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 .
Information Disclosure Statement
The information disclosure statements (IDSs) submitted on November 20, 2024 and on August 25, 2025 are in compliance with 37 CFR 1.97 and 1.98 and therefore have been considered by the examiner and placed in the file.
Claim Interpretation
The claims in this application are given their broadest reasonable interpretation (BRI) using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The BRIs are used for purposes of searching for prior art, but cannot be incorporated into the claims. Claim limitations must be given their plain meaning unless such meaning is inconsistent with the specification. MPEP 2111.01. BRIs for some of the claim limitations are provided below. Should Applicant believe that other interpretations are warranted, Applicant should point to the portions of the present disclosure that clearly show that a different interpretation is appropriate.
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.
Claim 14 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 14 does not use proper idiomatic English and its meaning cannot be understood. It appears to refer to mixing of the plurality of second parameters with the plurality of first parameters, but the examiner is unable to determine from the claim language what the mixing involves or what the result is.
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.
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.
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.
Claims 1, 2, 9-13, 17 and 18 re rejected under 35 U.S.C. 103 as being unpatentable over U.S. Publ. Appl. No. 2026/0156047 A1 to Wang et al. (hereinafter referred to as “Wang”) in view of U.S. Publ. Appl. No. 2018/0365512 A1 to Molchanov et al. (hereinafter referred to as “Molchanov”) and further in view of U.S. Publ. Appl. No. 2025/0104308 A1 to Zhou et al. (hereinafter referred to as “Zhou”).
Regarding claim 1, Wang discloses an image processing method (Paras. [0015]-[0016] disclose that the method performs image recognition and classification), comprising:
training a first neural network model configured to execute a first image processing, according to a plurality of training data, to generate a plurality of first parameters associated with the first neural network model, wherein the plurality of first parameters comprises a plurality of weights (Para. [0014], the first submodel of the learning model constitutes a first neural network. Para. [0180], the submodels are sub-neural networks of the learning model neural network. Para. [0040], steps are performed according to a “formula to train and update a target submodel, so as to obtain a parameter”. Paras. [0015] and [0020], the first image processing executed by the first neural network submodel is one of a plurality of image recognition tasks, such as processing images of license plates obtained by a camera of a vehicle to recognize the license plate. Wang does not explicitly disclose that the plurality of first parameters comprises a plurality of weights);
training a second neural network model configured to execute a second image processing, according to the plurality of training data and the plurality of weights, to generate a plurality of second parameters associated with the second neural network model, wherein the second image processing is different from the first image processing (Para. [0014], the second submodel of the learning model constitutes a second neural network. Para. [0180], the submodels are sub-neural networks of the learning model neural network. Para. [0040], steps are performed according to a “formula to train and update a target submodel, so as to obtain a parameter”. Paras. [0015] and [0021], the second image processing executed by the second neural network submodel is a different one of the plurality of image recognition tasks, such as processing images of a human body obtained by a camera to recognize the human body. Wang does not explicitly disclose that training of the second neural network is according to the weights of the first plurality of first parameters); and
mixing the plurality of first parameters with the plurality of second parameters, to generate a plurality of blending parameters for a blending neural network model, wherein the blending neural network model is configured to execute the first image processing and the second image processing on an input image, to output an optimized image (the BRI for this limitation, based on paras. [0006]-[0007] of the present disclosure, is that a combination of parameters of the pluralities of first and second parameters are used to generate parameters of a neural network model configured to execute the first and second image processing on an input image to produce an optimized output image. Para. [0020] of Wang discloses that the learning neural network model uses a combination of the updated parameters of the first and second submodel neural networks to determine updated parameters of the learning neural network model, which functions as a blending neural network model, and that the learning neural network model optimizes performance of the first and second image processing tasks based on the updated parameters of the learning neural network model to produce an optimized result: “[w]hen the learning model needs to optimize M (where M is an integer greater than or equal to 2) subtasks, a process of updating the learning model may be understood as updating M submodels (where the M submodels include the first submodel) that are in the learning model and that are used to execute the M subtasks. Correspondingly, a basis for determining the updated learning model may include the parameter of the updated first submodel and a parameter of the another submodel (for example, the second submodel).” Wang does not explicitly disclose outputting an image based on the first and second image processing tasks executed by the blending learning model).
As indicated above, Wang does not explicitly refer to the parameters comprising weights, but it is well known in the art that parameters of neural networks that are updated during training comprise weights and biases. Molchanov, in the same field of endeavor, discloses that parameters comprising weights are generated during updating of a classification neural network (Para. [0053]).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the present disclosure, for the parameters that are generated by the first submodel neural network of Wang to comprise weights. One of ordinary skill in the art would have been motivated to generate parameters that comprise weights during training to ensure that the neurons of the neural network are properly updated during training. The inclusion of weights would have been ensured by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because neural network software commonly used for configuring neural networks includes computer code for setting and updating weights as parameters of the neural networks.
As indicated above, Wang also does not explicitly disclose that training of the second neural network is according to the weights of the first plurality of first parameters. Zhou, in the same field of endeavor, discloses transferring model parameters from a first neural network model to a second neural network model (Fig. 2, G1 is the first neural network model and G2 is the second neural network model. Paras. [0040]-[0042], during transfer learning, the parameters that were used to train model G1 are transferred to model G2 and used by model G2. Para. [0035], model G1 performs a first image processing to generate an image of a face. Para. [0042], the second model G2 performs a second image processing to add “local regional clothing, hairstyles, hair accessories and makeups based on the original facial features of the user” such that the image that is output from model G2 is an optimized image that includes the face image output by model G1 with the effects added by model G2).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the present disclosure, to modify the method of Wang such that the parameters that are generated by the first submodel neural network of Wang are transferred to the second submodel neural network along with the training data and used along with the training data to train the second submodel neural network. One of ordinary skill in the art would have been motivated to use transfer learning for the second submodel neural network of Wang to improve training efficiency and accuracy by allowing the second submodel neural network to benefit from training already performed on the first submodel neural network. The inclusion modification could have been made by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because making the modification merely involves combining prior art elements according to known methods to yield predictable results (modifying the software executed by processor(s) of Wang to determine weight parameters of the second submodel according to the weight parameters of the first submodel).
As indicated above, Wang does not explicitly disclose outputting an optimized image based on the first and second image processing tasks executed by the blending learning model of Wang. While Wang teaches outputting an optimized result of the image recognition tasks, the result is not necessarily an image.
As indicated above, Zhou discloses that the second model G2 outputs an output image corresponding to a modified (i.e., stylized) input image. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the present disclosure, to modify the system and method of Wang such that the image processing tasks performed by the first and second submodels and by the blending learning neural network model are processes that produce an output image. One of ordinary skill in the art would have been motivated to make the modification to enable the method and system of Wang to be used to produce optimized output images. The modification could have been made by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because making the modification merely involves combining prior art elements according to known methods to yield predictable results (modifying the software executed by the processor(s) of Wang to perform other types of image processing tasks that produce an optimized output image).
Regarding claim 2, Wang discloses:
generating a plurality of first output data, through the first neural network model, after training the first neural network model is completed (Para. [0015], after the first neural network submodel is trained, it performs the task for which it was trained to generate first output data, which is the result of performing the task, such as the result of recognizing a license plate);
generating a plurality of second output data, through the second neural network model, after training the second neural network model is completed (Para. [0015], after the second neural network submodel is trained, it performs the task for which it was trained to generate second output data, which is the result of performing the task, such as the result of recognizing a human body); and
further training the blending neural network model, by using a blending loss function, to optimize the plurality of blending parameters (Para. [0167], Fig. 2B the blending learning neural network model back-propagates an error until a loss function reaches a minimum value at “the optimal point”).
Regarding claim 9, Wang does not explicitly disclose that the first submodel neural network is a convolutional neural network (CNN). Wang does not specify the type of neural network architectures that are used.
Molchanov discloses using a CNN for one or more of its neural networks (Fig. 1D, paras. [0036]-[0037]). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the present disclosure, to use a CNN for the first neural network submodel of Wang since CNNs are commonly used for image processing tasks, such as feature extraction and object recognition. One of ordinary skill in the art would have been motivated to use a CNN for this purpose to take advantage of the well-known benefits of CNNs, such as their ability to perform automatic feature extraction and object recognition with relatively high computational efficiency. The modification could have been made by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because making the modification merely involves combining prior art elements according to known methods to yield predictable results (modifying the software executed by the processor(s) of Wang to implement a neural network architecture for the first submodel neural network).
Regarding claim 10, Wang does not explicitly disclose that the second submodel neural network is a generative adversarial neural network (GAN). As indicated above, Wang does not specify the type of neural network architectures that are used.
Zhou discloses using a GAN for one or more of its neural networks (Para. [0036]-[0037]). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the present disclosure, to use a GAN for the second neural network submodel of Wang since GANs are commonly used for semi-supervised training when training datasets are limited or need to be augmented. One of ordinary skill in the art would have been motivated to use a GAN for this purpose to take advantage of the well-known benefits of GANs. The modification could have been made by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because making the modification merely involves combining prior art elements according to known methods to yield predictable results (modifying the software executed by the processor(s) of Wang to implement a GAN architecture for the second submodel neural network).
Regarding claim 11, as indicated above in the rejection of claim 10, it would have been obvious to use a GAN for the second submodel of Wang. The limitations recited in claim 11 merely describe the normal components of a GAN and the normal manner in which the generator and discriminator of the GAN operate during training. Zhou discloses that the GAN includes a generator network and a discriminator network that perform the operations recited in claim 11 (Para. [0036]-[0037]). The rejection of claim 10 applies mutatis mutandis to claim 11.
Regarding claim 12, the rejections of claims 9 and 10 apply mutatis mutandis to claim 12.
Regarding claim 13, Wang does not explicitly disclose that the blending learning neural network model is a CNN or a GAN. However, it would be obvious for the same reasons discussed above in the rejections of claims 9 and 10 to use either a CNN or a GAN for the blending learning neural network. CNNs and GANs are commonly used for image processing tasks. The rejections of claims 9 and 10 apply mutatis mutandis to claim 13.
Regarding claim 17, Wang discloses that a neural network processor performs the method described therein (Para. [0291]). As indicated above in the rejection of claim 1, the combined teachings of Wang, Molchanov and Zhou teach a blending neural network model configured to execute a plurality of image processings on an input image to output an optimized image, and that the plurality of model parameters of the blending learning neural network comprises a combination of the first plurality and second plurality of parameters of the first and second submodels, respectively.
Regarding the limitation of a “first proportion” and a “second proportion” of parameters, the BRI for this limitation, based on para. [0032] of the present specification, is that the plurality of model parameters of the blending neural network model is made up of some portion of the first plurality of parameters and some portion of the second pluralities of parameters, where the portions are determined as some ratio of the first plurality of parameters to the second plurality of parameters. Since Wang teaches that the plurality of parameters of the blending learning neural network model includes model parameters of both the first and second pluralities of model parameters, this meets the “proportion” limitation because there is some ratio of first plurality of parameters to the second plurality of parameters (e.g., 1:1) in the plurality of parameters of the blending learning neural network model.
Therefore, the rejections of claims 1, 9 and 10 apply mutatis mutandis to claim 17. As indicated above in the rejection of claims 9 and 10, it would have been obvious to use a CNN and a GAN for the first and second submodels of Wang, in which case the first plurality of model parameters would be model parameters of a CNN and the second plurality of model parameters would be model parameters of a CNN, which means that model parameters of the blending learning neural network would have some number of the model parameters of the CNN and some number of the model parameters of the GAN.
Regarding claim 18, the rejections of claims 1, 9 and 10 apply mutatis mutandis to claim 18. The BRI for the term “are correlated” is based on the plain meaning because this term is not defined in the specification. The Merriam-Webster online dictionary defines correlation as a relationship. As indicated above in the rejection of claim 1, Zhou discloses transferring model parameters from a first neural network model to a second neural network model and therefore the plurality of parameters of the second neural network are determined based on the plurality of parameters of the first neural network, which means there is a relationship between the pluralities of parameters since they are both used as blending parameters.
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Wang in view of Molchanov and Zhou and further in view of U.S. Pat. No. 11,640,519 B2 to Chen et al. (hereinafter referred to as “Chen”).
Regarding claim 4, the BRI for this claim limitation, based on para. [0042] of the present specification, is that the linear superpositioning means adding, summing or combining the loss functions. Wang discloses using fused parameters and determining a fused loss of the submodels and updating the parameters of the blending learning neural network model based on the updated fused parameters of the submodels. Wang discloses that doing this enhances the model generalization capability of the blending learning neural network. (Paras. [0046], [0080] and [0214]). However, the combined teachings of Wang, Molchanov and Zhou do not explicitly disclose that the loss function of the blending learning neural network model is a linear superposition of the loss functions of the submodels.
Chen, in the same field of endeavor, discloses adding a plurality of loss functions to obtain a blended loss function (Col. 11, lines 10-33). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the present disclosure, to linearly superposition the loss functions of the first and second submodels of Wang as taught by Chen to obtain a blended loss function for the blending learning neural network model of Wang. One of ordinary skill in the art would have been motivated to make the modification to improve the training and updating of the neural network system of Wang. The modification could have been made by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because making the modification merely involves combining prior art elements according to known methods to yield predictable results (modifying the software executed by the processor(s) of Wang to represent the loss function of the blending learning neural network of Wang as a summation of the loss functions of the submodel neural networks).
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Wang in view of Molchanov, Zhou and Chen as applied to claim 4 and further in view of U.S. Publ. Appl. No. 2023/0334833 A1 to Shi et al. (hereinafter referred to as “Shi”).
Regarding claim 5, the combined teachings of Wang, Molchanov, Zhou and Chen do not explicitly teach that the plurality of loss functions comprise at least one of a noise suppression loss function, a sharpening loss function, and an image-edge-enhancement loss function.
Shi, in the same field of endeavor, discloses that different image processing models perform respective image processing tasks including image enhancement and image sharpening and use respective loss functions (Paras. [0034]-[0035]) and summing the loss functions to generate a linear superposition of the loss functions (Para. [0099]: “[t]herefore, the process in which the network parameter of each network is updated (namely, corrected) in the first way may be as follows: the super-resolution loss function, the image quality loss function, the face loss function and the sharpening loss function are added, and then the network parameter of the super-resolution network is updated according to the loss function obtained by summation”).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the present disclosure, to linearly superposition the loss functions of the first and second submodels of Wang as taught by Chen and Shi, where the loss functions include at least a sharpening loss function as taught by Shi. One of ordinary skill in the art would have been motivated to make the modification to mutually constrain and influence the training of the models during parameter updating such that the models are “trained in a manner of mutual association, mutual integration, and mutual promotion” as taught by Shi to achieve “superimposed optimization” as taught by Shi. The modification could have been made by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because making the modification merely involves combining prior art elements according to known methods to yield predictable results (modifying the software executed by the processor(s) of Wang to represent the loss function of the blending learning neural network of Wang as a summation of the loss functions of the submodel neural networks).
Regarding claim 6, the rejection of claim 5 applies mutatis mutandis to claim 6. As indicated in the rejection of claim 5, Shi discloses that the blending loss function is at least a sharpening loss function as described above in the rejection of claim 5.
Claims 7 and 8 are rejected under 35 U.S.C. 103 as being unpatentable over Wang in view of Molchanov and Zhou as applied to claims 1, 2, 9-13, 17 and 18 and further in view of a printed publication entitled “Pre-treating images before machine learning”, published on Community Partners forum in image.sc in May 2022 (hereinafter referred to as “Pre-treating publication”).
Regarding claim 7, the combined teachings of Wang, Molchanov and Zhou do not explicitly teach executing pre-processing to optimize training data before training the first submodel neural network.
The Pre-treating publication, in the same field of endeavor, discloses that it is known to pre-process image data to optimize it before using it to train a machine learning model (Comments by Ruman Gerst and Janina Hanne regarding performing edge enhancement and brightness normalization on images before using them to train the models).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the present disclosure, to modify the system and method of Wang as modified based on the teachings of Molchanov and Zhou further based on the teachings of the Pre-treating publication to perform pre-processing on the image data before using the image data to train the first submodel. One of ordinary skill in the art would have been motivated to make the modification to improve the performance of the submodel and reduce training time. The modification could have been made by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because making the modification merely involves combining prior art elements according to known methods to yield predictable results (modifying the software executed by the processor(s) of Wang to perform pre-processing of the image data before using it to train the submodel).
Regarding claim 8, the rejection of claim 7 applies mutatis mutandis to claim 8. As indicated in the rejection of claim 7, the Pre-treating publication discloses that the pre-processing comprises an edge enhancement process.
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Wang in view of Molchanov and Zhou as applied to claims 1, 2, 9-13, 17 and 18 and further in view of U.S. Publ. Appl. No. 2023/0385986 A1 to Park et al. (hereinafter referred to as “Park”) and U.S. Publ. Appl. No. 2026/0134867 A1 to Shao et al. (hereinafter referred to as “Shao”).
Regarding claim 15, the limitation that begins “training a plurality of neural networks” recites features that are recited in claim 1. Likewise, the limitation that begins “mixing the plurality of sets” recites features that are recited in claim 1. Regarding these features, the rejection of claim 1 applies mutatis mutandis to claim 15.
Regarding the order of training the neural network models, since no order is specified in the claim, the BRI for this limitation is that the neural network models are trained in some order. The combined teachings of Wang, Molchanov and Zhou do not explicitly teach training the neural network submodels in any particular order. Park, in the same field of endeavor, discloses training neural network models trained in a particular order (Para. [0086]: “[a]s shown in FIG. 4, two neural network models, that is, the first and second neural network models 100 and 200, may be simultaneously trained in a state where the first neural network model 100 and the second neural network model 200 are connected, or the first neural network model 100 may be first trained and then the second neural network model 200 may be trained in a state where the trained first neural network model 100 is fixed (parameters included in the first neural network model 100 are fixed). Also, each of the first neural network model 100 and the second neural network model 200 may be separately trained.”).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the present disclosure, to modify the system and method of Wang as modified based on the teachings of Molchanov and Zhou further based on the teachings of Parker to train the submodels of Wang in a particular order as taught by Parker. One of ordinary skill in the art would have been motivated to make the modification to ensure that the submodels are trained in the best order that improves learning efficiency and prevents inadvertent unlearning of parameters. The modification could have been made by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because making the modification merely involves combining prior art elements according to known methods to yield predictable results (modifying the software executed by the processor(s) of Wang to select the training order that makes the most logical sense or to provide the user with the ability to select the order).
Regarding the limitation in claim 15 of adjusting the set of blending parameters according to a set of output data of one of the plurality of neural network models and a blending loss function, the combined teachings of Wang, Molchanov, Zhou and Park do not explicitly teach this limitation.
As indicated above in the rejection of claims 4 and 5, which apply mutatis mutandis to claim 15, Wang discloses adjusting the set of blending parameters according to a blending loss function during updating of the blending learning neural network, but Wang does not explicitly teach adjusting the set of blending parameters according to a set of output data of one of the neural network submodels.
Shao, in the same field of endeavor, discloses adjusting the set of blending parameters according to a set of output data of one of a plurality of neural network models. In Shao, neural network language submodels 410, 420 and 430 are used for performing tasks associated with recognizing speech (Abstract, paras. [0068]-[0071] and Fig. 4. The submodel 410 receives output data and parameters that are output from the fully connected layer of submodel 420 in the adders of submodel 410 in order to achieve feature fusion between the implicit representations outputted from the fully connected layers of submodels 410 and 420 (Para. [0073]: “[i]n this way, it is possible to achieve a sufficient fusion of the first implicit representation and the features in the first language sub-model, so that the sharing of network parameters in the first language sub-model and network parameters in the second language sub-model may be enhanced, the accuracy of the obtained first language probability may be improved, and a speech recognition accuracy may be improved.”).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the present disclosure, to modify the system and method of Wang as modified based on the teachings of Molchanov and Zhou further based on the teachings of Shao to update the blending parameters of Wang based not only on the first and second parameters of the first and second submodels, respectively, but also on output data of one or more of the submodels. One of ordinary skill in the art would have been motivated to make the modification to provide improved recognition accuracy. The modification could have been made by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because making the modification merely involves combining prior art elements according to known methods to yield predictable results (modifying the software executed by the processor(s) of Wang to select the provide output data of the submodel(s) along with their parameters to the blending learning neural network model).
Allowable Subject Matter
Claims 3 and 16 are 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.
The following is a statement of reasons for the indication of allowable subject matter:
Regarding claim 3, none of the art of record teaches or suggests, in combination with the other limitations recited in claims 1 and 2, training the blending neural network model according to the first output data or the second output data generated by the first and second neural network models, respectively, based on the types of neural networks that are used for the first, second and blending neural networks.
Regarding claim 16, none of the art of record teaches or suggests, in combination with the limitations recited in claim 15, determining a training order of the plurality of neural network models according to convergence difficulty.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DANIEL J SANTOS whose telephone number is (571)272-2867. The examiner can normally be reached M-F 9-5.
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, Matt Bella can be reached at (571)272-7778. 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.
/DANIEL J. SANTOS/Examiner, Art Unit 2667
/MATTHEW C BELLA/Supervisory Patent Examiner, Art Unit 2667
Prosecution Insights