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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 1 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
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.
Claim(s) 1-7, 10-20 are rejected under 35 U.S.C. 103 as being unpatentable over Castiglione (US 20220114399 A1), in view of Zhou (US 20240104422 A1).
Regarding claim 1, Castiglione discloses “performing a gradient analysis of the inference model using [a] test data set to obtain a latent bias score for the test data set, the latent bias score indicating a relationship between input features of the test data set and an ability of the inference model to predict a bias feature;” (See [0024]; gradient analysis is performed to determine if the gradient of the model exceeds a predetermined threshold. A heat map is generated to show bias indicator values)
“making a determination regarding whether the latent bias score exceeds a latent bias score threshold;” (See [0024]; latent bias score threshold is exceeded if the predetermined threshold is exceeded by the gradient alignment of the model)
“in an instance of the determination in which the latent bias score exceeds the latent bias score threshold:” (See [0024]; latent bias score threshold has been determined to be exceeded)
“obtaining a new inference model using the updated training data set to replace the inference model” (See [0080]; after evaluating that a new inference model is needed, train a new inference model to replace the old model).
Castiglione fails to explicitly disclose, “obtaining a test data set, the test data set comprising at least a portion of the first training data set;”. Castiglione fails to further explicitly disclose, “removing the input features of the test data set from the first training data set to obtain an updated training data set;”.
Zhou teaches “obtaining a test data set, the test data set comprising at least a portion of the first training data set;” (See [0060]; test data is obtained by pruning a set of training data). Zhou further teaches “removing the input features of the test data set from the first training data set to obtain an updated training data set;” (See [0070]; a subset of input features is removed from the training data that are known to cause bias to update the training set).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention having Castiglione and Zhou before them to modify Castiglione to obtain the test data sets and to further remove input features from it to obtain an updated training data set. One would be motivated to do so in order to obtain a test data set without having to search for a new data set by pruning the first training data set of any input features that are known to cause bias, see e.g., [0070], where Zhou describes removing a subset of features from the training data that are known bias factors, as well as removing any other data features that are considered irrelevant to the desired criteria of the model, see e.g., [0060], where Zhou discloses removing a desired amount of data features according to various criteria for the purpose of improving the model’s accuracy.
Regarding claim 2, Castiglione fails to explicitly disclose “obtaining the test data set comprises: obtaining the first training data set, the first training data set comprising a listing of entries; for each entry of the listing of the entries: obtaining an input, an output, and a bias feature label; obtaining a subset of input features associated with the input; obtaining a new entry for the test data set, the new entry comprising: an input comprising the subset of the input features; the output; and the bias feature label; and adding the new entry to the test data set.”.
However, Zhou discloses “obtaining the test data set comprises: obtaining the first training data set, the first training data set comprising a listing of entries; for each entry of the listing of the entries: obtaining an input, an output, and a bias feature label;” (See [0024], [0025], [0060], [0064]; test data is obtained by pruning a set of training data using a dataset pruning component. Then, the dataset pruning component finds a set of features that are sufficient for accurately training a model over the training dataset. This includes the input features, the output, and the bias feature label.)
“obtaining a subset of input features associated with the input;” (See [0051]; a subset of input features (labels) is selected from the input)
“obtaining a new entry for the test data set, the new entry comprising: an input comprising the subset of the input features; the output; and the bias feature label; and adding the new entry to the test data set.” (See [0060], [0064]; This is similar to the previous reference and only adds "adding the new entry to the set").
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention having Castiglione and Zhou before them to modify Castiglione to by the implementation of obtaining the following data: the test data set and description of its contents, a subset of input features associated with the input, and a new entry for the test data set. One would be motivated to do so in order to specify what data the model requires and how this data can be selected for the purposes of accurately training the model, see e.g., [0064], where Zhou discloses that pruning a training data set to obtain a test data set with a specific set of features is done to accurately train and test a model.
Regarding claim 3, Castiglione discloses “performing the gradient analysis comprises: obtaining a listing of partial latent bias scores; and aggregating the partial latent bias scores of the listing of partial latent bias scores to obtain the latent bias score.” (See [0151], [0025]; Castiglione obtains a listing of partial latent bias scores by finding a first gradient, which is done by comparing a gradient of the machine learning model to a gradient of the auxiliary model to generate a fairness indicator value, obtaining a bias score as a result. Castiglione then generates aggregate info about the data to complete the gradient analysis.).
Regarding claim 4, Castiglione discloses “obtaining the listing of partial latent bias scores comprises: for each entry of the test data set: obtaining a first gradient, the first gradient indicating a relationship between the input and the ability of the inference model to predict the output; obtaining a second gradient, the second gradient indicating a relationship between the input and the ability of the inference model to predict the bias feature label; obtaining a first partial latent bias score using the first gradient and the second gradient;” (See [0151]; Castiglione finds a first and second gradient, and then obtains a bias score using the gradients.)
“and adding the first partial latent bias score to the listing of partial latent bias scores.” (See [0092]; Castiglione discloses storing latent or separate variables in a collection of data.).
Regarding claim 5, Castiglione discloses “a negative first gradient with a larger magnitude indicates a stronger relationship between the input and the ability of the inference model to faithfully predict the output.” (See [0143], [0396]; Castiglione shows that the gradient magnitude impacts performance, and having a large gradient as shown in chart (c) FIG. 15 shows that worse performance is exhibited. Conversely, a smaller or negative gradient exhibits a stronger performance. Castiglione configures the model to generate target predictions that predict performance data, which includes predicting the output (e.g, the target feature)).
Regarding claim 6, Castiglione discloses “a negative second gradient with a larger magnitude indicates a stronger relationship between the input and the ability of the inference model to faithfully predict the bias feature.” (See [0143], [0396]; Castiglione shows that the gradient magnitude impacts performance, and having a large gradient as shown in chart (c) FIG. 15 shows that worse performance is exhibited. Conversely, a smaller or negative gradient exhibits a stronger performance. Castiglione configures the model to generate target predictions that predict performance data, which includes predicting the bias feature, features that cause latent bias).
Regarding claim 7, Castiglione discloses “the latent bias score threshold is exceeded when the latent bias score is negative and has a magnitude that exceeds a magnitude indicated by the latent bias score threshold.” (See [0024]; Castiglione shows that the bias score threshold is exceeded when the gradient of the auxiliary model is more aligned with the gradient of the machine learning model).
Regarding claims 10 and 16, these claims are similar in scope to claim 1.
Regarding claims 11 and 17, these claims are similar in scope to claim 2.
Regarding claims 12 and 18, these claims are similar in scope to claim 3.
Regarding claims 13 and 19, these claims are similar in scope to claim 4.
Regarding claims 14 and 20, these claims are similar in scope to claim 5.
Regarding claim 15, this claim is similar in scope to claim 6.
Allowable Subject Matter
Claim 8 objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims. None of the prior art of recording, including Castiglione and Zhou, discloses an inference model comprising: a first inference path comprising a shared body and a first head, and a second inference path comprising a shared body and a second head. Further, a complete and thorough search did not uncover any other prior art. While the concept of dual headed inference models appears to be known, no art was found that disclosed a shared body for the models in combination with the other required limitations of the claim.
Claim 9 objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims. None of the prior art of recording, including Castiglione and Zhou, discloses a first inference path that is trained to generate inferences to predict the output and a second inference path that is trained to generate inferences to predict the bias feature. Further, a complete and thorough search did not uncover any other prior art. While the concept of generating inferences to predict the output or the bias feature appears to be known, no art was found that disclosed a shared body for the models in claim 8 in combination with the other required limitations of the claim.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DAVID KIM whose telephone number is (571)272-4331. The examiner can normally be reached 7:30 AM - 4:30 PM.
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, Matthew Ell can be reached at (571) 270-3264. 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.
/D.K./Examiner, Art Unit 2141
/MATTHEW ELL/Supervisory Patent Examiner, Art Unit 2141
Prosecution Insights