to . . . an edge device associated with substrate processing equipment” (Pg. 10, Para. 3) (internal quotation marks omitted).
Additionally, Applicant points out that Dong fails to teach or suggest elements of the amended claims relating to recitations of “reference distribution data” and “current distribution data” (Pg. 10, Para. 2-5).
According to MPEP 2111, “During patent examination, the pending claims must be given their broadest reasonable interpretation consistent with the specification” (internal quotation marks omitted) (see also Phillips v. AWH Corp., 415 F.3d 1303, 1316, 75 USPQ2d 1321, 1329 (Fed. Cir. 2005)).
Additionally, according to MPEP 2145, “One cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references” (see also In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981)).
Furthermore, according to 37 C.F.R. 1.111(b), “In order to be entitled to reconsideration or further examination, . . . The reply by the applicant or patent owner must . . . specifically points out the supposed errors in the examiner’s action . . . A general allegation that the claims define a patentable invention without specifically pointing out how the language of the claims patentably distinguishes them from the references does not comply with the requirements of this section”.
Here, as discussed in detail above, Dong discloses components of a method, which is within the broadest reasonable interpretation of an edge device associated with substrate processing equipment (see above for details, section “Claim Rejections - 35 USC § 103”). Additionally, Applicant’s assertion that Dong is silent on these limitations does not specifically point out any supposed errors with this interpretation of Dong. Furthermore, while Applicant is correct that Dong fails to teach or suggest elements of the amended claims relating to recitations of “reference distribution data” and “current distribution data”, Xu is relied upon as part of a combination with Dong to teach these elements. Therefore, arguments against Dong, individually, are not persuasive. Specifically, while Applicant asserts that each reference individually fails to disclose entire limitations of amended claim 1, Applicant has not specifically pointed out any shortcomings in the prior art of record to teach elements which they are actually relied upon to teach. Additionally, Applicant has not argued against the motivation to combine the prior art of record.
As a result, the argument is not persuasive.
3) Third, Applicant argues amended claims 9 and 14 are patentable because each contains language similar to amended claim 1, which Applicant asserts is patentable. Based on this argument, Applicant further argues dependent claims 2-8, 10-13, and 15-20 are patentable because each depends upon and includes elements of one of the independent claims. However, as discussed in detail above, arguments in favor of the patentability of amended claim 1 are not persuasive.
As a result, the argument is not persuasive.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any 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 MATTHEW BRYCE GOLAN whose telephone number is (571)272-5159. The examiner can normally be reached Monday through Friday, 8:00 AM to 5:00 PM ET.
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, Alexey Shmatov can be reached at (571) 270-3428. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/MATTHEW BRYCE GOLAN/Examiner, Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123
Prosecution Insights