DETAILED ACTION
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 .
Status of Claims
The present application is being examined under the claims filed 11/13/2023.
Claims 1-20 are pending.
Information Disclosure Statement
The information disclosure statement filed 02/25/2025 fails to comply with 37 CFR 1.98(a)(3)(i) because it does not include a concise explanation of the relevance, as it is presently understood by the individual designated in 37 CFR 1.56(c) most knowledgeable about the content of the information, of each reference listed that is not in the English language. It has been placed in the application file, but the information referred to therein has not been considered. Particularly, Pareto “Manuele di economia politica con una introduzione alla scienza sociale” is both illegible and contains no English explanation of relevance.
Applicant's Information Disclosure Statements, filed on 02/25/2025 have been received, and entered into the record. However, it is impractical for the examiner to review the references thoroughly with the number of references cited in this case. By initializing each of the cited references on the accompanying 1449 forms, the examiner is merely acknowledging the submission of the cited references and merely indicating that only a cursory review has been made of the cited references. 4. MPEP § 2004.13 states: It is desirable to avoid the submission of long lists of documents if it can be avoided. Eliminate clearly irrelevant and marginally pertinent cumulative information. If a long list is submitted, highlight those documents which have been specifically brought to applicant's attention and/or are known to be of most significance. See Penn Yan Boats, Inc. v. Sea Lark Boats, Inc., 359 F. Supp. 948, 175 USPQ 260 (S.D. Fla. 1972), aft 'd, 479 F.2d 1338, 178 USPQ 577 (Sth Cir. 1973), cert. denied, 414 U.S. 874 (1974). But cf. Molins PLC v. Textron Inc., 48 F.3d 1172, 33 USPQ2d 1823 (Fed. Cir. 1995). 5. Further, it should be noted that an applicant's duty of disclosure of material and information is not satisfied by presenting a patent examiner with "a mountain of largely irrelevant material from which he is presumed to have been able, with his experience and with adequate time, to have found the critical [material]. It ignores the real world conditions under which examiners work." Rohm & Haas Co. v. Crystal Chemical co., 722 F.2d 1556, 1573 [220 USPQ 289] (Fed. Cir. 1983), cert. Denied, 469 U.S. 851 (1984). Patent applicant has a duty not just to disclose pertinent prior art references but to make a disclosure in such a way as not to "bury" it within other disclosures of less relevant prior art; see Golden Valley Microwave Foods Inc. v. Weaver Popcorn Co. Inc., 24 USPQ2d 180i (N~D. Ind. 1992); Molins PLC v. Textron Inc., 26 USPQ2d 1889, at 1899 (D.Del 1992); Penn Yan Boats, Inc. v. Sea Lark Boats, Inc. et al., 175 USPQ 260, at 272 (S.D. FI. 1972).
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 2-9 and 14-20 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.
Regarding Claims 2 and 15
Claim 2 recites the limitation “splitting, by at least one computing device during a preparatory step, the training instances of the training dataset into a plurality of sets, wherein each set comprises a first subset of training instances and second subset of training instances, such that each training instance is present in at least one of a second subset of the plurality of sets”. If each training instance is present in at least one of a second subset as claimed, it appears impossible to examiner to split the data into subsets achieve this result. Claim 15 is rejected for including similar language.
Regarding Claim 3 and 16
Claim 3 recites “using, by at least one computing device, the one or more protected-attribute variables (Z) as model inputs when training the machine-learning model with the one or more input variables (X) as model inputs and the each target variable (Y) as model output.” It is unclear whether Z is an input variable to machine learning or an output variable. Claim 16 is further rejected for including similar language.
Regarding Claim 4 and 17
Claim 4 recites “using, by at least one computing device, the one or more target variables (Y) as model inputs when training the machine-learning model with the one or more input variables (X) as model inputs and the each protected-attribute variable (Z) as model output.” It is unclear whether Y is an input variable to machine learning or an output variable. Claim 17 is further rejected for including similar language.
Regarding Claim 14 and 20
Claim 14 recites the terms “the training dataset” “each instance of the training dataset” “each target variable” “the one or more target variables” “the one or more input variables” “the one or more protected-attribute variables” and similar terms throughout the claim. There is insufficient antecedent basis for these terms in the claims. For purposes of examination, the examiner interprets the claim as if it reflected the language of claim 1. Examiner suggests amending the claim accordingly. Claim 20 is further rejected for including similar language.
Regarding Claims 15-18
Claims 15-18 are further rejected for including the same terms lacking antecedent basis that rendered claim 14 indefinite.
Claims
5-9 are dependent upon claim 2
15-19 are dependent upon claim 14
18-19 are dependent upon claim 15
and are therefore similarly rejected for including the deficiencies of claims 2, 14, and 15 respectively.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1, 3-4, 10, 14, 16-17, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Du et al. “Fairness in Deep Learning: a Computational Perspective” herein referred to as Du.
Regarding Claim 1
Du teaches:
A method for machine-learning fairness-aware data valuation processing for supervised learning,
(page 6 column 2 section 4.2) “The goal is to reduce representation bias while at the same
time preserve useful prediction properties of DNNs”; [*Examiner notes: Equation 2 of page 7 shows the supervised learning process.]
from a training dataset comprising a plurality of data records each containing data for a training instance (i) for the supervised learning, wherein the data for each training instance comprises one or more target variables (Y), one or more input variables (X) and one or more protected-attribute variables (Z), wherein fairness is defined as minimizing a data bias present in the training set in respect of the one or more protected variables (Z), the method comprising:
(page 7 section 4.2.2) “A predictor and an adversarial classifier are learned simultaneously. The goal of the predictor is to learn a high-level representation which is maximally informative for the major prediction task, while the role of adversarial classifier is to minimize the predictor’s ability to predict the protected attribute (Fig. 2(b)).”; (Page 7 Equation 2); (page 4 figure 2b)
PNG
media_image1.png
155
546
media_image1.png
Greyscale
PNG
media_image2.png
158
326
media_image2.png
Greyscale
for each target variable of the one or more target variables (Y), training, by at least one computing device, a machine-learning model, using the training dataset, with the one or more input variables (X) as model inputs and the each target variable (Y) as model output, for each protected-attribute variable of the one or more protected-attribute variables (Z),training, by at least one computing device, a machine-learning model, using the training dataset, with the one or more input variables (X) as model inputs and the each protected-attribute variable (Z) as model output;
(page 7 column 1 below equation 2) “where the adversarial classifier is to penalize the representation of h(x) if protected attribute z is predictable, parameter λ is used to negotiate the trade-off between maximizing utility and fairness. The training is iteratively performed between the main classifier f(x) and the adversarial classifier g(h(x)).”
obtaining, by at least one computing device, a prediction of each target variable (Y) and a prediction for each protected-attribute variable (Z), for each instance of the training dataset;
(page 6 column 2 paragraph 2) “We use gender and race as protected attribute for the two datasets respectively. Each dataset is split into 50% for training, 20% for validation and 30% for testing. The base DNN model is a multilayer perceptron (MLP) with 3 layers 3. We evaluate the metrics with the best performing model on validation set.”
obtaining, by at least one computing device, a performance entropy from the predictions of each the target variable (Y) and a protected-attribute entropy from the predictions of each the protected-attribute variable (Z); and outputting, by at least one computing device, the performance entropy and the protected-attribute entropy.
(page 7 column 1 last paragraph) “Some methods implement adversarial using general cross entropy loss [2], [34]”
PNG
media_image3.png
209
635
media_image3.png
Greyscale
Regarding Claim 3
Du teaches:
The method according to claim 1
(see rejection of claim 1)
further comprising: using, by at least one computing device, the one or more protected-attribute variables (Z) as model inputs when training the machine-learning model with the one or more input variables (X) as model inputs and the each target variable (Y) as model output.
(Equation 2)
PNG
media_image4.png
59
290
media_image4.png
Greyscale
Regarding Claim 4
Du teaches:
The method according to claim 1,
(see rejection of claim 1)
further comprising: using, by at least one computing device, the one or more target variables (Y) as model inputs when training the machine-learning model with the one or more input variables (X) as model inputs and the each protected-attribute variable (Z) as model output.
(Equation 2)
PNG
media_image4.png
59
290
media_image4.png
Greyscale
Regarding Claim 10
Du teaches:
The method according to claim 1
(see rejection of claim 1)
further comprising: obtaining, by at least one computing device, a utility metric, for each training instance, as a combination of the obtained performance entropy or entropies and of the obtained protected- attribute entropy or entropies.
PNG
media_image5.png
232
519
media_image5.png
Greyscale
Regarding Claim 14
Claim 14 is a computer-readable medium claim corresponding to method claim 1. The entirety of claim 14 is taught by the rejection of claim 1 and is thus rejected analogously.
Regarding Claim 16
Claim 16 is a computer-readable medium claim corresponding to method claim 3. The entirety of claim 16 is taught by the rejection of claim 3 and is thus rejected analogously.
Regarding Claim 17
Claim 17 is a computer-readable medium claim corresponding to method claim 4. The entirety of claim 14 is taught by the rejection of claim 4 and is thus rejected analogously.
Regarding Claim 20
Claim 20 is a computer-readable medium claim corresponding to method claim 1. The entirety of claim 20 is taught by the rejection of claim 1 and is thus rejected analogously.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 2, 6, 15, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Du in view of Nti et al. “Performance of Machine Learning Algorithms with Different K Values in K-fold Cross Validation” et al. herein referred to as Nti.
Regarding Claim 2
Du teaches:
The method according to claim 1
(see rejection of claim 1)
Du does not explicitly teach:
further comprising: splitting, by at least one computing device during a preparatory step, the training instances of the training dataset into a plurality of sets, wherein each set comprises a first subset of training instances and second subset of training instances, such that each training instance is present in at least one of a second subset of the plurality of sets;
when training the machine-learning models and obtaining the prediction, further comprising: for each set and for each target variable of the one or more target variables (Y): training, by at least one computing device using the first subset of the each set, a machine-learning model with the one or more input variables (X) as model inputs and the each target variable (Y) as model output; for each set and for each protected-attribute variable of the one or more protected- attribute variables (Z): training, by at least one computing device using the first subset of the each set, a machine-learning model with the one or more input variables (X) as model inputs and the each protected-attribute variable (Z) as model output
the method further comprising, when obtaining a prediction, the step of: obtaining, by at least one computing device, a prediction of each target variable (Y) and a prediction for each protected-attribute variable (Z), for each instance of the second subsets.
[*Examiner notes: Some of the limitations are taught by the combination of Du with Nti. See comments below.]
Nti teaches:
further comprising: splitting, by at least one computing device during a preparatory step, the training instances of the training dataset into a plurality of sets, wherein each set comprises a first subset of training instances and second subset of training instances, such that each training instance is present in at least one of a second subset of the plurality of sets;
(page 63 last paragraph) “In k-fold CV, the dataset is divided into k fold. A fold is used in each iteration once as testing data, while the remaining folds are used as training data [24]. Thus, the process is repetitive until all dataset is evaluated.”; (Figure 3)
PNG
media_image6.png
180
398
media_image6.png
Greyscale
when training the machine-learning models and obtaining the prediction, further comprising: for each set and for each target variable of the one or more target variables (Y): training, by at least one computing device using the first subset of the each set, a machine-learning model with the one or more input variables (X) as model inputs and the each target variable (Y) as model output; for each set and for each protected-attribute variable of the one or more protected- attribute variables (Z): training, by at least one computing device using the first subset of the each set, a machine-learning model with the one or more input variables (X) as model inputs and the each protected-attribute variable (Z) as model output
(page 63) “A fold is used in each iteration once as testing data, while the remaining folds are used as training data [24]”; [*Examiner notes: The limitation is met in combination when the k-fold cross validation taught by Nti is used for the machine learning taught by Du]
the method further comprising, when obtaining a prediction, the step of: obtaining, by at least one computing device, a prediction of each target variable (Y) and a prediction for each protected-attribute variable (Z), for each instance of the second subsets.
(page 63) “A fold is used in each iteration once as testing data, while the remaining folds are used as training data [24]”; [*Examiner notes: The limitation is met in combination when the k-fold cross validation taught by Nti is used for the machine learning taught by Du.]
Du, Nti, and the instant application are analogous because they are all directed to machine learning training.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the present invention to modify the fairness-aware machine learning of Du by using k-fold cross validation during training as taught by Nti because (Nti page 63 last paragraph) “In this study, different k values [3, 5, 7, 10, 15 and 20] were selected and compared to identify the optimal k value that improves accuracy in a classification task.”
Regarding Claim 6
Du in view of Nti teaches:
The method according to claim 2
(see rejection of claim 2)
And Nti further teaches:
wherein the splitting is overlapping between second subsets such that each training instance is present in one or more second subsets of the plurality of sets.
(page 63 last paragraph) “In k-fold CV, the dataset is divided into k fold. A fold is used in each iteration once as testing data, while the remaining folds are used as training data [24].”; [*Examiner notes: Observe that the subsets overlap in that e.g. fold 3 is used in the training data in both iteration 1 and iteration 2. Further, each training instance is included in one of the k folds, and each of the k folds is used in the testing data in at least one iteration.]
PNG
media_image7.png
207
568
media_image7.png
Greyscale
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the present invention to combine Du with Nti for the same reasons given in claim 2 above.
Regarding Claim 15
Claim 15 is a computer-readable medium claim corresponding to method claim 2. The entirety of claim 15 is taught by the rejection of claim 2 and is thus rejected analogously.
Regarding Claim 19
Du in view of Nti teaches:
The system according to claim 15
(see rejection of claim 15)
And Nti further teaches:
wherein the splitting is overlapping between second subsets such that one or more training instances are present in two or more second subsets of the plurality of sets.
(page 63 last paragraph) “In k-fold CV, the dataset is divided into k fold. A fold is used in each iteration once as testing data, while the remaining folds are used as training data [24].”; [*Examiner notes: Observe that the subsets overlap in that e.g. fold 3 is used in the training data in both iteration 1 and iteration 2. Further, each training instance is included in one of the k folds, and each of the k folds is used in the testing data in at least one iteration.]
PNG
media_image7.png
207
568
media_image7.png
Greyscale
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the present invention to combine Du with Nti for the same reasons given in claim 15 (see claim 2) above.
Claims 5, 7, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Du in view of Nti, and further in view of NPL reference “Cross-Validation (Statistics)” herein referred to as Wikipedia CV.
Regarding Claim 5
Du in view of Nti teaches:
The method according to claim 2
(see rejection of claim 2)
Du in view of Nti does not explicitly teach:
wherein the splitting is random and repeated until each training instance is present in at least one of second subsets, and further wherein each training instance is randomly allocated to one of the second subsets.
[*Examiner notes: While Nti does not explicitly teach the process claimed in claim 5, it is widely known as a part of how k-fold cross validation works. Evidence is supplied from Wikipedia CV as it existed before the effective filing date.]
However, Wikipedia CV teaches:
wherein the splitting is random and repeated until each training instance is present in at least one of second subsets, and further wherein each training instance is randomly allocated to one of the second subsets.
(page 5 of pdf section “k-fold cross-validation”) “In k-fold cross-validation, the original sample is randomly partitioned into k equal sized subsamples. Of the k subsamples, a single subsample is retained as the validation data for testing the model, and the remaining k − 1 subsamples are used as training data. The cross-validation process is then repeated k times, with each of the k subsamples used exactly once as the validation data. The k results can then be averaged to produce a single estimation.”
Du, Nti, Wikipedia CV, and the instant application are analogous because they are all directed to machine learning training.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the present invention to modify the fairness-aware machine learning of Du in view of Nti with the random splitting taught by Wikipedia CV because (Wikipedia CV page 5 of pdf section “k-fold cross-validation”) “The advantage of this method over repeated random sub-sampling (see below) is that all observations are used for both training and validation, and each observation is used for validation exactly once. 10-fold cross-validation is commonly used,[16] but in general k remains an unfixed parameter.”
Regarding Claim 7
Du in view of Nti:
The method according to any claim 2
(see rejection of claim 2)
Du in view of Nti does not explicitly teach:
further comprising: after splitting and before training, sampling, for each set:sampling, by at least one computing device without replacement within each set and with replacement across sets, training instances for subsequent training.
[*Examiner notes: While Nti does not explicitly teach the process claimed in claim 7, it is widely known as a part of how k-fold cross validation works. Evidence is supplied from Wikipedia as it existed before the effective filing date.]
However, Wikipedia CV teaches:
further comprising: after splitting and before training, sampling, for each set:sampling, by at least one computing device without replacement within each set and with replacement across sets, training instances for subsequent training.
[*Examiner notes: sampling without replacement is a term of the art that refers to sampling repeatedly from a population without ever choosing the same instance twice (i.e. )]; (page 5 of pdf section “k-fold cross-validation”) “In k-fold cross-validation, the original sample is randomly partitioned into k equal sized subsamples. Of the k subsamples, a single subsample is retained as the validation data for testing the model, and the remaining k − 1 subsamples are used as training data. The cross-validation process is then repeated k times, with each of the k subsamples used exactly once as the validation data[*Examiner notes: sampling without replacement]. The k results can then be averaged to produce a single estimation.”
Du, Nti, Wikipedia CV, and the instant application are analogous because they are all directed to machine learning training.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the present invention to modify the fairness-aware machine learning of Du in view of Nti with the random splitting taught by Wikipedia CV because (Wikipedia CV page 5 of pdf section “k-fold cross-validation”) “The advantage of this method over repeated random sub-sampling (see below) is that all observations are used for both training and validation, and each observation is used for validation exactly once. 10-fold cross-validation is commonly used,[16] but in general k remains an unfixed parameter.”
Regarding Claim 18
Claim 18 is a computer-readable medium claim corresponding to method claim 5. The entirety of claim 18 is taught by the rejection of claim 5 and is thus rejected analogously.
Claims 8-9 are rejected under 35 U.S.C. 103 as being unpatentable over Du in view of Nti, and further in view of NPL reference Lee et al. “Multi-Objective Instance Weighting-Based Deep Transfer Learning Network for Intelligent Fault Diagnosis” herein referred to as Lee.
Regarding Claim 8
Du in view of Nti teaches:
The method according to claim 2,
(see rejection of claim 2)
Du in view of Nti does not explicitly teach:
further comprising: obtaining, by at least one computing device, a central value statistic from the obtained performance entropy and the protected-attribute entropy over the trained models and over the second subsets of the each set; and outputting, by at least one computing device, the obtained central value statistic as an instance weighing for supervised training.
However, Lee teaches:
further comprising: obtaining, by at least one computing device, a central value statistic from the obtained performance entropy and the protected-attribute entropy over the trained models and over the second subsets of the each set; and outputting, by at least one computing device, the obtained central value statistic as an instance weighing for supervised training.
(page 11 paragraph 2) “Both KLD and MMD of certain source instance have positive values. These two indicators were standardized and converted to be used as weights 𝒲𝐾𝐿𝐷 𝑖 and 𝒲𝑀𝑀𝐷 𝑖, respectively. Instances with a high value in a specific indicator should have a low weight[*Examiner notes: instance weighting] value calculated from that indicator.”; (equation 9) [*Examiner notes: When the k-fold cross validation is combined with instance weighting as taught by Lee, the limitation is met]
PNG
media_image8.png
190
820
media_image8.png
Greyscale
Du, Nti, Lee, and the instant application are analogous because they are all directed to machine learning training.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the present invention to modify the fairness-aware machine learning of Du in view of Nti with the instance weighting approach as taught by Lee because (Lee page 19 conclusion) “The proposed method is based on the transfer learning and instance weighting strategy that complementarily utilize two indicators, KLD and MMD, to minimize discrepancy between the two domains used for the transfer learning model. Through this, the accuracy of target diagnosis is improved by using data with different conditions under a general situation where only a small number of data were collected from the target system condition.” That is, instance weighting improves accuracy by prioritizing data under different conditions on how well the instances adhere to those conditions.
Regarding Claim 9
Du in view of Nti and Lee teaches:
The method according to claim 8
(see rejection of claim 8)
And Lee further teaches:
wherein the obtaining of the central value statistic comprises obtaining an average of obtained entropy over the trained models and over the second subsets of the each set.
(equation 9)
PNG
media_image8.png
190
820
media_image8.png
Greyscale
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the present invention to combine Du and Nti with Lee for the same reasons given in claim 8 above.
Claim 11 is rejected under 35 U.S.C. 103 as being unpatentable over Du in view of NPL reference “Multi-objective optimization” herein referred to as Wikipedia MOO.
Regarding Claim 11
[*Examiner notes: Claims 11 and 12 involve complex equations, as do the prior art and as such screenshots of the claims and prior art are used to show teaching of the claims]
Du teaches the method of claim 10. Du does not explicitly teach the remaining limitations of the claim. However, Wikipedia MOO teaches the remaining limitations of the claim:
PNG
media_image9.png
107
653
media_image9.png
Greyscale
PNG
media_image10.png
345
669
media_image10.png
Greyscale
(pdf page 8-9)
PNG
media_image11.png
142
633
media_image11.png
Greyscale
Du, Wikipedia MOO, and the instant application are analogous because they are all directed to machine learning training.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the present invention to modify the fairness-aware machine learning of Du by using the linear scalarization as taught by Wikipedia MOO because (Wikipedia MOO page 8) “Scalarizing a multi-objective optimization problem is an a priori method, which means formulating a single-objective optimization problem such that optimal solutions to the single-objective optimization problem are Pareto optimal solutions to the multi-objective optimization problem.[2] In addition, it is often required that every Pareto optimal solution can be reached with some parameters of the scalarization.[2] With different parameters for the scalarization, different Pareto optimal solutions are produced.”
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over Du in view of NPL reference Deng et al. “A Unified Energy Efficiency and Spectral Efficiency Tradeoff Metric in Wireless Networks” herein referred to as Deng.
Regarding Claim 12
Du teaches the method of claim 10. Du does not explicitly teach the remaining limitations of the claim. However, Deng teaches the remaining limitations of the claim:
PNG
media_image12.png
246
668
media_image12.png
Greyscale
(page 56 column 2 above equation 10)
PNG
media_image13.png
223
412
media_image13.png
Greyscale
Du, Deng, and the instant application are analogous because they are all directed to machine learning training and/or optimization.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the present invention to modify the fairness-aware machine learning of Du by using the multiplicative scalarization as taught by Deng because (Deng page 55 abstract) “We further show that the objective function of this SOO problem, i.e., our proposed EE and SE tradeoff (EST) metric, is quasiconcave with the transmit power and a unique globally optimal solution is derived. Numerical results validate the effectiveness of the proposed unified EST metric.”
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Du in view of Lee.
Regarding Claim 13
Du teaches:
The method according to claim 1
(see rejection of claim 1)
Du does not explicitly teach:
further comprising carrying out supervised learning, by at least one computing device, using a utility metric obtained from a combination of the obtained performance entropy or entropies and of the obtained protected-attribute entropy or entropies, for each training instance, as instance weighing.
However, Lee teaches:
further comprising carrying out supervised learning, by at least one computing device, using a utility metric obtained from a combination of the obtained performance entropy or entropies and of the obtained protected-attribute entropy or entropies, for each training instance, as instance weighing.
(page 11 paragraph 2) “Both KLD and MMD of certain source instance have positive values. These two indicators were standardized and converted to be used as weights 𝒲𝐾𝐿𝐷 𝑖 and 𝒲𝑀𝑀𝐷 𝑖, respectively. Instances with a high value in a specific indicator should have a low weight[*Examiner notes: instance weighting] value calculated from that indicator.”; (equation 9) [*Examiner notes: Combining the learning metrics of Du with the combined metric as instance weighting meets the limitation in combination]
PNG
media_image8.png
190
820
media_image8.png
Greyscale
Du, Lee, and the instant application are analogous because they are all directed to machine learning training.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the present invention to modify the fairness-aware machine learning of Du with the instance weighting approach as taught by Lee because (Lee page 19 conclusion) “The proposed method is based on the transfer learning and instance weighting strategy that complementarily utilize two indicators, KLD and MMD, to minimize discrepancy between the two domains used for the transfer learning model. Through this, the accuracy of target diagnosis is improved by using data with different conditions under a general situation where only a small number of data were collected from the target system condition.” That is, instance weighting improves accuracy by prioritizing data under different conditions on how well the instances adhere to those conditions.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure: NPL reference Kamani et al. "Pareto Efficient Fairness in Supervised Learning: From Extraction to Tracing" teaches multi-objective optimization in a fairness context.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Ezra J Baker whose telephone number is (703)756-1087. The examiner can normally be reached Monday - Friday 10:00 am - 8:00 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, David Yi can be reached at (571) 270-7519. 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.
/E.J.B./Examiner, Art Unit 2126
/DAVID YI/Supervisory Patent Examiner, Art Unit 2126
Prosecution Insights