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 .
Election/Restrictions
Applicant’s election without traverse of Group I (claims 1-15) in the reply filed on 01/09/2026 is acknowledged.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1, 6-9, and 14-15 are rejected under 35 U.S.C. 103 as being unpatentable over Watson (US 20190354850) in view of Betthauser (“Discovering Distribution Shifts using Latent Space Representations” Feb, 17, 2022) and Watson2 (US 20200082210)..
Regarding claim 1:
Watson discloses: a method for measuring similarities between sets of feature vectors in a machine learning application (FIG. 11), comprising:
generating a feature representation from existing sources of datasets (¶ [0094] discloses generating respective representation characterizing one or more source data sets, and see ¶ [0051] – ¶ [0053]);
generating representation of metric distances of the existing sources to each other (¶ [0095] discloses using one or more distance computation techniques regarding the first vector representation and/or the second vector representation to assess one or more similarity metrics between the one or more source data sets and/or the one or more sample data sets, and ¶ [0054] discloses the distance computing techniques) ;
generating feature representation of a target dataset (¶ [0094] discloses generating respective representation characterizing one or more sample data sets, and see ¶ [0051] – ¶ [0053]); and
generating representation of the metric distances of each of the target dataset (¶ [0095] discloses using one or more distance computation techniques regarding the first vector representation and/or the second vector representation to assess one or more similarity metrics between the one or more source data sets and/or the one or more sample data sets, ¶ [0054] discloses the distance computing techniques, and ¶ [0051] discloses the equation representing the distance D on aggregated feature representation of target data set ti and source data set sj..”. ¶ [0042] “the assessment component 112 can assess one or more similarity metrics between the one or more target feature representations and the one or more source feature representations …the assessment component 112 can utilize one or more distance computation techniques to assess the similarity and/or dissimilarity between the one or more target feature representations and/or the one or more source feature representations… Thus, the one or more similarity metrics can indicate how similar and/or dissimilar the one or more sample data sets, and thereby the target data sets, are from the one or more source data sets. “);
Watson does not use energy distance, it lists examples such as KL-divergence, L2 distance, cosine similarity, Manhattan distance, Minkowski distance, Jaccard similarity, chi-square distance, a combination thereof, and/or the like. And Watson does not teach: wherein the energy distance measure comprises a statistical distance between probability distributions.
However, in a related field, Betthauser teaches the use of energy distance (page 3, “3 approach”; “We assume we are given two sets of embeddings, X and Y of dimension d, generated by the same model based on two datasets”; page 4, “3.2 Distribution Shift Tests” “we compute pairwise distances between reference subsamples to estimate the range of distances to be considered “in-distribution”, page 5, Algorithm 2 and “3.3 Distance Metrics”; “Energy Distance is a nonparametric distribution distance metric [17], and is an instance of Maximum Mean Discrepancy with kernel equal to the negative Euclidean norm between elements. The function requires no assumptions or explicit functional form, and can be easily evaluated over collections of high-dimensional data, such as two sets of embeddings, X and Y”);
wherein the energy distance measure comprises a statistical distance between probability distributions (page 5, sec. “3.3 distance metrics”, “Energy Distance is a nonparametric distribution distance metric” that can be evaluated over two high-dimensional embeddings sets X and Y. A nonparametric distribution distance metric is a statistical measure of distance between the probability distributions represented by X and Y).
Therefore, it would have been obvious to a person of ordinary skill in the art prior to the effective filing date of the claimed invention to have modified Watson to incorporate the teachings of Betthauser by including: using energy distance in order to obtain an alternative way of quantifying similarity between source and target datasets.
Watson in view of Betthauser does not specifically teach: generating pseudo-labels for the target dataset based on the metric distances; and training a machine learning model using the generated pseudo-labels.
However, in the same field of endeavor, Watson2 teaches: generating pseudo-labels for the target dataset based on the metric distances (¶¶ [0003] “The computer-implemented method assigns pseudo-labels to unlabeled examples of data using a similarity metric on an embedding space to produce pseudo-labeled examples”, [0017] “A pseudo-label is a listing of a similarities of data elements to different data elements (e.g., a relative label or relative name). Pseudo-label are automatically generated based on a distancing metric applied to unlabeled data elements of the data set to determine a distance, which itself is used as a pseudo-label of these unlabeled data items”, and [0055] – [0058]); and
training a machine learning model using the generated pseudo-labels (¶¶ [0003] “A curriculum learning model is trained using the pseudo-labeled examples. The curriculum learning model trained with the pseudo-labeled examples is employed in a fine-tuning task to enhance classification accuracy of the data.”, and [0060] – [0061]).
Therefore, it would have been obvious to a person of ordinary skill in the art prior to the effective filing date of the claimed invention to have modified Watson in view of Betthauser to incorporate the teachings of Watson2 in order to produce pseudo-labeled target examples from the computed energy-distance based similarity relationship and train a machine learning model using these labels by applying a predictable implementation.
Regarding claim 6:
Watson2 further discloses: wherein the target data set comprises images (¶ [0055] “Forward prediction comprises an image feature extraction where the pre-trained model penultimate layer's outputs are feature vectors, from which each category's average feature vectors is computed as a category feature representation. Note that the category feature representation for both source and target data sets are computed. This is repeated for other target layers.” And see ¶ [0059]).
Regarding claim 7:
Watson in view of Betthauser and Watson2 teaches the limitations of claim 1 as applied above.
Betthauser teaches the use of energy distance (page 3, “3 approach”; “We assume we are given two sets of embeddings, X and Y of dimension d, generated by the same model based on two datasets”; page 4, “3.2 Distribution Shift Tests” “we compute pairwise distances between reference subsamples to estimate the range of distances to be considered “in-distribution”, page 5, Algorithm 2 and “3.3 Distance Metrics”; “Energy Distance is a nonparametric distribution distance metric [17], and is an instance of Maximum Mean Discrepancy with kernel equal to the negative Euclidean norm between elements. The function requires no assumptions or explicit functional form, and can be easily evaluated over collections of high-dimensional data, such as two sets of embeddings, X and Y”).
Watson in view of Betthauser and Kender does not specifically teach: further comprising outputting pseudo labeling.
Watson2 teaches: further comprising outputting pseudo labeling (abstract, “…assigns pseudo-labels to unlabeled examples of data using a similarity metric on an embedding space to produce pseudo-labeled examples”; ¶ [0058] “At block 620, pseudo labels generated and assigned to unlabeled data. For instance, the individual dissimilarity scores are placed in a new vector (formed to hold a plurality of dissimilarity scores for each unlabeled element). The combined new vector and scores therein is the pseudo label for the unlabeled element.”; ¶ [0061] “At block 630, a final model is outputted.”).
Regarding claim 8: the claim limitations are similar to those of claim 1; therefore, rejected in the same manner as applied above. Watson further discloses a system in FIGS. 1 and 2 and a crm in ¶ [0032].
Regarding claim 9: the claim limitations are similar to those of claim 1; therefore, rejected in the same manner as applied above. Watson further discloses a system in FIGS. 1 and 2.
Regarding claim 14: the claim limitations are similar to those of claim 6; therefore, rejected in the same manner as applied above.
Regarding claim 15: the claim limitations are similar to those of claim 7; therefore, rejected in the same manner as applied above.
Claim(s) 2-5, and 10-13 are rejected under 35 U.S.C. 103 as being unpatentable over Watson (US 20190354850) in view of Betthauser (“Discovering Distribution Shifts using Latent Space Representations” Feb, 17, 2022), Watson2 (US 20200082210), and Kender (“A Geometric Approach for Generating Pseudo-labels that Improve Transfer Learning” July 7, 2022).
Regarding claim 2:
Watson in view of Betthauser and Watson2 does not specifically teach: providing a choice of existing sources which further extremizes a geometric content of a hypervolume described by a pseudo label sequence.
However, in a related field, Kender teaches: providing a choice of existing sources which further extremizes a geometric content of a hypervolume described by a pseudo label sequence (abstract; “We generate pseudo-labels according to an efficient and extensible algorithm that is based on a classical result from the geometry of high dimensions, the Cayley-Menger determinant. This G2L (“geometry to label”) method incrementally builds up pseudo-labels using a greedy computation of hypervolume content”; “1 Introduction”; “…. Our method creates pseudo-labels for this data by determining their geometric relationships to the feature space of existing labeled data… pseudo-labels are then constructed for the incoming datapoints based on these distances, or, more accurately, based on “contents”, which is the high-dimensional generalization of distances, calculated using a geometric approach. Each pseudo-label then consists of a sequence of semantically descriptive names: for example, ⟨tool, plant⟩”).
Therefore, it would have been obvious to a person of ordinary skill in the art prior to the effective filing date of the claimed invention to have modified Watson in view of Betthauser and Watson2 to incorporate the teachings of Kender by including: providing a choice of existing sources which further extremizes a geometric content of a hypervolume described by a pseudo label sequence in order to generate richer pseudo-labels for transfer learning targets from known source datasets.
Regarding claim 3:
Watson in view of Betthauser, Watson2, and Kender teaches the limitations of claim 2 as applied above.
Kender further teaches: repeating the providing of the choice until an empirically-determined stopping criterion is performed (abstract “method incrementally builds up pseudo-labels using a greedy computation of hypervolume content.”; “4.4 Empirical Properties of Pseudo-labels” and “4.3 Pseudo-label Creation”; “After a stopping criteria, this sequence gives the pseudo-label.:… The full G2L algorithm is summarized in Algorithm 1. The algorithm requires a number of hyperparameters that are set by experiment”).
Regarding claims 4 and 5:
Watson in view of Betthauser and Watson2, and Kender teaches the limitations of claims 2 and 3 as applied above.
Kender further teaches: further comprising repeating the providing of the choice using differing extremizing criteria (“4.3 Pseudo-label Creation”; “…the process repeats, and at each step the sequence is extended with the name of the anchor point that best extremizes the content—the area, volume, hyper volume, etc.— of the evolving polytope formed by these selected points… Extremizing policies. The extrema decision sequence Pol, and its summarizing notation, are best explained by a walkthrough of the algorithm… The four choices of extremizing policy at any dimension are therefore captured by the quaternary alphabet {c,f,C,F}. And in particular, the policy ⟨C⟩ is the special case already explored in prior work [8], which forms pseudo-labels consisting of the names of ⟨closest, farthest⟩ pairs.”).
Regarding claims 10-13: the claim limitations are similar to those of claims 2-5; therefore, rejected in the same manner as applied above.
Response to Arguments
Applicant’s arguments, see pp. 6-9 of the remarks, filed 06/17/2026, with respect to the 101 rejection have been fully considered and are persuasive. The 101 rejection of claims 1-15 has been withdrawn.
Applicant's arguments filed 06/17/2026 with respect to the 103 rejection have been fully considered but they are not persuasive
Applicant’s reliance on former claim 6 is not persuasive. Original claim 6 required energy distance calculations of the targets to each other and to the sources, together with reciting that the target dataset comprises images. Amended claim 1 require target to source distance but omits target to target distances. This partial inclusion of claim 6 into claim 1 is now rejected in view of Watson2 as applied above in claim 1 rejection.
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 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to WASSIM MAHROUKA whose telephone number is (571)272-2945. The examiner can normally be reached Monday-Thursday 8:00-5:00 EST.
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, Stephen Koziol can be reached at (408) 918-7630. 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.
/WASSIM MAHROUKA/Primary Examiner, Art Unit 2665
Prosecution Insights