Detailed Action
1. Claims 6-38 are pending in this Application.
Notice of Pre-AIA or AIA Status
2. 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 Election /Restriction
3 Applicant’s election without traverse of Group II claims 6-38 in the reply filed on 04/28/2026 is acknowledged.
4 Claims 1-5 are withdrawn from further consideration pursuant to 37 CFR 1.142(b) as being drawn to a nonelected invention, there being no allowable generic or linking claim
Claim Rejections - 35 USC § 101
5. 35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
6. Claims 6-38 are rejected under 35 U.S.C 101.
Regarding claim 6,
Step 1: This claim is directed towards a method .
Step 2A Prong one :
The claim recites “ A method, comprising: at a device: computing one or more diffusion bridges between a first version of data and a second version of the data”; and
The “diffusion bridges between a first version of data and a second version of the data” limitation is mere data gathering recited at a high level of generality and is insignificant extra- solution activity.
Using one more diffusion bridges. The one more diffusion bridges are used to apply the abstract idea( i.e., perform data-to- data translation ) without placing any limitation how the diffusion bridges operates.
The limitation amounts to mere instruction to implement an abstract idea on a computer or merely use of use of computer as a tool to perform data -to-data conversion process.
Step 2A Prong two
The claim further recite “ using the one or more diffusion bridges to train a conditional diffusion model to perform data-to-data translation”
The above limitations are mere data gathering recited at a high level of generality and is insignificant extra- solution activity.
Using diffusion bridges of a machine learning model to perform data-to-data translation in order to obtain training data without placing any limitation how the attention engine operates. The limitation amounts to mere instruction to implement an abstract idea on a computer or merely use of use of computer as a tool to perform an abstract idea (see MPEP 2106. 05 (f)
Even when viewed in combination this additional element elements do not integrate the recited judicial exception into a particular application.
Step 2B:
The claim does not provide an inventive concepts (significantly more than the abstract idea ). The claim is ineligible .As explained in Step 2A the limitation “using the one or more diffusion bridges to train a conditional diffusion model to perform data-to-data translation” was mere instruction to apply an exception. This conclusion does not change step 2B change. It is also well-understood , routine and conventional the diffusion bridges of the machine learning model to receive and process and generate trained image data. Even when considering in combination, these additional elements represent mere instruction to apply an exception and insignificantly extra-solution activity, which cannot provide an inventive concept.
Regarding claim 23,
Step 1: This claim is directed towards a system
Step 2A Prong one :
The claim recites “ A system, comprising: a non-transitory memory storage comprising instructions; and one or more processors in communication with the memory, wherein the one or more processors execute the instructions to: compute one or more diffusion bridges between a first version of data and a second version of the data
The “diffusion bridges between a first version of data and a second version of the data” limitation is mere data gathering recited at a high level of generality and is insignificant extra- solution activity.
Using one more diffusion bridges. The one more diffusion bridges are used to apply the abstract idea( i.e., perform data-to- data translation ) without placing any limitation how the diffusion bridges operates.
The limitation amounts to mere instruction to implement an abstract idea merely use of computer as a tool to perform data-to-date conversion process.
Step 2A Prong two
The claim further recite “ using the one or more diffusion bridges to train a conditional diffusion model to perform data-to-data translation”
The above limitations are mere data gathering recited at a high level of generality and is insignificant extra- solution activity.
Using diffusion bridges of a machine learning model to perform data-to-data translation in order to obtain training data without placing any limitation how the attention engine operates. The limitation amounts to mere instruction to implement an abstract idea merely use of computer as a tool to perform an abstract idea (see MPEP 2106. 05 (f)
Even when viewed in combination this additional element elements do not integrate the recited judicial exception into a particular application.
Step 2B:
The claim does not provide an inventive concepts (significantly more than the abstract idea ). The claim is ineligible .As explained in Step 2A the limitation “using the one or more diffusion bridges to train a conditional diffusion model to perform data-to-data translation” was mere instruction to apply an exception. This conclusion does not change step 2B change. It is also well-understood , routine and conventional the diffusion bridges of the machine learning model to receive and process and generate trained image data. Even when considering in combination, these additional elements represent mere instruction to apply an exception and insignificantly extra-solution activity, which cannot provide an inventive concept.
Claim 36 is rejected the same as claim 23 except claim 36 is directed to a computer program claim. Thus , the analysis applied to claim 23 above is also applicable to claim 23.
Dependent claims
Regarding claims 7, 9 and 10
Step 1: The claims are directed towards a method
Step 2A, prong 1: The claims are a part of the judicial exception as noted above in claim 6.
Step 2A, prong 2: The claims further recites “ the first version of the data is a degraded version of the data and the second version of the data is a clean target version of the data.”, “the data is an image.” and “the first version of the image is a degraded version of the image and the second version of the image is a clean target version of the image” respectively.
However, the additional elements amount to mere insignificant pre-extra solution steps of data gathering and therefore do not integrate the judicial exception into a practical application.
Step 2B: Claims7,9 and 10, recite limitations do not integrate into a practical application or amount to significantly more for the reasons provided in claim 6.
Accordingly, claims 7,9 and 10 are directed to non-eligible patent subject matter and is therefore rejected.
Claims 24,26 and 27 are rejected the same as claims 7,9 and 10 respectively except claims 24,26 and 27 are directed to a system claims . Thus , the analysis applied to claims 7,9 and 10 above is also applicable to claims 24,26 and 27.
Regarding claims 8, 11,12, and 13
Step 1: The claims are directed towards a method
Step 2A, prong 1: The claims are a part of the judicial exception as noted above in claim 6.
Step 2A, prong 2: The claims further recites “the data-to-data translation is a data restoration task.”, “the degraded version of the image has a lower resolution than the clean target version of the image.”, “the degraded version of the image is a corrupted version of the clean target version of the image.”, and “the degraded version of the image includes more blurring than the clean target version of the image.” respectively . However, the additional elements amount to mere insignificant pre-extra solution steps of data gathering and data manipulation, and therefore do not integrate the judicial exception into a practical application.
Step 2B: Claims 8, 11, 12 and 13 recite limitations do not integrate into a practical application or amount to significantly more for the reasons provided in claim 6.
Accordingly, claims 8,11,12 and 13 are directed to non-eligible patent subject matter and is therefore rejected.
Claims 25 and 28 are rejected the same as claims 8,11,12 and 13 respectively except claims 25 and 28 are directed to a system claims . Thus , the analysis applied to claims 8,11,12 and 13 above is also applicable to claims 25 and 28.
Regarding claims 14-15
Step 1: The claims are directed towards a method
Step 2A, prong 1: The claims are a part of the judicial exception as noted above in claim 6.
Step 2A, prong 2: The claims further recites “the degraded version of the image is captured by a camera.”, and “the camera is a component of an autonomous driving system.” respectively . However, the additional elements amount to mere insignificant pre-extra solution steps of data gathering, and therefore do not integrate the judicial exception into a practical application.
Step 2B: Claims 14-15 recite limitations do not integrate into a practical application or amount to significantly more for the reasons provided in claim 6.
Accordingly, claims 14-15 are directed to non-eligible patent subject matter and is therefore rejected.
Claims 29-30 are rejected the same as claims 14-15 respectively except claims 29- 30 are directed to a system claims . Thus , the analysis applied to claims 14-15 above is also applicable to claims 29 and 30.
Regarding claims 16-22
Step 1: The claims are directed towards a method
Step 2A, prong 1: The claims are a part of the judicial exception as noted above in claim 6.
Step 2A, prong 2: The claims further recites “the one or more diffusion bridges are tractable, interpretable, and efficient.”, “ the one or more diffusion bridges between the first version of the data and the second version of the data are nonlinear.”, “the one or more diffusion bridges correspond to one or more time steps existing between the first version of the data and the second version of the data.”, “the one or more diffusion bridges train a score function to perform the data-to-data translation.”, “the conditional diffusion model is trained to perform the data-to-data translation for a data compression application.”, “the conditional diffusion model is trained to perform the data-to-data translation for a robotics application.” and “the conditional diffusion model is trained to perform the data-to-data translation for an autonomous driving application.” respectively. However, the additional elements amount to mere insignificant pre-extra solution steps of data gathering, and data manipulation, and therefore do not integrate the judicial exception into a practical application.
Step 2B: Claims 16-22 recite limitations do not integrate into a practical application or amount to significantly more for the reasons provided in claim 6.
Accordingly, claims 16-22 are directed to non-eligible patent subject matter and is therefore rejected.
Claim 35 is rejected the same as claims 20-22 respectively except claim 35 are directed to a system claims . Thus , the analysis applied to claims 20-22 above is also applicable to claims 35.
Claim 37-38 are rejected the same as claims 18-19 respectively except claim 37-38 are directed to a computer program claims . Thus , the analysis applied to claims 18-19 above is also applicable to claims 37-38.
Claim Rejections - 35 USC § 102
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 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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless -
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale or otherwise available to the public before the effective filing date of the claimed invention.
7. Claims 6-13, 16-20, 23-28 and 31-34 and 36-38 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Tianrong et al., (hereafter Tianrong), “LIKELIHOOD TRAINING OF SCHRÖDINGER BRIDGE USING FORWARD-BACKWARD SDES THEORY” published as a conference paper at ICLR 2022, pub. 2022
As to claim 6, Tianrong teaches A method, comprising: at a device: computing one or more diffusion bridges between a first version of data and a second version of the data ( page 1 section 1, 1st par., Forward-Backward stochastic differential equation (SDE) method for computing the log-likelihood of Schrödinger Bridge (SB). The method includes data-to-noise diffusion using the Schrödinger Bridge (SB) models).
As to claim 7, Tianrong teaches the first version of the data is a degraded version of the data and the second version of the data is a clean target version of the data(as discussed in claim 1 above the method include data-to-noise diffusion. The degraded version of the data and the clean target version of the data corresponds to the noise data and the data respectively).
As to claim 8, Tianrong teaches the data-to-data translation is a data restoration task( Fig.1, as shown in Fig.1. The Schrödinger Bridge (SB) enhances generative AI by connecting two specific distributions (noisy and clean data) with an optimal trajectory., SB treats denoising as a data-to-data process, utilizing intermediate guidance to efficiently clean and reconstruct data.
As to claim 9, Tianrong teaches the data is an image (see Fig.1).
As to claim 10, Tianrong teaches the first version of the image is a degraded version of the image and the second version of the image is a clean target version of the image (Fig.1 illustrates the noise image and clean image, and as shown in Fig. 1 the data-to-noise diffusion using SB, where both the clean data and noise data are image data)
As to claim 11, Tianrong teaches the degraded version of the image has a lower resolution than the clean target version of the image (Fig.1 shown a sequence of images, where the first image from right to left is a noise image of a dog (i.e., low resolution image) while the last image is a clean image of the dog which is obtained by denoising the noise image)
As to claim 12, Tianrong teaches the degraded version of the image is a corrupted version of the clean target version of the image (as discussed in claim 11, above the clean image of the doge is obtained from the noise image of the dog, thus, the noise image of the dog is the degraded version of the clean image of the dog. ).
As to claim 13, Tianrong teaches the degraded version of the image includes more blurring than the clean target version of the image ( inherent and as shown in Fig 1. where the first image from right to left is a noise image which is blurred image, while the last image is a clean image of the dog which is obtained by deblurring the noise image)
As to claim 16, Tianrong teaches the one or more diffusion bridges are tractable, interpretable, and efficient( page 2 3rd- 4th pars,. inherent, The Diffusion Schrödinger Bridge (DSB) makes score-based generative modeling (SGM) efficient by replacing the arbitrarily long, fixed forward noising process with an optimal transport path. Instead of diffusing data until it reaches an isotropic Gaussian, DSB learns both the forward and backward SDEs, requiring significantly fewer discretization steps to generate high-quality sample)
As to claim 17, Tianrong teaches the one or more diffusion bridges between the first version of the data and the second version of the data are nonlinear (page 5 section 3, page 3 1st par., page 6 equation A Diffusion Schrödinger Bridge (DSB) is fundamentally nonlinear. While a DSB problem can be structurally decomposed into two linear partial differential equations (the forward and backward Fokker-Planck/Feynman-Kac equations) for computational tractability, the dynamic diffusion process itself directly learns the underlying nonlinear transformations between two arbitrary probability distributions).
As to claim 18, Tianrong teaches the one or more diffusion bridges correspond to one or more time steps existing between the first version of the data and the second version of the data (Fig.1 shown a sequence of for image images at different time, where the first image from right to left is a noise image of a dog while the last image is a clean image of the dog which is obtained by denoising the noise image, and two intermediate images ).
As to claim 19, Tianrong teaches the one or more diffusion bridges train a score function to perform the data-to-data translation(Fig.2, section 5 page 6, Score-based Generative Model, and facilitate applications of modern generative training for the Schrödinger Bridge (SB), which enhances generative AI. As shown in Fig.2 the data-to-noise diffusion, SB learns the process).
As to claim 20, Tianrong teaches the conditional diffusion model (Abstract) is trained to perform the data-to-data translation for a data compression application ( Fig. 7 Figure 7: Network architecture for toy datasets includes position encoding)
As to claim 23, Tianrong teaches A system, comprising: a non-transitory memory storage comprising instructions; and one or more processors in communication with the memory (page 71st , par., algorithms 1-2, the algorithms illustrates a computer program code stored in a memory that carry image processing related to data to noise diffusion based on likelihood training of SB), wherein the one or more processors execute the instructions to: compute one or more diffusion bridges between a first version of data and a second version of the data; and use the one or more diffusion bridges to train a conditional diffusion model to perform data-to-data translation (page 1 section 1, 1st par., all this limitation discussed in claim 6 above).
Claim 24 is rejected the same as claim 7 except claim 24 is directed to a system claim . Thus, argument analogous to that presented above for claim 7 is also applicable to claim 24.
Claim 25 is rejected the same as claim 8 except claim 25 is directed to a system claim . Thus, argument analogous to that presented above for claim 8 is also applicable to claim 25.
Claim 26 is rejected the same as claim 9 except claim 26 is directed to a system claim . Thus, argument analogous to that presented above for claim 9 is also applicable to claim 26.
Claim 27 is rejected the same as claim 10 except claim 27 is directed to a system claim . Thus, argument analogous to that presented above for claim 10 is also applicable to claim 27.
Claim 28 is rejected the same as claims 11-13 except claim 28 is directed to a system claim. Thus, argument analogous to that presented above for claims11-13 are also applicable to claim 28.
Claim 31 is rejected the same as claim 16 except claim 31is directed to a system claim . Thus, argument analogous to that presented above for claim 16 is also applicable to claim 31.
Claim 32 is rejected the same as claim 17 except claim 32is directed to a system claim . Thus, argument analogous to that presented above for claim 17 is also applicable to claim 32.
Claim 33 is rejected the same as claim 18 except claim 33 is directed to a system claim . Thus, argument analogous to that presented above for claim 18 is also applicable to claim 33.
Claim 34 is rejected the same as claim 19 except claim 34 is directed to a system claim . Thus, argument analogous to that presented above for claim 19 is also applicable to claim 34.
Claim 36 is rejected the same as claim 23 except claim 36 is directed to a computer program claim .Thus, argument analogous to that presented above for claim 23 is also applicable to claim 36.
Claim 37 is rejected the same as claim 18 except claim 37 is directed to a computer program claim. Thus, argument analogous to that presented above for claim 18 is also applicable to claim 37.
Claim 38 is rejected the same as claim 19 except claim 38 is directed to a computer program claim. Thus, argument analogous to that presented above for claim 19 is also applicable to claim 38.
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 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.
6. Claims 14-15 and 21-22, 29-30 and 35 are rejected under 35 U.S.C. 103 as being unpatentable over Tianrong, “LIKELIHOOD TRAINING OF SCHRÖDINGER BRIDGE USING FORWARD-BACKWARD SDES THEORY”, in view of Song; Jiaming( hereafter Song), US 20240046422 A1, filed Aug. 3, 2022 (claims the benefit of U.S. Provisional Application No. 63/394,776)
As to claim 14, Tianrong teaches the degraded version of the image (Fig. 1 noise image) is captured by a camera.
However, it is noted that Tianrong does not specifically teach the degraded version of the image is captured by a camera
On the other hand in the same filed of endeavor a pseudoinverse guidance for data restoration with diffusion models of Song teaches “ the degraded version of the image is captured by a camera ([0056], [0087]image processing includes image super-resolution (recover high-resolution images from low-resolution images), image deblurring (reduce blurring artifacts in images, such as optical blur and motion blur), JPEG artifact restoration, where the low- resolution image is captured by a digital camera).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to incorporate a parallel processing unit (PPU) 400
which is configured to accelerate numerous deep learning systems and applications taught by Song (see [0059])into Tianrong.
The suggestion/motivation for doing so would have been to allow user of Tianrong
to improve the learning algorithm by extending it to generate autonomous vehicle and robotics training data (see [0059])
Claim 29 is rejected the same as claim 14 except claim 29 is directed to a system claim . Thus, argument analogous to that presented above for claim 14 is also applicable to claim 29.
As to claim 15, Song teaches the camera is a component of an autonomous driving system([0058], [0087, The PPU 400 may be configured to accelerate numerous deep learning systems and applications for autonomous vehicles, where the TPPU 400 may be included in, supercomputers, a smart-phone, personal digital assistant (PDA), a digital camera,],
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to incorporate a parallel processing unit (PPU) 400
which is configured to accelerate numerous deep learning systems and applications taught by Song (see [0059])into Tianrong
The suggestion/motivation for doing so would have been to allow user of Tianrong
to improve the learning algorithm by extending it to generate autonomous vehicle and robotics training data (see [0059])
Claim 30 is rejected the same as claim 15 except claim 30 is directed to a system claim . Thus, argument analogous to that presented above for claim 15 is also applicable to claim 30.
As to claim 21, Tianrong teaches the conditional diffusion model (Abstract) is trained to perform the data-to-data translation (Abstract, Fig.1, While SGM requires pre-specifying the data-to-noise diffusion, SB instead learns the process);
However, it is noted that Tianrong does not specifically teach “ perform the data-to-data translation for a robotics application”
On the other hand in the same filed of endeavor a pseudoinverse guidance for data restoration with diffusion models of Song teaches the conditional diffusion model is trained to perform the data-to-data translation for a robotics application([0057]-[0059], FIG. 4 illustrates a parallel processing unit (PPU) 400, The PPU 400 may be used to implement one or more of the diffusion model 260. The PPU 400 may be configured to accelerate numerous deep learning systems and applications for autonomous vehicles, robotics, big data analytics, and alike)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to incorporate a parallel processing unit (PPU) 400
which is configured to accelerate numerous deep learning systems and applications taught by Song (see [0059])into Tianrong.
The suggestion/motivation for doing so would have been to allow user of Tianrong
to improve the learning algorithm by extending it to generate autonomous vehicle and robotics training data (see [0059])
As to claim 22, Tianrong teaches the conditional diffusion model (Abstract) is trained to perform the data-to-data translation (Abstract, Fig.1, While SGM requires pre-specifying the data-to-noise diffusion, SB instead learns the process);
However, it is noted that Tianrong does not specifically teach “ perform the data-to-data translation for an autonomous driving application”
On the other hand in the same filed of endeavor a pseudoinverse guidance for data restoration with diffusion models of Song teaches the conditional diffusion model is trained to perform the data-to-data translation for an autonomous driving application (0057]-[0059], FIG. 4 illustrates a parallel processing unit (PPU) 400, The PPU 400 may be used to implement one or more of the diffusion model 260. The PPU 400 may be configured to accelerate numerous deep learning systems and applications for autonomous vehicles, robotics, big data analytics, and alike).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to incorporate a parallel processing unit (PPU) 400
which is configured to accelerate numerous deep learning systems and applications taught by Song (see [0059])into Tianrong
The suggestion/motivation for doing so would have been to allow user of Tianrong
to improve the learning algorithm by extending it to generate autonomous vehicle and robotics training data (see [0059])
Claim 35 is rejected the same as claims 21-22 except claim 35 is directed to a system claim. Thus, argument analogous to that presented above for claims 21-22 are also applicable to claim 35.
Prior art not used in rejections but pertinent to the claims or disclosure.
“Diffusion Schrödinger Bridge with Applications to Score-Based Generative Modelin.” 35th Conference on Neural Information Processing Systems (NeurIPS 2021), pub. 2021 to Valentin De Bortoli et al., disclosed:
“Progressively applying Gaussian noise transforms complex data distributions to
approximately Gaussian. Reversing this dynamic defines a generative model. When
the forward noising process is given by a Stochastic Differential Equation (SDE),
Song et al. (2021) demonstrate how the time inhomogeneous drift of the associated reverse-time SDE may be estimated using score-matching. A limitation of this approach is that the forward-time SDE must be run for a sufficiently long time for the final distribution to be approximately Gaussian while ensuring that the corresponding time-discretization error is controlled. In contrast, solving the Schrödinger Bridge (SB) problem, i.e. an entropy-regularized optimal transport problem on path spaces, yields diffusions which generate samples from the data distribution in finite time. We present Diffusion SB (DSB), an original approximation of the Iterative Proportional Fitting (IPF) procedure to solve the SB problem, and provide theoretical analysis along with generative modeling experiments. The first DSB iteration recovers the methodology proposed by Song et al. (2021), with the flexibility of using shorter time intervals, as subsequent DSB iterations reduce the discrepancy between the final-time marginal of the forward (resp. backward) SDE with respect to the Gaussian prior (resp. data) distribution. Beyond generative modeling, DSB offers a computational optimal transport tool as the continuous state-space analogue of the popular Sinkhorn algorithm (Cuturi, 2013), see Abstract.
Contact Information
Any inquiry concerning this communication or earlier communication from the examiner should be directed to Mekonen Bekele whose telephone number is (469) 295-9077.The examiner can normally be reached on Monday -Friday from 9:00AM to 6:50 PM Eastern Time.
If attempt to reach the examiner by telephone are unsuccessful, the examiner’s supervisor Eng, George can be reached on (571) 272-7495.The fax phone number for the organization where the application or proceeding is assigned is 571-237-8300. Information regarding the status of an application may be obtained from the patent Application Information Retrieval (PAIR) system. Status information for published application may be obtained from either Private PAIR or Public PAIR.
Status information for unpublished application is available through Privet PAIR only.
For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have question on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866.217-919 (tool-free)
/MEKONEN T BEKELE/Primary Examiner, Art Unit 2699
Prosecution Insights