CTNF 18/241,450 CTNF 89305 Notice of Pre-AIA or AIA Status 07-03-aia AIA 15-10-aia 1. 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 07-04-01 AIA 07-04 2. 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. 3 . Claims 1-8 and 10-14 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. Taking independent claim 1 as representative, the claims recite, in part: A method comprising: estimating an empirical entropy of a training set on which a generative model has been trained ; generating a typicality score for the distance between 1) an augmented training element of a training element included in the training set and 2) a typicality set of the trained generative model, wherein the typicality score is based on the estimated empirical entropy ; and comparing the typicality score to a threshold to determine whether the augmented training element is suitable for inclusion into the training set . The claim is directed to a judicial exception, and specifically an abstract idea. For purposes of analyzing the claim, the Examiner has bolded the portions in the claim above that are akin to mental steps and underlined the portions that might be considered additional elements. Focusing on the mental steps found in the claim, the steps for estimating an entropy, generating a score, and comparing the score with a threshold are generally mental steps, evaluations, judgments, or even just mathematical computations/expressions. Hence, the claim, based on the bolded features, is largely based on an abstract idea. The subject of the estimation, e.g., a training set on which a generative model has been trained, is indicated by the Examiner as an additional element. The Examiner notes that the model related to the training set / subject of estimation is not an active participant in the claim. It is merely enough that it exists and that someone can evaluate it, e.g. provide an estimation of its character (such as entropy of the data used to train it). Hence, this additional element language is not sufficient to ground the abstract idea into a practical integration, because again the claim only seems to require the existence of data used to train a model, and evaluating/estimating a character of that data is not persuasive to the involved mental step to be integrated into a practical application. Nor does the Examiner reason that the additional element discussed above is sufficient to amount to significantly more than the judicial exception. For example, the training of the model is outside the active scope of the claim. Regarding the model itself, the claim only involves the evaluation of an aspect of the model, e.g. its training data entropy, and with the model being passively involved in the claim’s active mental steps. Independent claim 10 is essentially directed to the same limitations as claim 1 discussed above, and is therefore likewise rejected under the same rationale as being directed to an abstract idea that is not practically integrated and not providing significantly more. Independent claim 11 is essentially a reorganization of the steps of claims 1 and 10, but with an iterative structure added so as to apply those same steps many times for many elements. Even with this differentiable understanding, the Examiner believes the same reasoning discussed per claim 1 for example would likewise apply for claim 11. The dependent claims 2-8 and 12-14 are likewise rejected for inheriting the same deficiencies discussed above per claim 1 for example without otherwise curing them. The Examiner will address some of these dependent claims here: Claim 2 clarifies a type of the model, which as noted is only passively involved in the active mental steps of the independent claim. Hence, this clarification is not persuasive to making those mental steps as discussed per claim 1 any more practically integrated or as providing significantly more. Claims 3-5 and 7 clarify mathematical concepts involved in claim 1’s mental steps, but do not otherwise make the mathematical feature any less abstract. Claim 6 is merely an additional mental step, and hence its scope is rejected under the same rationale given above per claim 1. Claim 8 involved an action to include a training element into a training set as a result of an evaluation, judgement, etc. Hence, it understood by the Examiner as extra-solution activity that does not otherwise make the recitations of claim 1 any less abstract. The Examiner notes that further detail that actively does something with this expanded training set could be understood to overcome the Examiner’s 101 rationale, provided it related to an improvement in the functioning of a computer, a field of invention, etc. Claims 12-14 feature the same limitations as claims 4-5 and 7 discussed above, and are therefore rejected under the same reasoning. Claim Rejections - 35 USC § 112 4. The following is a quotation of 35 U.S.C. 112(b): (b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention. 5. Claims 5 and 12-3 are rejected under 35 U.S.C. 112(b) as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor regards as the invention. Claims 5 and 12 both feature a recitation for the estimated empirical energy Ĥ p , for which the Examiner believes there is improper antecedent basis. It has the effect of rendering the claims vague and indefinite. Claim 13 depends from claim 12, and does not otherwise cure this deficiency; hence, it too is rejected under the same rationale. The Examiner notes that claims 4 and 11, from which claims 5 and 12 depend, recite a similar feature but with a different name: “an estimated empirical entropy .” The Examiner speculates this is the term intended by Applicants for express recitation in claims 5 and 12. Claim Rejections - 35 USC § 103 07-06 AIA 15-10-15 6. 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. 07-20-aia AIA 7. 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. 07-23-aia AIA 8. The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows: 1. Determining the scope and contents of the prior art. 2. Ascertaining the differences between the prior art and the claims at issue. 3. Resolving the level of ordinary skill in the pertinent art. 4. Considering objective evidence present in the application indicating obviousness or nonobviousness. 07-21-aia AIA 9. Claim s 1-16 are rejected under 35 U.S.C. 103 as being unpatentable over Non-Patent Literature “Detecting Out-of-Distribution Inputs to Deep Generative Models Using Typicality” (“ Nalisnick ”) in view of U.S. Patent Application Publication No. 2024/0394527 (“ Boiarov ”) . Regarding claim 1 , NALISNICK teaches A method comprising: estimating an empirical entropy of a training set on which a generative model has been trained (section 3.3 on page 4, discussing an entropy estimator for deep generative models, where the entropy as determined is based on sampling the data for the model (where the data as referenced here relates to the training data, as the Examiner understands based on a reading of the Abstract and Introduction sections)) ; generating a typicality score for the distance between 1) an ... training element of a training element included in the training set and 2) a typicality set of the trained generative model, wherein the typicality score is based on the estimated empirical entropy and comparing the typicality score to a threshold to determine whether the ... training element is suitable for inclusion into the training set (staying with section 3.3, at the bottom of page 4 through top of page 5, discussing a threshold used to determine out-of-distribution for a particular input, such that if the input is rejected under this analysis it is deemed to not reside in the typical set, from which the Examiner reasons the analysis examines the gap between the typical set and the input (i.e., the gap is akin to a typicality score as recited), and where this portion references Algorithm 1 in the reference’s Appendix – which the Examiner understands to be a distance computation as made clear in both Appendix A’s A.1 and A.2 sections and also the if-conditional at the bottom of Algorithm 1 (top of page 13)) . Nalisnick does not teach that the training element, as discussed above, is an augmented training element. Rather, the Examiner relies upon BOIAROV to teach what Nalisnick lacks , see e.g., Boiarov’s comparable framework of detecting outliers in machine learning training ([0001], [0036], [0039]) that is similarly threshold and distance-based ([0031]-[0032], [0039]), and the Examiner particularly notes the involvement of this taught approach as part of a data augmentation aspect ([0035]-[0036]) as would be used to improve the machine learning by way of improving the training data, in part by discarding outliers among the training data and/or augmented data to be added thereto. Both references relate to generative modeling, the improving of training therefor, and related aspects. Hence, they are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to extend Nalisnick’s analysis to determine OOD instances among training data, as the aforementioned reference contemplates, into scenarios that also concern augmented data added to training data, with a reasonable expectation of success, to improve the training data and hence the relevant machine learning model trained using that data. Regarding claim 2 , Nalisnick in view of Boiarov teach the method of claim 1, as discussed above. The aforementioned references teach the additional limitations wherein the trained generative model is a normalizing flow, a diffusion model, or a variational autoencoder (Nalisnick’s section 3.1 on page 3, discussing that the DGM may be a normalizing flow or a VAE). The motivation for combining the references is as discussed above in relation to claim 1. Regarding claim 3 , Nalisnick in view of Boiarov teach the method of claim 1, as discussed above. The aforementioned references teach the additional limitations wherein the typicality score is an estimate of the distance between the log-density of the trained generative model, wherein x is the augmented training element (Nalisnick’s section 3.3, equations 4-6, and specifically, at the bottom of page 4 through top of page 5, discussing a threshold used to determine out-of-distribution for a particular input, such that if the input is rejected under this analysis it is deemed to not reside in the typical set, from which the Examiner reasons the analysis examines the gap between the typical set and the input (i.e., the gap is akin to a typicality score as recited), and where this portion references Algorithm 1 in the reference’s Appendix – which the Examiner understands to be a distance computation as made clear in both Appendix A’s A.1 and A.2 sections and also the if-conditional at the bottom of Algorithm 1 (top of page 13), and where the Examiner reasons such a distance based determination per Nalisnick could be extended to determining OOD not just for existing training data but new augmented training data as Boiarov contemplates the use of to improve training for the machine learning model). The motivation for combining the references is as discussed above in relation to claim 1. Regarding claim 4 , Nalisnick in view of Boiarov teach the method of claim 1, as discussed above. The aforementioned references teach the additional limitations wherein the typicality score is an estimate according to TS(x)= ❘ -logp (x;θ) - Hˆp ❘ , where TS(x) is the typicality score of sample x, p(x; θ) is the density of sample x with respect to the trained generative model having parameters θ, and Ĥp is an estimated empirical entropy of the trained generative model (Nalisnick’s section 3.3 as discussed above in relation to claim 1, and specifically equations 4-6). The motivation for combining the references is as discussed above in relation to claim 1. Regarding claim 5 , Nalisnick in view of Boiarov teach the method of claim 4, as discussed above. The aforementioned references teach the additional limitations wherein the estimated empirical energy Ĥp is computed over the training set as Hˆp = Ex ∼ Dtrain [-logp (x;θ)], where Ex˜D train is the expected value of the samples x taken from the training set Dtrain (Nalisnick’s section 3.3 as discussed above in relation to claim 1, and specifically equations 4-6). The motivation for combining the references is as discussed above in relation to claim 1. Regarding claim 6 , Nalisnick in view of Boiarov teach the method of claim 1, as discussed above. The aforementioned references teach the additional limitations for determining the augmented training sample is suitable for inclusion into the training set when the typicality score of the augmented training sample is less than or equal to the threshold (as discussed per claim 1, Nalisnick: section 3.3, at the bottom of page 4 through top of page 5, discussing a threshold used to determine out-of-distribution for a particular input, such that if the input is rejected under this analysis it is deemed to not reside in the typical set, from which the Examiner reasons the analysis examines the gap between the typical set and the input (i.e., the gap is akin to a typicality score as recited), and where this portion references Algorithm 1 in the reference’s Appendix – which the Examiner understands to be a distance computation as made clear in both Appendix A’s A.1 and A.2 sections and also the if-conditional at the bottom of Algorithm 1 (top of page 13)). The motivation for combining the references is as discussed above in relation to claim 1. Regarding claim 7 , Nalisnick in view of Boiarov teach the method of claim 1, as discussed above. The aforementioned references teach the additional limitations wherein the threshold is estimated from samples in the training set as Α = max x ∈ Dtrain TS(x)+ϵ, where Dtrain is the training set, x is a sample in the training set, and ϵ is a tunable parameter (as discussed per claim 1, Nalisnick: section 3.3, at the bottom of page 4 through top of page 5, discussing a threshold used to determine out-of-distribution for a particular input, such that if the input is rejected under this analysis it is deemed to not reside in the typical set, from which the Examiner reasons the analysis examines the gap between the typical set and the input (i.e., the gap is akin to a typicality score as recited), and where this portion references Algorithm 1 in the reference’s Appendix – which the Examiner understands to be a distance computation as made clear in both Appendix A’s A.1 and A.2 sections and also the if-conditional at the bottom of Algorithm 1 (top of page 13)). The motivation for combining the references is as discussed above in relation to claim 1. Regarding claim 8 , Nalisnick in view of Boiarov teach the method of claim 1, as discussed above. The aforementioned references teach the additional limitations for including the augmented training element into the training set when the augmented training element is determined to be suitable for inclusion into the training set (as discussed per claim 1, Nalisnick: section 3.3, at the bottom of page 4 through top of page 5, discussing a threshold used to determine out-of-distribution for a particular input, such that if the input is rejected under this analysis it is deemed to not reside in the typical set, from which the Examiner reasons the analysis examines the gap between the typical set and the input (i.e., the gap is akin to a typicality score as recited), and where this portion references Algorithm 1 in the reference’s Appendix – which the Examiner understands to be a distance computation as made clear in both Appendix A’s A.1 and A.2 sections and also the if-conditional at the bottom of Algorithm 1 (top of page 13), and where the understanding is that such augmentation as Boiarov more explicitly considers is in aim of improving training the model). The motivation for combining the references is as discussed above in relation to claim 1. Regarding claim 9 , Nalisnick in view of Boiarov teach the method of claim 8, as discussed above. The aforementioned references teach the additional limitations for training a machine learning model on the training set including the augmented training element (as discussed per claim 1, Nalisnick: section 3.3, at the bottom of page 4 through top of page 5, discussing a threshold used to determine out-of-distribution for a particular input, such that if the input is rejected under this analysis it is deemed to not reside in the typical set, from which the Examiner reasons the analysis examines the gap between the typical set and the input (i.e., the gap is akin to a typicality score as recited), and where this portion references Algorithm 1 in the reference’s Appendix – which the Examiner understands to be a distance computation as made clear in both Appendix A’s A.1 and A.2 sections and also the if-conditional at the bottom of Algorithm 1 (top of page 13), and where the understanding is that such augmentation as Boiarov more explicitly considers is in aim of improving training the model). The motivation for combining the references is as discussed above in relation to claim 1. Regarding claim 10 , the claim includes the same or similar limitations as claim 1 discussed above, and is therefore rejected under the same rationale. The claim additionally recites a non-transitory memory including processor-executable instructions which the Examiner believes are taught per Nalisnick’s computer-based experimentation of its taught method (section 5 on page 6) and/or Boiarov’s system which is clearly a general purpose computer per [0043] and would be understood to feature such a memory aspect as is recited. Regarding claim 11 , the claim includes the same or similar limitations as claim 1 discussed above, and is therefore rejected under the same rationale. The claim additionally recites a non-transitory memory including processor-executable instructions and one or more processors, which the Examiner believes are taught per Nalisnick’s computer-based experimentation of its taught method (section 5 on page 6) and/or Boiarov’s system which is clearly a general purpose computer per [0043] and would be understood to feature such a memory aspect as is recited. The present claim organizes the same steps of claim 1 but in a more iterative structure for executing across more than one instance. However, that merely involves more repetitions of the same steps and hence the same teachings still apply. Regarding claim 12 , the claim includes the same or similar limitations as claim 5 discussed above, and is therefore rejected under the same rationale. Regarding claim 13 , the claim includes the same or similar limitations as claim 4 discussed above, and is therefore rejected under the same rationale. Regarding claim 14 , the claim includes the same or similar limitations as claim 7 discussed above, and is therefore rejected under the same rationale. Regarding claim 15 , the claim includes the same or similar limitations as claim 9 discussed above, and is therefore rejected under the same rationale. Regarding claim 16 , Nalisnick in view of Boiarov teach the system of claim 15, as discussed above. The aforementioned references teach the additional limitation that different models may be subject to the taught framework, which the Examiner believes Nalisnick’s computer-based experimentation of its taught method (section 5 on page 6) reads on. Conclusion 07-96 10. The prior art made of record and not relied upon is considered pertinent to Applicants’ disclosure: US 20230119132 A1 US 20240394528 A1 US 12579786 B2 US 12229845 B2 US 11967097 B2 Non-Patent Literature “Boosting Out-of-distribution Detection with Typical Features” 11. Any inquiry concerning this communication or earlier communications from the examiner should be directed to SHOURJO DASGUPTA whose telephone number is (571) 272-7207. The examiner can normally be reached M-F 8am-5pm CST. 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 USA OR CANADA) or 571-272-1000. /SHOURJO DASGUPTA/Primary Examiner, Art Unit 2144 Application/Control Number: 18/241,450 Page 2 Art Unit: 2144 Application/Control Number: 18/241,450 Page 3 Art Unit: 2144 Application/Control Number: 18/241,450 Page 4 Art Unit: 2144 Application/Control Number: 18/241,450 Page 5 Art Unit: 2144 Application/Control Number: 18/241,450 Page 6 Art Unit: 2144 Application/Control Number: 18/241,450 Page 7 Art Unit: 2144 Application/Control Number: 18/241,450 Page 8 Art Unit: 2144 Application/Control Number: 18/241,450 Page 9 Art Unit: 2144 Application/Control Number: 18/241,450 Page 10 Art Unit: 2144 Application/Control Number: 18/241,450 Page 11 Art Unit: 2144 Application/Control Number: 18/241,450 Page 12 Art Unit: 2144 Application/Control Number: 18/241,450 Page 13 Art Unit: 2144
Prosecution Insights