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 .
DETAILED ACTION
This communication is a non-Final office action on merit. Clams 1, 3-20, after amendment, are presently pending and have been considered below.
Request for Continued Examination
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 3/19/2026 has been entered.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 10/24/2024 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Objections
Claims 3-5 are objected to because of the following informalities:
Claims 3-5 depend from claim 2 directly or indirectly and claim 2 has been canceled.. Appropriate correction is required.
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 of this title, 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 set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied 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, 3-7, 14-17, 20 are rejected under 35 U.S.C. 103 as being unpatentable over US 2022/0076135 A1, Chen et al. (hereinafter Chen) in view of “Variable Rate Deep Image Compression with Modulated Autoencoder”, Yang et al. Journal of LATEX class files, Vol. 14. 14, No. 8, August 2019, an IDS submission (hereinafter Yang) and further in view of US2021/0089682 A1, Bercich et al. (hereinafter Bercich).
As to claim 1, Chen discloses a method comprising:
computing, by one or more processors of a computing system, a first conditional meta embedded feature based on inputting a first set of optimized meta-control variables into a first learning model (Figs 1, 7-10; pars 0004-0006, 0023, 0025-0029, meta-learning/controlled features being extracted over a meta learning model/system); and a second conditional meta embedded feature based on inputting a second set of optimized meta-control variables into a second modulation learning model (Figs 1, 7-10; pars 0004-0006, 0023, 0025-0029, meta-learning/controlled features being extracted over a meta learning model/system, note such features include domain dependence features and domain agnostic features through the model and model variables optimization with respective inputs or decompositions of inputs. It is also note that the first and second features here do not have to be domain dependent and domain agnostic features, the learning model can be applied and optimized for more than one set of features based on their respective inputs);
tuning, by the one or more processors, parameter of an auto-encoder based on the first conditional meta embedding feature (Fig 1; pars 0019-0025, 0029-0030, 0041, meta-learning adaptation process tuning/adapting the model parameters of an auto-encoder, joint distribution being appropriately conditioned to maintain consistency with the interpretability and transferred to an appreciated laten space for domain dependent and domain agnostic latent representation); and
computing, by the one or more processors configured by the auto-encoder, a latent representation of an input picture (Fig 1; pars 0022-0023, 0025, 0027, 0029-0030, providing latent representation of input image) and
computing, by the one or more processors, a statistical measure describing the latent representation based on a hyperprior of the latent representation derived from the second conditional meta embedded feature (Fig 1; pars 0022-0023, 0025, 0027, 0029-0030, 0046, rationale for the first computing above applies here; note the measurements of maximum mean discrepancy, correlation alignment distance etc. are statistical in nature, and the inferred variables (parameters) from the observed data being optimized are hyperprior in nature (see pars 0025, 0046)).
Chen does not expressly disclose the autoencoder being a modulated autoencoder.
Yang, in the same or similar field of endeavor, additionally teaches a modulation learning model with an inputting of a first set of optimized meta-control variables (Abstract; Figs 1-2; I-II., a modulated autoencoder with variable rate-distortion control) and tuning, by the one or more processors, a conditional hyperparameter of an auto-encoder based on the first conditional meta embedding feature (Fig 2; II-III, tuning within range of a list of hyperparameters λ for autoencoder training) and keep encoder and decoder fixed (II. B, Bottleneck scaling); and computing, by the one or more processors, a statistical measure describing the latent representation based on the second conditional meta embedded feature (Abstract; Figs 1-2; I-II, more than on input images being processed with respective features extracted).
Additionally, Bercich, in the same or similar field of endeavor, also teaches parameters or weights in one or more nodes of an auto-encoder can be fixed during training or encoding process (pars 0035-0038, claim 5).
Therefore, consider Chen, Yang, and Bercich’s teachings as a whole, it would have been obvious to one of skill in the art before the filing date of invention to incorporate Yang and Bercich’s teachings in Chen’s method to provide a modulated autoencoder training for meta learning utilizing hyperparameters for training latent related features with different characteristics.
2. (Canceled)
As to claim 3, Chen as modified discloses the method of claim 2, further comprising: coding, by the one or more processors, the latent representation as a coded picture based on the statistical measure (Chen: pars 0022-0025, 0027, 0029; Yang: abstract; Figs 1-2); and transmitting, by the one or more processors, the coded picture and the second set of optimized meta-control variables in a bitstream (Yang: Figs 1-2; I-II, binary representations (bitstream) being transmitted).
As to claim 4, Chen as modified discloses the method of claim 2, wherein the first modulation learning model and the second modulation learning model each comprises a respective plurality of fully- connected layers (Yang: Fig 2, fully connect layers) and a respective plurality of activation layers (Yang: Fig 2, ReLU commonly contains/equips with activation layers to performs activation function in a deep learning neural network model).
As to claim 5, Chen as modified discloses the method of claim 3, further comprising: transmitting, by the one or more processors, a third set of optimized meta-control variables in a bitstream; wherein the first, second, and third sets of optimized meta-control variables are each learned by optimizing a rate distortion (RD) loss during online meta-learning by stochastic gradient descent (SGD) (Chen: Fig 2; pars0036, LSTM being updated with SGD; Yang: Fig 2; II., SGD adaptation).
As to claim 6, Chen as modified discloses the method of claim 1, wherein tuning a conditional hyperparameter of the auto-encoder based on the first conditional meta embedding feature comprises receiving, by the one or more processors, the first conditional meta embedding feature at a plurality of conditional feature modulation inputs, wherein each conditional feature modulation input corresponds to a respective encoding block of the auto-encoder (Yang: I-II; Figs 1-2).
As to claim 7, Chen as modified discloses the method of claim 6, wherein tuning a conditional hyperparameter of the auto-encoder based on the first conditional meta embedding feature further comprises computing, by the one or more processors, a multiplication operation between the conditional meta embedding feature and an output of a respective encoding block of the auto-encoder (Yang: Fig 2; II-III).
As to claim 14, Chen as modified discloses a method comprising: reading, by one or more processors of a computing system, a coded picture and a first set of optimized meta-control variables and a second set of optimized meta-control variables from a bitstream (Chen: pars 0019, 0025, 0028; Yang: Figs 1-2; I-III; also see citation and rejection in claim 1); computing, by one or more processors of a computing system, a first conditional meta embedded feature based on inputting the first set of optimized meta-control variables into a first modulation learning model and a second conditional meta embedded feature based on inputting the second set of optimized meta-control variables into a second modulation learning model (see citation and rejection in claim 1); tuning, by the one or more processors, a conditional hyperparameter of an entropy decoder based on the first conditional meta embedding feature while holding fixed model parameters of the auto-encoder (Yang: Figs 1-2, I-III, parameters of the encoder and decoder are learned from certain image data by jointly minimizing rate and distortion at a particular R-D tradeoff; Bercich: pars 0035-0038, claim 5, also see rejection in claim 1); and decoding, by the one or more processors, a decoded latent representation based on inputting the coded picture into the entropy decoder (Yang: Figs 1-2; I-III, entropy decoder to obtain a decoded latent representation in a modulated autoencoder); and computing, by the one or more processors, a statistical measure describing the decoded latent representation based on decoding the bitstream and based on a context model (Chen: pars 0019, 0025, 0028; Yang: Figs 1-2; also see rejection in claim 1). Consider Chen, Yang, and Bercich’s teachings as a whole, it would have been obvious to one of skill in the art before the filing date of invention to incorporate Yang’s teachings on entropy decoding in Chen’s method to complete encoding and decoding processes of the deep learning neural network.
As to claim 15, Chen as modified discloses the method of claim 14, further comprising: tuning, by the one or more processors, a conditional hyperparameter of an auto-decoder based on the second conditional meta embedded feature (Yang: Figs 1-2; I-III).
As to claim 16, Chen as modified discloses the method of claim 15, further comprising: computing, by the one or more processors, a reconstructed picture by inputting the decoded latent representation into the auto-decoder (Yang: Figs 2, 5; II-III, decoder reconstruct the output image for reconstructed representation).
As to claim 17, it is rejected with the same reason as set forth in claim 4.
As to claim 20, it is rejected with the same reason as set forth in claim 14.
Claims 8, 10-13 are rejected under 35 U.S.C. 103 as being unpatentable over Yang in view of Chen.
As to claim 8, Yang discloses a method comprising: computing, by the one or more processors, a training latent representation based on a first training conditional meta embedded feature and an input training picture (Abstract; I-II; Figs 1-2; jointly training modulated autoencoder and the modulation network under R-D tradeoff constraint);
computing, by the one or more processors, a decoded training latent based on a second training conditional meta embedded feature by dequantization (Figs 1-2; I-II, dequantized decoding to reconstruct representation); and computing, by the one or more processors, a reconstructed picture based on the decoded training latent and a third training conditional meta embedded feature (Figs 1-2; I-II, reconstructed image/picture being decoded with optimized R-D tradeoff);
wherein the first, second, and third training conditional meta embedded features are respectively derived from a first, second, and third set of optimized meta-control variables each learned by optimizing a rate distortion (RD) loss during online meta- learning by stochastic gradient descent (SGD) (Abstract; I-III, Figs 1-2).
Yang does not expressly disclose the conditional embedded feature being a meta based feature. Chen, in the same or similar field of endeavor, further teaches a meta learning system (Figs 1, 7-10; pars 0004-0006, 0019-0021, 0023). Chen also teaches set of optimized meta-control variables each learned by optimizing a rate distortion (RD) loss during online meta- learning by stochastic gradient descent (SGD) (Figs 1-2, 5; pars 0036, 0049-0050, the update rule being stochastic gradient decent (SGD)),
Therefore, consider Yang and Chen’s teachings as a whole, it would have been obvious to one of skill in the art before the filing date of invention to incorporate Chen’s teachings in Yang’s method to provide a meta based learning model for training latent representation with given R-D tradeoff.
As to claim 10, Yang as modified discloses the method of claim 8, further comprising: computing, by the one or more processors, statistical measures describing the training latent representation (Chen: pars 0025, 0046, 0057; Yang: Abstract; Figs 1-2; I-II); coding, by the one or more processors, a coded training picture based on the training statistical measures (Yang: Figs 1-2; I-II, IV, entropy encoding for latent representation with R-D tradeoff); and computing, by the one or more processors, a rate loss based on the statistical measures (Yang: Abstract; Figs 1-2; I-II, IV; the rate loss/distortion being minimized with distortion).
As to claim 11, Yang as modified discloses the method of claim 10, further comprising: computing, by the one or more processors, a distortion loss based on the input training picture and the reconstructed picture (Yang: Abstract; I-III, Figs 1-2); and computing, by the one or more processors, an updated RD loss based on the estimated rate loss and the distortion loss (Yang: Abstract; I-III, Figs 1-2).
As to claim 12, Yang as modified discloses the method of claim 10, wherein the training latent representation and the statistical measures are each transmitted in a bitstream (Yang: Figs 1-2; I-II, binary representations (bitstream) being transmitted), the decoded training latent is derived from the training latent representation and the statistical measures transmitted in the bitstream, and the rate loss is based on a bitrate of the bitstream (Yang: Abstract; I-III, Figs 1-2).
As to claim 13, Yang as modified discloses the method of claim 11, wherein the first, second, and third set of optimized meta-control variables are each updated based on the updated RD loss during the online meta-learning (Yang: Abstract; I-III, Figs 1-2; Chen: Fig 2; par 0036, LSTM being updated with SGD).
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Yang in view of Chen and further in view of US 2020/0372301 A1, Kearney et al. (hereinafter Kearney).
As to claim 9, Yang as modified discloses the method of claim 8, further comprising: updating a meta- control variable based on the RD loss; and learning, by the one or more processors, the meta-control variable based on a stochastic gradient descent (SGD) computed from the rate distortion loss (see claim 8) but does not expressly teach determining a step size for updating the meta-control variable learning process.
Kearney, in the same or similar field of endeavor, further teaches determining, by the one or more processors, a step size for updating a meta- control variable based on the RD loss (pars 0064, 0077, 0151, 170, estimate the step size for the training process). Therefore, consider Yang as modified and Kearney’s teachings as a whole, it would have been obvious to one of skill in the art before the fling date of invention to incorporate Kearney’s teachings in Yang as modified’s method to properly control the convergent rate of the learning process.
Allowable Subject Matter
Claims 18-19 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.
Reasons for Allowance
Prior art of record (Chen, Yang, and Kearney) neither discloses alone nor teaches in combination functions and features recited in claim 18. Claim 19 depends from claim 18.
Response to Arguments
Applicant’s arguments have been considered but they are not persuasive. Applicant’s arguments and concerns have bee addressed in merit rejection above where additional rationale and citations have been provided.
Examiner’s Note
Examiner has cited particular column, line number, paragraphs and/or figure(s) in the reference(s) as applied to the claims for the convenience of the Applicant. Although the specified citations are representative of the teachings of the art and are applied to the specific limitations within the individual claim, other passages and figures may apply as well. It is respectfully requested from the applicant in preparing responses, to fully consider the reference(s) in entirety as potentially teaching all or part of the claimed invention, as well as the context of the passage as taught by the prior art or disclosed by the Examiner.
Contact Information
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Qun Shen whose telephone number is (571) 270-7927. The examiner can normally be reached on Mon-Friday from 9:00-5:00. If attempts to reach the examiner by telephone are unsuccessful, the examiner's Supervisor, Amandeep Saini can be reached on (571) 272-3382. The fax phone number for the organization where this application or proceeding is assigned is (571) 273-8300. Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
/QUN SHEN/
Primary Examiner, Art Unit 2662
Prosecution Insights