DETAILED ACTION
This action is in response to the amendment filed 28 October 2025.
Claims 1–21 are pending. Claims 1 and 13 are independent.
Claims 1–21 are rejected.
Notice of Pre-AIA or AIA Status
The present application, filed on or after 16 March 2013, is being examined under the first inventor to file provisions of the AIA .
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.
Response to Arguments
The objection to claims 12 and 20 is withdrawn in light of the amendment and accompanying arguments (remarks, p. 8).
Applicant's arguments filed 28 October 2025 have been fully considered but they are not persuasive.
Applicant argues that Rodriguez Lopez does not teach a reconstruction error (remarks, p. 10). The examiner respectfully disagrees. Rodriguez Lopez teaches that the adversarial loss function includes optimizing a binary cross-entropy [a type of reconstruction loss function] between a target value ỹ given input x, and the prediction f(x). Therefore, the rejection is maintained.
Applicant argues that Rodriguez Lopez does not teach “occurs since previously decreasing” as in claim 6 (remarks, p. 10), but does not provide further explanation. Therefore, the rejection is maintained.
Applicant further argues that Rodriguez Lopez does not teach “not limited by an unperturbed value” as in claim 8 (remarks, p. 11). However, Rodriguez Lopez teaches that each perturbation vector may comprise a number of elements, and each element may be a variation of a respective element in the initial vector, but that a variation may be equal to zero. Therefore, a perturbation of one element/feature is not limited by the perturbation of another element/feature. Therefore, the rejection is maintained.
Applicant further argues that Rodriguez Lopez does not teach initializing a perturbed value with a random or predefined value, as in claim 9 (remarks, p. 11). However, Rodriguez Lopez teaches that the modified vector [perturbed tuple] is based on a combination of an initial vector and perturbation vector, and the perturbation vector may be set by a predefined value. Therefore, the values of the modified vector are ultimately based on the predefined value. Therefore, the rejection is maintained.
Applicant further argues that Rodriguez Lopez does not teach “error of…not the perturbed feature” as in claim 10 (remarks, p. 11). However, Rodriguez Lopez teaches perturbing/varying multiple elements/features of the perturbation vector while keeping other elements of the perturbation vector at zero (i.e., not perturbing that element/feature) (Rodriguez Lopez, ¶¶ 77–80). Therefore, the rejection is maintained.
Applicant further argues for new claim 21 (remarks, p. 12). See the updated rejections below.
Claim Rejections—35 U.S.C. § 103
The following is a quotation of 35 U.S.C. § 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 C.F.R. § 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. § 102(b)(2)(C) for any potential 35 U.S.C. § 102(a)(2) prior art against the later invention.
Claims 1–11 and 13–19 are rejected under 35 U.S.C. § 103 as being unpatentable over Rodriguez Lopez et al. (US 2022/0130143 A1) [hereinafter Rodriguez Lopez] in view of Ravi et al. (“General Frameworks for Anomaly Detection Explainability: Comparative Study”) [hereinafter Ravi].
Regarding independent claim 1, Rodriguez Lopez teaches [a] method comprising: generating, from a […] tuple, [a] […] tuple that contains a perturbed value of a perturbed feature of a plurality of features; A latent vector [tuple] is combined with a perturbation vector to obtain a modified vector (Rodriguez Lopez, ¶¶ 13, 17, 56). modifying, in the […] tuple, the perturbed value of the perturbed feature to cause a change in reconstruction error for the anomalous tuple, wherein the change in reconstruction error comprises at least one selected from the group consisting of: The perturbation vectors are iteratively generated (Rodriguez Lopez, ¶ 80). The perturbation vector includes modifying, e.g., the value of one element and not the remaining element values (Rodriguez Lopez, ¶ 77). a decrease in reconstruction error of the perturbed feature, and The explainer searches for perturbations by minimizing loss functions [reconstruction error] including an adversarial loss function that optimizes [decreases] the binary cross-entropy between a target and prediction (Rodriguez Lopez, ¶¶ 111–114). an increase in a sum of reconstruction error of the plurality of features that are not the perturbed feature; automatically generating, after said modifying the perturbed value, an explanation that identifies an identified feature of the plurality of features as a cause of the anomalous tuple being anomalous; The explanation identifies sets of attributes [features] that have the most effect on the model’s output (Rodriguez Lopez, ¶ 100). detecting whether the identified feature of the explanation is the perturbed feature; The explanation includes a perturbed version of input data that emphasizes features that contributed to the model’s output (Rodriguez Lopez, ¶¶ 3, 5). wherein the method is performed by one or more computers. The method is performed by a computer or multiple computers (Rodriguez Lopez, ¶¶ 9, 13, 39, 47, 51, 52, 54, 57).
Rodriguez Lopez teaches explaining an autoencoder model, but does not expressly teach a model for detecting anomalies. However, Ravi teaches: [generating, from a] non-anomalous [tuple, an] anomalous [tuple that contains a perturbed value of a perturbed feature of a plurality of features] Methods of explaining autoencoder models, wherein the methods include perturbing normal [non-anomalous] samples/images to generate anomalous samples/images (Ravi, §§ 3.2, 3.4).
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to combine the teachings of Rodriguez Lopez with those of Ravi. Doing so would have been a matter of applying a known technique [explaining autoencoder models that classify the content of images] to a known [method] ready for improvement [explaining autoencoder models that classify data as anomalous or non-anomalous] to yield predictable results [using perturbation to generate anomalous data to explain a model].
Regarding dependent claim 2, the rejection of claim 1 is incorporated and Rodriguez Lopez/Ravi further teaches: wherein said modifying the perturbed value is based on maximizing a quantity that is not a sum of reconstruction error of the plurality of features. The perturbation vector may be obtained by maximizing the value of an element based on the semantic concept it represents (Rodriguez Lopez, ¶ 79).
Regarding dependent claim 3, the rejection of claim 1 is incorporated and Rodriguez Lopez/Ravi further teaches: wherein said modifying the perturbed value is partly based on minimizing reconstruction error of the perturbed feature. The explainer searches for perturbations by minimizing loss functions [reconstruction error] including an adversarial loss function that optimizes [minimizes] the binary cross-entropy between a target and prediction (Rodriguez Lopez, ¶¶ 111–114).
Regarding dependent claim 4, the rejection of claim 1 is incorporated and Rodriguez Lopez/Ravi further teaches: wherein: said modifying the perturbed value is repeated in multiple iterations; The perturbation vectors may be generated iteratively (Rodriguez Lopez, ¶ 80). said modifying the perturbed value in an iteration comprises adjusting the perturbed value based on a multiplicative product that is based on a scaling factor; The modified vectors [perturbed values] are generated by combining the latent vector with the perturbation vector (Rodriguez Lopez, ¶ 56). the method further comprises decreasing the scaling factor for at least one iteration. The steps are repeated to find the minimal value for the first element of the perturbation vector [i.e., the value is decreased in each repetition] (Rodriguez Lopez, ¶ 78).
Regarding dependent claim 5, the rejection of claim 4 is incorporated and Rodriguez Lopez/Ravi further teaches: wherein the decreasing the scaling factor is geometric. An element of the perturbation vector may be reduced by half, e.g., from 1 to 0.5 (Rodriguez Lopez, ¶ 77).
Regarding dependent claim 6, the rejection of claim 4 is incorporated and Rodriguez Lopez/Ravi further teaches: wherein said decreasing the scaling factor is conditioned on at least one selected from the group consisting of: a predefined count of iterations occurs since previously decreasing the scaling factor; The perturbations are iteratively optimized until, e.g., a maximum number of τ iterations is reached (Rodriguez Lopez, ¶ 133). the respective reconstruction error of the perturbed feature is not less than the respective reconstruction error of the perturbed feature was when previously decreasing the scaling factor; and a sum of respective reconstruction errors of the plurality of features that are not the perturbed feature is not less than a sum of respective reconstruction errors of the plurality of features that are not the perturbed feature when previously decreasing the scaling factor.
Regarding dependent claim 7, the rejection of claim 4 is incorporated and Rodriguez Lopez/Ravi further teaches: wherein the multiplicative product is based on a gradient. The perturbation vector is updated based on a gradient (Rodriguez Lopez, ¶¶ 17, 56, 66, 68, 72–74, 77).
Regarding dependent claim 8, the rejection of claim 1 is incorporated and Rodriguez Lopez/Ravi further teaches: wherein during said automatically generating the explanation, at least one selected from the group consisting of: the non-anomalous tuple and the anomalous tuple have identical values for the plurality of features that are not the perturbed feature, and a range of the perturbed value of the perturbed feature is not limited by an unperturbed value of the perturbed feature in the non-anomalous tuple. The perturbation vectors are generated based on a gradient [i.e., are not generated/limited by an unperturbed value] (Rodriguez Lopez, ¶ 56).
Regarding dependent claim 9, the rejection of claim 1 is incorporated and Rodriguez Lopez/Ravi further teaches: further comprising initializing the perturbed value with one selected from the group consisting of: a random value and a predefined value. The first perturbation vector may be set by the gradient [predefined value] (Rodriguez Lopez, ¶ 77).
Regarding dependent claim 10, the rejection of claim 1 is incorporated and Rodriguez Lopez/Ravi further teaches: wherein after said modifying the perturbed value, at least one selected from the group consisting of: the reconstruction error of the perturbed feature is less than each respective reconstruction error of the plurality of features that are not the perturbed feature, Generating the perturbation vector includes generating perturbation vectors for different elements/characteristics [features] of an image (Rodriguez Lopez, ¶¶ 67, 68, 71, 72, 74). The group of perturbation vectors that correspond to reconstructed vectors that have been assigned a different class, are identified [i.e., some perturbation vectors, for different features, may have lower reconstruction error than others] (Rodriguez Lopez, ¶ 76). the reconstruction error of the perturbed feature is less than at least one respective reconstruction error of the plurality of features that are not the perturbed feature, and the reconstruction error of the perturbed feature is less than an average respective reconstruction error of the plurality of features that are not the perturbed feature.
Regarding dependent claim 11, the rejection of claim 1 is incorporated and Rodriguez Lopez/Ravi further teaches: wherein: the explanation identifies multiple features of the plurality of features as a cause of the anomalous tuple being anomalous; The explanation may identify multiple attributes [features] that have the most effect on model output (Rodriguez Lopez, ¶¶ 100, 102). An autoencoder may be explained using counterfactuals (Ravi, § 3.4) or using feature importance methods like LIME or SHAP, which determine the attributes [features] that cause the model to make a decision (Ravi, §§ 3.2–3.3). the multiple features do not include the perturbed feature. The explanation may be generated by a feature importance method, e.g., LIME or SHAP, which determine the contribution of each feature by perturbation/replacement [i.e., the explanations do not necessarily include the perturbed feature(s)] (Ravi, §§ 3.2–3.3). [Applicant likewise discloses using LIME, SHAP, or Kernel SHAP, in para. 56 of the specification.]
Regarding independent claim 13, this claim recites limitations similar to those of claim 1, and is rejected for the same reasons.
Regarding dependent claim 14, this claim recites limitations similar to those of claim 2, and is rejected for the same reasons.
Regarding dependent claim 15, this claim recites limitations similar to those of claim 3, and is rejected for the same reasons.
Regarding dependent claim 16, this claim recites limitations similar to those of claim 4, and is rejected for the same reasons.
Regarding dependent claim 17, this claim recites limitations similar to those of claim 6, and is rejected for the same reasons.
Regarding dependent claim 18, this claim recites limitations similar to those of claim 8, and is rejected for the same reasons.
Regarding dependent claim 19, this claim recites limitations similar to those of claim 9, and is rejected for the same reasons.
Claims 12 and 20 are rejected under 35 U.S.C. § 103 as being unpatentable over Rodriguez Lopez et al. (US 2022/0130143 A1) [hereinafter Rodriguez Lopez] in view of Ravi et al. (“General Frameworks for Anomaly Detection Explainability: Comparative Study”) [hereinafter Ravi].
Regarding dependent claim 12, the rejection of claim 1 is incorporated. Rodriguez Lopez/Ravi teaches generating explanations that explain which features contribute to a model’s decision, but does not expressly teach allocating fractions to features using percentages. However, Roelofs teaches: wherein at least one selected from the group consisting of: the explanation allocates a fraction of attribution to a feature that is not the perturbed feature that is more than one selected from the group consisting of: fifty percent, seventy percent, and eighty percent; An explanation for an autoencoder model is generated; importances [attributions] of features are output as percentages, e.g., a feature importance of 80% [more than fifty or seventy percent] (Roelofs, pp. 5–6, § 4.3). the explanation allocates a fraction of attribution to the perturbed feature that is less than one selected from the group consisting of: twenty percent, ten percent, and five percent. An explanation for an autoencoder model is generated; importances [attributions] of features are output as percentages, e.g., a feature importance of 10%, 12%, or 15% [less than twenty percent] (Roelofs, pp. 5–6, § 4.3, pp. 6–7, §§ 5.3.1, 5.3.2).
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to combine the teachings of Rodriguez Lopez/Ravi with those of Roelofs. Doing so would have been a matter of simple substitution of one known element [outputting feature attribution as a number other than a percentage] for another [outputting feature attribution as a percentage] to obtain predictable results [a method of explaining an anomaly detection model wherein feature attributions are expressed as percentages].
Regarding dependent claim 20, this claim recites limitations similar to those of claim 12, and is rejected for the same reasons.
Claim 21 is rejected under 35 U.S.C. § 103 as being unpatentable over Rodriguez Lopez et al. (US 2022/0130143 A1) [hereinafter Rodriguez Lopez] in view of Ravi et al. (“General Frameworks for Anomaly Detection Explainability: Comparative Study”) [hereinafter Ravi], further in view of Takeishi (“Shapley Values of Reconstruction Errors of PCA for Explaining Anomaly Detection”).
Regarding dependent claim 21, Rodriguez Lopez/Ravi teaches a reconstruction error of an autoencoder model, but does not expressly teach that of a non-autoencoder model. However, Takeishi teaches: wherein said reconstruction error is a reconstruction error of a reconstructive machine learning model that is not an autoencoder. Shapley values are calculated based on reconstruction errors of PCA-based [principal component analysis] anomaly detection models (Takeishi, p. 795, § III).
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to combine the teachings of Rodriguez Lopez/Ravi with those of Takeishi. Doing so would have been a matter of simple substitution of one known element [an autoencoder-based model] for another [a PCA-based model] to obtain predictable results [an explanation system for explaining a PCA-based model].
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 C.F.R. § 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 C.F.R. § 1.17(a)) pursuant to 37 C.F.R. § 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.
The prior art made of record and not relied upon is considered pertinent to Applicant's disclosure.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Tyler Schallhorn whose telephone number is 571-270-3178. The examiner can normally be reached Monday through Friday, 8:30 a.m. to 6 p.m. (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, Tamara Kyle can be reached at 571-272-4241. 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 the USA or Canada) or 571-272-1000.
/Tyler Schallhorn/Examiner, Art Unit 2144
/TAMARA T KYLE/Supervisory Patent Examiner, Art Unit 2144