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 .
Response to Amendment
This Office Action is in response to applicant’s communication filed 20 April 2026, in response to the Office Action mailed 18 November 2025. The applicant’s remarks and any amendments to the claims or specification have been considered, with the results that follow.
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 18 November 2025 has been entered.
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.
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 CFR 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.
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.
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.
Claim(s) 1-3, 8, 9, 13, and 15-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Shi et al. (Probabilistic Face Embeddings, April 2019, pgs. 1-12 – cited in an IDS) in view of Takahashi (US 2015/0317284).
As per claim 1, Shi teaches a computer-implemented method of modelling a latent space of a trained neural network [a system that converts existing deterministic face embeddings in a latent space into uncertainty aware Probabilistic Face Embeddings (PFE) which represent each face image as a Gaussian distribution in the latent space including learning a model for the PFEs (abstract; pg. 1, fig. 1; pg. 4, section 4; pg. 5, section 4.3; etc.)] comprising: receiving input data for the trained neural network [To address the above problems, we propose Probabilistic Face Embeddings (PFEs), which give a distributional estimation instead of a point estimation in the latent space for each input face image (pg. 1, section 1; etc.); where the input face images are the received input data]; observing the latent space of the trained neural network during inference of the trained neural network, the latent space comprising data from one or more hidden layers of the trained neural network and generating observations of the latent space [given a pre-trained model f which produces a deterministic face embedding in the latent space, we learn an optimized uncertainty module estimate for f(x), where f(x) represents the most likely features of a given input in the latent space using multiple layers of a network (pg. 5, section 4.3; etc.)]; generating a probabilistic model of the latent space of the trained neural network using the observations of the latent space [given a pre-trained model f which produces a deterministic face embedding in the latent space, we learn an optimized uncertainty module estimate for f(x), where f(x) represents the most likely features of a given input in the latent space (pg. 5, section 4.3; etc.), which is a probabilistic model of the latent space]; generating a prediction for a confidence value for an output of the trained neural network [the PFE model can also be used to estimate (predict) the uncertainty/confidence of the model output, for each input (pg. 2, section 1; pg. 4, section 4; etc.)], wherein the confidence value is determined based on an evaluation of a latent space location of the input data using the probabilistic model of the latent space [the PFE model can also be used to estimate (prediction) the uncertainty/confidence of the model output, for each input (pg. 2, section 1; pg. 4, section 4; etc.), where the uncertainty module is used as part of the PFE to estimate the confidence value (pg. 5, section 4.3; etc.); thus using a probabilistic model to generate the prediction of confidence for each (face image) input]; and filtering input data to the trained neural network to provide a control mechanism for the trained neural network and improve an accuracy of the trained neural network [the PFE can be used to filter/reject input images if it is not confident, in order to improve the performance of the system (pg. 8, section 6; etc.)], wherein filtering the input data includes: determining whether the predicted confidence value exceeds a predetermined confidence threshold and only permitting input data having a predicted confidence value that exceeds the predetermined confidence threshold to be processed by the trained neural network to generate an output [In many scenarios, we may expect a higher performance than our system is able to achieve or we may want to make sure the system’s performance can be controlled when facing complex application scenarios. Therefore, we would expect the model to reject input images if it is not confident. A common solution for this is to filter the images with a quality assessment tool. We show that PFE provides a natural solution for this task. We take all the images from LFW and IJB-A datasets for image-level face verification (We do not follow the original protocols here). The system is allowed to “filter out” a proportion of all images to maintain a better performance. We then report the TAR@FAR= 0:001% against the “Filter Out Rate”. We consider two criteria for filtering: (1) the detection score of MTCNN and (2) a confidence value predicted by our uncertainty module (pg. 8, section 6 and fig. 11; etc.); where using the predicted confidence value to filter input images at a pre-specified “Filter Out Rate” would mean that the predicted confidence exceeds a predetermined confidence threshold (e.g., if the confidence-based “Filter Out Rate” is 20% then the confidence threshold is the 20th percentile confidence)].
While Shi teaches producing deterministic face embeddings in the latent space and learning an optimized uncertainty module estimate for f(x), where f(x) represents the most likely features of a given input in the latent space (pg. 5, section 4.3; etc.), it has not been relied upon for teaching generating a probabilistic model of the latent space by modelling the latent space with a probability density function.
Takahashi teaches generating a probabilistic model of the latent space by modelling the latent space with a probability density function [the generation unit generates a model representing a probability distribution of the latent space data b by generating a joint probability density function (paras. 0044-45, etc.)].
Shi and Takahashi are analogous art, as they are within the same field of endeavor, namely learning mapping input image/sensor data to a latent space.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to generate a model representing the probability distribution of the latent space data, as taught by Takahashi, for the latent space face embeddings in the system taught by Shi.
Takahashi provides motivation as [embedding the sensor data in a low-dimensional latent space and then reconstructing the data reduces the amount of calculations necessary and improves processing of the data, and allows time series data to be handled with complicated correlations or changes over time (paras. 0018-22, etc.)].
As per claim 2, Shi/Takahashi teaches receiving input data for the trained neural network during the inference of the trained neural network [given a pre-trained model f which produces a deterministic face embedding in the latent space, we learn an optimized uncertainty module estimate for f(x), which takes the same inputs as the deterministic model (Shi: pg. 5, section 4.3; etc.) To address the above problems, we propose Probabilistic Face Embeddings (PFEs), which give a distributional estimation instead of a point estimation in the latent space for each input face image (Shi: pg. 1, section 1; etc.); where the input face images are the received input data during inference for the trained network]; and wherein the generating of the probabilistic model of the latent space of the trained neural network using the observations of the latent space comprises generating the probabilistic model of the latent space of the trained neural network using the observations of the latent space and the input data received by the trained neural network [given a pre-trained model f which produces a deterministic face embedding in the latent space, we learn an optimized uncertainty module estimate for f(x), which takes the same inputs as the deterministic model (Shi: pg. 5, section 4.3; etc.)].
As per claim 3, Shi/Takahashi teaches receiving output of the trained neural network during inference of the trained neural network [given a pre-trained model f which produces a deterministic face embedding in the latent space, we learn an optimized uncertainty module estimate for f(x) (the output of the trained neural network), which takes the same inputs as the deterministic model (Shi: pg. 5, section 4.3; etc.)]; and wherein the generating of the probabilistic model of the latent space of the trained neural network using the observations of the latent space comprises generating the probabilistic model of the latent space of the trained neural network using the observations of the latent space and the input data received by the trained neural network during inference and the output of the trained neural network during inference of the trained neural network [given a pre-trained model f which produces a deterministic face embedding in the latent space, we learn an optimized uncertainty module estimate for f(x) (the output of the trained neural network), which takes the same inputs as the deterministic model (Shi: pg. 5, section 4.3; etc.)].
As per claim 8, Shi/Takahashi teaches predicting a confidence value for output of a trained neural network having a given input, wherein predicting the confidence value includes using the probabilistic model to generate a prediction of confidence for each of one or more input data to the trained neural network [the PFE model can also be used to estimate (prediction) the uncertainty/confidence of the model output, for each input (Shi: pg. 2, section 1; pg. 4, section 4; etc.), where the uncertainty module is used as part of the PFE to estimate the confidence value (Shi: pg. 5, section 4.3; etc.); thus using a probabilistic model to generate the prediction of confidence for each (face image) input].
As per claim 9, Shi/Takahashi teaches predicting a confidence value for output of a trained neural network having a given input, wherein predicting the confidence value includes using the probabilistic model to generate a prediction of confidence for each of one or more input data to the trained neural network [the PFE model can also be used to estimate (prediction) the uncertainty/confidence of the model output, for each input (Shi: pg. 2, section 1; pg. 4, section 4; etc.), where the uncertainty module is used as part of the PFE to estimate the confidence value (Shi: pg. 5, section 4.3; etc.); thus using a probabilistic model to generate the prediction of confidence for each (face image) input].
As per claim 13, see the rejection of claim 1, above, wherein Shi/Takahashi further teaches a non-transitory computer-readable storage medium storing instructions that, when executed by one or more processors of a computing system, cause the computing system to perform [the method] [the PFE system is combined with a base model which is a 64-layer network trained with AM-Softmax on the MS-Celeb-1M dataset (Shi: pg. 6, section 5.2); which requires program code stored in a memory and executed by at least one processor].
As per claim 15, see the rejection of claim 2, above.
As per claim 16, Shi/Takahashi teaches wherein generating of the probabilistic model of the latent space of the trained neural network using the observations of the latent space includes using the output of the trained neural network during inference of the trained neural network [given a pre-trained model f which produces a deterministic face embedding in the latent space, we learn an optimized uncertainty module estimate for f(x) (the output of the trained neural network), which takes the same inputs as the deterministic model (Shi: pg. 5, section 4.3; etc.); which is using the outputs of the trained model (during inference)].
As per claim 17, see the rejection of claim 8, above.
As per claim 18, see the rejection of claim 1, above, wherein Shi/Takahashi further teaches a computer system comprising one or more processors operable to perform operations for modelling a latent space of a trained neural network comprising [the method steps] [the PFE system is combined with a base model which is a 64-layer network trained with AM-Softmax on the MS-Celeb-1M dataset (Shi: pg. 6, section 5.2); which requires program code stored in a memory and executed by at least one processor].
As per claim 19, see the rejection of claim 2, above.
As per claim 20, see the rejection of claim 8, above.
Response to Arguments
The rejection of claims 13 and 15-17 under 35 U.S.C. 101 has been withdrawn due to the amendments filed.
The objection to claims 15-17 has been withdrawn due to the amendments filed.
Applicant’s arguments, see the remarks, filed 20 April 2026, with respect to the rejections under 35 U.S.C. 101 and providing improvements to the machine learning technology have been fully considered and are persuasive. The rejections of claims 1-3, 8, 9, and 18-20 has been withdrawn.
Applicant’s arguments, see the remarks, filed 20 April 2026, with respect to the rejection(s) of claim(s) 1-3, 8, 9, 13, and 15-20 under 35 U.S.C. 102 have been fully considered and are persuasive in view of the amendments made to the independent claims. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Takahashi, which has been relied upon for teaching generating a probabilistic model of the latent space by modelling the latent space with a probability density function.
Conclusion
The following is a summary of the treatment and status of all claims in the application as recommended by M.P.E.P. 707.07(i): claims 4-7, 10-12, and 14 are cancelled; claims 1-3, 8, 9, 13, and 15-20 are rejected.
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Gonzalez Aguirre (US 2019/0135300) – discloses performing probabilistic modeling of sensor data in a latent space using unsupervised auto-encoding and using the model to predict a confidence based on sensor inputs, with which to perform anomaly detection.
Amores (US 2020/0243069) – discloses determining a confidence score for portions of text, and comparing the confidence to a threshold to determine whether to use that portion of text as training data.
Borland (US 2016/0094567) – discloses storing a confidence level with training data records, and ignoring records for which the confidence level does not meet a minimum threshold.
Braho (US 2006/0178882) – discloses assigning confidence to speech signals, and if the confidence factor is below the acceptance threshold, rejecting/ignoring the speech input.
Northcutt et al. (Learning with Confident Examples: Rank Pruning for Robust Classification with Noisy Labels, Aug 2017, pgs. 1-27) – discloses a rank pruning algorithm that selects confident examples and removes the rest, before training on the pruned set.
Zubarev (US 2020/0401111) – discloses systems/methods utilizing a VAE(s) with encoders used to encode a Gaussian probability density of a latent space(s).
Beggel (US 2019/0130279) – discloses a system/method for detecting anomalous images using an adversarial autoencoder, including determining a probability density of a distribution for a latent space.
The examiner requests, in response to this Office action, that support be shown for language added to any original claims on amendment and any new claims. That is, indicate support for newly added claim language by specifically pointing to page(s) and line number(s) in the specification and/or drawing figure(s). This will assist the examiner in prosecuting the application.
When responding to this office action, Applicant is advised to clearly point out the patentable novelty which he or she thinks the claims present, in view of the state of the art disclosed by the references cited or the objections made. He or she must also show how the amendments avoid such references or objections. See 37 CFR 1.111(c).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GEORGE GIROUX whose telephone number is (571)272-9769. The examiner can normally be reached M-F 10am-6pm.
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, Omar Fernandez Rivas can be reached on 571-272-2589. 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.
/GEORGE GIROUX/Primary Examiner, Art Unit 2128
Prosecution Insights