DETAILED ACTION
This action is in response to the application filed on 08/28/2023. Claims 1-35 are pending and have been examined.
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 .
The information disclosure statement (IDS) submitted on 01/03/2024, 03/21/2025, and 02/18/2026. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 101
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.
Claims 7 and 24 rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 7:
Subject Matter of Eligibility Analysis Step 1:
Claim 7 recites a method, which is directed to a process, and thus is one of the four statutory categories of patentable subject matter.
Subject Matter of Eligibility Analysis Step 2A Prong 1:
Claim 7 recites
generating the synthetic data using a physics based simulation, and wherein the synthetic data is generated to mimic real world regional data (this limitation is a mental process as it encompasses a human mentally generating synthetic data).
Therefore, claim 7 recites an abstract idea.
Subject Matter of Eligibility Analysis Step 2A Prong 2:
Claim 7 recites additional elements of
performing self-supervised learning on a first dataset to initially train the machine learning model is an instruction to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
performing region specific training on the initially trained machine learning model using a second dataset is an instruction to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
refining the machine learning model using a third dataset to train the machine learning model to perform a particular inference task is an instruction to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 7 is not integrated into a practical application.
Subject Matter of Eligibility Analysis Step 2B:
The additional elements of claim 7 do not provide significantly more than the abstract idea itself, take alone or in combination because
performing self-supervised learning on a first dataset to initially train the machine learning model (this element does not integrate the abstract idea into a practical application because it amounts to mere instructions to apply (see MPEP 21.06(f))).
performing region specific training on the initially trained machine learning model using a second dataset (this element does not integrate the abstract idea into a practical application because it amounts to mere instructions to apply (see MPEP 21.06(f))).
refining the machine learning model using a third dataset to train the machine learning model to perform a particular inference task (this element does not integrate the abstract idea into a practical application because it amounts to mere instructions to apply (see MPEP 21.06(f))).
Therefore, claim 7 is subject matter ineligible.
Regarding claim 24:
Subject Matter of Eligibility Analysis Step 1:
Claim 24 recites a system, which is directed to a machine, and thus is one of the four statutory categories of patentable subject matter.
Subject Matter of Eligibility Analysis Step 2A Prong 1:
Claim 24 recites
generating the synthetic data using a physics based simulation, and wherein the synthetic data is generated to mimic real world regional data (this limitation is a mental process as it encompasses a human mentally generating synthetic data).
Therefore, claim 24 recites an abstract idea.
Subject Matter of Eligibility Analysis Step 2A Prong 2:
Claim 24 recites additional elements of
performing self-supervised learning on a first dataset to initially train the machine learning model is an instruction to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
performing region specific training on the initially trained machine learning model using a second dataset is an instruction to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
refining the machine learning model using a third dataset to train the machine learning model to perform a particular inference task is an instruction to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 24 is not integrated into a practical application.
Subject Matter of Eligibility Analysis Step 2B:
The additional elements of claim 24 do not provide significantly more than the abstract idea itself, take alone or in combination because
performing self-supervised learning on a first dataset to initially train the machine learning model (this element does not integrate the abstract idea into a practical application because it amounts to mere instructions to apply (see MPEP 21.06(f))).
performing region specific training on the initially trained machine learning model using a second dataset (this element does not integrate the abstract idea into a practical application because it amounts to mere instructions to apply (see MPEP 21.06(f))).
refining the machine learning model using a third dataset to train the machine learning model to perform a particular inference task (this element does not integrate the abstract idea into a practical application because it amounts to mere instructions to apply (see MPEP 21.06(f))).
Therefore, claim 24 is subject matter ineligible.
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.
Claim(s) 1-11, 16, 18-28, and 35 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li et al. (WO 2020146863) (hereafter referred to as Li) in view of Denli et al. (WO 2022140717) (hereafter referred to as Denli).
Regarding claim 1, Li teaches
performing region specific training on the initially trained machine learning model using a second dataset (Li, paragraph 00160, “the output can be for a target region (e.g., a selected region) where a second ML model is to be trained in a supervised manner”).
refining the machine learning model using a third dataset to train the machine learning model to perform a particular inference task (Li, paragraph 00154, “an interpreter may decide to refine one or more regions, which may be accomplished by selecting a region, automatically selecting the training data (e.g., as labeled) and adjusting and/or augmenting the training data to retrain a specific ML model for that region, which can then be executed to process the corresponding seismic data, which may generate a refined earth model for that region”).
Li does not teach, but Denli does teach
performing self-supervised learning on a first dataset to initially train the machine learning model (Denli, paragraph 0052, “the initial or first stage may comprise unsupervised (or self-supervised) machine learning with a larger dataset that is configured for pre-training”).
Li and Denli are considered analogous to the claimed invention because they both deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li to use self-supervised learning from Denli. One of the ordinary skill in the art would have known that both unsupervised learning and self-supervised learning serve the same purpose of learning features from unlabeled data. Therefore, substituting Li’s unsupervised learning with Denli’s self-supervised learning would have yield the predictable result of learning feature representation from unlabeled seismic data. (See MPEP 2141 (III)(B) Simple substitution of one known element for another to obtain predictable results).
Regarding claim 2, Li and Denli teach the method of claim 1, Li further teaches
the first dataset comprises unlabeled data (Li, abstract, “A method can include receiving a first trained machine model trained via unsupervised learning using unlabeled seismic image data”).
Regarding claim 3, Li and Denli teach the method of claim 1, Li further teaches
the second dataset is associated with a particular geographic region (Li, paragraph 00160, “the output can be for a target region (e.g., a selected region) where a second ML model is to be trained in a supervised manner”).
Regarding claim 4, Li and Denli teach the method of claim 1, Li further teaches
the third dataset comprises labeled data (Li, paragraph 00154, “an interpreter may decide to refine one or more regions, which may be accomplished by selecting a region, automatically selecting the training data (e.g., as labeled) and adjusting and/or augmenting the training data to retrain a specific ML model for that region”).
Regarding claim 5, Li and Denli teach the method of claim 1, Denli further teaches
the third dataset comprises synthetic data (Denli, paragraph 0074, “Embedding space may thus be used in a variety of contexts, such as analyzing different types of datasets (e.g., seismic images (including real or synthetic data), dialogs, etc.), tracking, or cataloging”).
Li and Denli are considered analogous to the claimed invention because they both deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li to use synthetic data from Denli. One of the ordinary skill in the art would have known to apply the known technique of including synthetic data into a dataset. Therefore, applying Denli’s technique would yield the predictable results of . (See MPEP 2141 (III)(B) Simple substitution of one known element for another to obtain predictable results of reducing cost/time by using synthetic data over real-world data. (See MPEP 2141 (III)(D) Applying a known technique to a known device ready for improvement to yield predicable results).
Regarding claim 6, Li and Denli teach the method of claim 1, Li further teaches
the third dataset is less than 10 percent the size of the first dataset (Li, paragraph 00162, “In terms of “how much” data are involved in unsupervised training (e.g., unsupervised learning) and supervised training (e.g., supervised learning), where, in the foregoing example, unsupervised training utilizes the entire Solsikke seismic cube, supervised training may utilize a relatively small percentage (e.g., less than approximately 1 percent, less than approximately 0.5 percent, or less than approximately 0.1 percent).“)
Regarding claim 7, Li and Denli teach the method of claim 1, Li further teaches
generating the synthetic data using a physics based simulation, and wherein the synthetic data is generated to mimic real world regional data (Li, paragraph 0030, “In an example embodiment, the simulation component 120 may rely on entities 122. Entities 122 may include earth entities or geological objects such as wells, surfaces, bodies, reservoirs, etc. In the system 100, the entities 122 can include virtual representations of actual physical entities that are reconstructed for purposes of simulation”).
Regarding claim 8, Li and Denli teach the method of claim 1, Denli further teaches
the particular inference task comprises wave picking to identify at least one of: a geographic fault; a geographic layer; P-wave arrival; S-wave arrival; de-noising; synthetic data generation; horizon picking; event identification; or a location of a subsurface feature (Denli, paragraph 0027, “ ’Seismic data’ is also intended to include any data (e.g., seismic image, migration image, reverse-time migration image, pre-stack image, partially-stack image, full-stack image, poststack image or seismic attribute image) or interpretation quantities, including geophysical properties such as one or more of: elastic properties (e.g., P and/or S wave velocity, P- Impedance, S-Impedance, density, attenuation, anisotropy and the like); and porosity, permeability or the like, that the ordinarily skilled artisan at the time of this disclosure will recognize may be inferred or otherwise derived from such data received and/or recorded as part of the seismic surveying and interpretation process” and “Examples of geological features include, without limitation salt, fault, channel, environment of deposition (EoD), facies, carbonate, rock types (e.g., sand and shale), horizon, stratigraphy, or geological time” (Denli, paragraph 0029)).
Li and Denli are considered analogous to the claimed invention because they both deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li to include all the types of inference tasks from Denli. One of the ordinary skill in the art would have known to apply the known technique of to use the types of inference tasks from Denli to train a seismic machine learning model. Therefore, applying Denli’s technique would yield the predictable results of using seismic inference tasks to improve the interpretation of the data. (See MPEP 2141 (III)(D) Applying a known technique to a known device ready for improvement to yield predicable results).
Regarding claim 9, Li and Denli teach the method of claim 1, Denli further teaches
the first, second, and third datasets are distributed acoustic sensing (DAS) datasets (Denli, paragraph 0028, “ ‘Seismic data’ may also include data derived from traditional seismic (e.g., acoustic) data sets in conjunction with other geophysical data, including, for example, gravity plus seismic; gravity plus electromagnetic plus seismic data, etc.” Examiner notes that the seismic acoustic data is mapped to the distributed acoustic sensing datasets).
Li and Denli are considered analogous to the claimed invention because they both deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li to use the seismic acoustic data from Denli. One of the ordinary skill in the art would have known to apply the known technique of to use seismic acoustic data to train a seismic machine learning model. Therefore, applying Denli’s technique would yield the predictable results of using cost-effective real-world data. (See MPEP 2141 (III)(D) Applying a known technique to a known device ready for improvement to yield predicable results).
Regarding claim 10, Li and Denli teach the method of claim 1, Li further teaches
the first, second, and third datasets are seismic imaging datasets (Li, Abstract, “ A method can include receiving a first trained machine model trained via unsupervised learning using unlabeled seismic image data; receiving labeled seismic image data acquired via an interactive interpretation process; and building a second trained machine model, as initialized from the first trained machine model, via supervised learning using the received labels, where the second trained machine model predicts stratigraphy of a geologic region from seismic image data of the geologic region”).
Regarding claim 11, Li and Denli teach the method of claim 1, Denli further teaches
the first dataset comprises synthetic data (Denli, paragraph 0074, “Embedding space may thus be used in a variety of contexts, such as analyzing different types of datasets (e.g., seismic images (including real or synthetic data), dialogs, etc.), tracking, or cataloging”).
Li and Denli are considered analogous to the claimed invention because they both deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li to use synthetic data from Denli. One of the ordinary skill in the art would have known to apply the known technique of including synthetic data into a dataset. Therefore, applying Denli’s technique would yield the predictable results of reducing cost/time by using synthetic data over real-world data. (See MPEP 2141 (III)(D) Applying a known technique to a known device ready for improvement to yield predicable results).
Regarding claim 16, Li and Denli teach the method of claim 1, Denli further teaches
refining the machine learning model using the third dataset comprises performing supervised learning training methods to learn feature extraction on the third dataset (Denli, Claim 2: The method of claim 1, wherein generating the embedding model comprises performing unsupervised or self-supervised machine learning; and wherein tailoring the embedding model comprises performing supervised machine learning. Claim 8: …wherein tailoring the embedding model in order to identify the one or more geological features of interest comprises…. Examiner notes that claim 2 maps to the supervised learning and claim 8 maps to the feature extraction).
Regarding claim 18, Li teaches
one or more processors (Li, paragraph 0003, “A system can include a processor; memory operatively coupled to the processor; and processor-executable instructions stored in the memory to instruct the system”).
one or more tangible, non-transitory media operably connectable to the one or more processors and storing instructions that, when executed, cause the one or more processors to perform operations (Li, paragraph 0062, “ As to the one or more computers 254, each computer may include one or more processors (e.g., or processing cores) 256 and memory 258 for storing instructions (e.g., one or more of the one or more sets of instructions 270), for example, executable by at least one of the one or more processors 256”).
performing region specific training on the initially trained machine learning model using a second dataset (Li, paragraph 00160, “the output can be for a target region (e.g., a selected region) where a second ML model is to be trained in a supervised manner”).
refining the machine learning model using a third dataset to train the machine learning model to perform a particular inference task (Li, paragraph 00154, “an interpreter may decide to refine one or more regions, which may be accomplished by selecting a region, automatically selecting the training data (e.g., as labeled) and adjusting and/or augmenting the training data to retrain a specific ML model for that region, which can then be executed to process the corresponding seismic data, which may generate a refined earth model for that region”).
Li does not teach, but Denli does teach
performing self-supervised learning on a first dataset to initially train the machine learning model (Denli, paragraph 0052, “the initial or first stage may comprise unsupervised (or self-supervised) machine learning with a larger dataset that is configured for pre-training”).
Li and Denli are considered analogous to the claimed invention because they both deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li to use self-supervised learning from Denli. One of the ordinary skill in the art would have known that both unsupervised learning and self-supervised learning serve the same purpose of learning features from unlabeled data. Therefore, substituting Li’s unsupervised learning with Denli’s self-supervised learning would have yield the predictable result of learning feature representation from unlabeled seismic data. (See MPEP 2141 (III)(B) Simple substitution of one known element for another to obtain predictable results).
Regarding claim 19, Li and Denli teach the system of claim 18, Li further teaches
the first dataset comprises unlabeled data (Li, abstract, “A method can include receiving a first trained machine model trained via unsupervised learning using unlabeled seismic image data”).
Regarding claim 20, Li and Denli teach the method of claim 18, Li further teaches
the second dataset is associated with a particular geographic region (Li, paragraph 00160, “the output can be for a target region (e.g., a selected region) where a second ML model is to be trained in a supervised manner”).
Regarding claim 21, Li and Denli teach the method of claim 18, Li further teaches
the third dataset comprises labeled data (Li, paragraph 00154, “an interpreter may decide to refine one or more regions, which may be accomplished by selecting a region, automatically selecting the training data (e.g., as labeled) and adjusting and/or augmenting the training data to retrain a specific ML model for that region”).
Regarding claim 22, Li and Denli teach the method of claim 18, Denli further teaches
the third dataset comprises synthetic data (Denli, paragraph 0074, “Embedding space may thus be used in a variety of contexts, such as analyzing different types of datasets (e.g., seismic images (including real or synthetic data), dialogs, etc.), tracking, or cataloging”).
Li and Denli are considered analogous to the claimed invention because they both deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li to use synthetic data from Denli. One of the ordinary skill in the art would have known to apply the known technique of including synthetic data into a dataset. Therefore, applying Denli’s technique would yield the predictable results of . (See MPEP 2141 (III)(B) Simple substitution of one known element for another to obtain predictable results of reducing cost/time by using synthetic data over real-world data. (See MPEP 2141 (III)(D) Applying a known technique to a known device ready for improvement to yield predicable results).
Regarding claim 23, Li and Denli teach the system of claim 18, Li further teaches
the third dataset is less than 10 percent the size of the first dataset (Li, paragraph 00162, “In terms of “how much” data are involved in unsupervised training (e.g., unsupervised learning) and supervised training (e.g., supervised learning), where, in the foregoing example, unsupervised training utilizes the entire Solsikke seismic cube, supervised training may utilize a relatively small percentage (e.g., less than approximately 1 percent, less than approximately 0.5 percent, or less than approximately 0.1 percent).“)
Regarding claim 24, Li and Denli teach the system of claim 18, Li further teaches
generating the synthetic data using a physics based simulation, and wherein the synthetic data is generated to mimic real world regional data (Li, paragraph 0030, “In an example embodiment, the simulation component 120 may rely on entities 122. Entities 122 may include earth entities or geological objects such as wells, surfaces, bodies, reservoirs, etc. In the system 100, the entities 122 can include virtual representations of actual physical entities that are reconstructed for purposes of simulation”).
Regarding claim 25, Li and Denli teach the system of claim 18, Denli further teaches
the particular inference task comprises wave picking to identify at least one of: a geographic fault; a geographic layer; P-wave arrival; S-wave arrival; de-noising; synthetic data generation; horizon picking; event identification; or a location of a subsurface feature (Denli, paragraph 0027, “ ’Seismic data’ is also intended to include any data (e.g., seismic image, migration image, reverse-time migration image, pre-stack image, partially-stack image, full-stack image, poststack image or seismic attribute image) or interpretation quantities, including geophysical properties such as one or more of: elastic properties (e.g., P and/or S wave velocity, P- Impedance, S-Impedance, density, attenuation, anisotropy and the like); and porosity, permeability or the like, that the ordinarily skilled artisan at the time of this disclosure will recognize may be inferred or otherwise derived from such data received and/or recorded as part of the seismic surveying and interpretation process” and “Examples of geological features include, without limitation salt, fault, channel, environment of deposition (EoD), facies, carbonate, rock types (e.g., sand and shale), horizon, stratigraphy, or geological time” (Denli, paragraph 0029)).
Li and Denli are considered analogous to the claimed invention because they both deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li to include all the types of inference tasks from Denli. One of the ordinary skill in the art would have known to apply the known technique of to use the types of inference tasks from Denli to train a seismic machine learning model. Therefore, applying Denli’s technique would yield the predictable results of using seismic inference tasks to improve the interpretation of the data. (See MPEP 2141 (III)(D) Applying a known technique to a known device ready for improvement to yield predicable results).
Regarding claim 26, Li and Denli teach the system of claim 18, Denli further teaches
the first, second, and third datasets are distributed acoustic sensing (DAS) datasets (Denli, paragraph 0028, “ ‘Seismic data’ may also include data derived from traditional seismic (e.g., acoustic) data sets in conjunction with other geophysical data, including, for example, gravity plus seismic; gravity plus electromagnetic plus seismic data, etc.” Examiner notes that the seismic acoustic data is mapped to the distributed acoustic sensing datasets).
Li and Denli are considered analogous to the claimed invention because they both deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li to use the seismic acoustic data from Denli. One of the ordinary skill in the art would have known to apply the known technique of to use seismic acoustic data to train a seismic machine learning model. Therefore, applying Denli’s technique would yield the predictable results of using cost-effective real-world data. (See MPEP 2141 (III)(D) Applying a known technique to a known device ready for improvement to yield predicable results).
Regarding claim 27, Li and Denli teach the system of claim 18, Li further teaches
the first, second, and third datasets are seismic imaging datasets (Li, Abstract, “ A method can include receiving a first trained machine model trained via unsupervised learning using unlabeled seismic image data; receiving labeled seismic image data acquired via an interactive interpretation process; and building a second trained machine model, as initialized from the first trained machine model, via supervised learning using the received labels, where the second trained machine model predicts stratigraphy of a geologic region from seismic image data of the geologic region”).
Regarding claim 28, Li and Denli teach the system of claim 18, Denli further teaches
the first dataset comprises synthetic data (Denli, paragraph 0074, “Embedding space may thus be used in a variety of contexts, such as analyzing different types of datasets (e.g., seismic images (including real or synthetic data), dialogs, etc.), tracking, or cataloging”).
Li and Denli are considered analogous to the claimed invention because they both deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li to use synthetic data from Denli. One of the ordinary skill in the art would have known to apply the known technique of including synthetic data into a dataset. Therefore, applying Denli’s technique would yield the predictable results of reducing cost/time by using synthetic data over real-world data. (See MPEP 2141 (III)(D) Applying a known technique to a known device ready for improvement to yield predicable results).
Regarding claim 35, Li teaches
A non-transitory computer readable storage medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform operations for training a machine learning model, the operations (Li, paragraph 0167, “As shown, the method 900 may be implemented in part via one or more computer-readable storage media (CRM) blocks 91 1 , 921 , 931 , 941 , 951 , 961 , 971 , 981 , 985 and 991 . Such CRM blocks include instructions executable by a processor to instruct a device such as a computing device, a computing system, a controller, etc. A computer-readable storage medium or media (CRM) is or are a non-transitory medium or media that is or are not a carrier wave and not a signal”).
performing region specific training on the initially trained machine learning model using a second dataset (Li, paragraph 00160, “the output can be for a target region (e.g., a selected region) where a second ML model is to be trained in a supervised manner”).
refining the machine learning model using a third dataset to train the machine learning model to perform a particular inference task (Li, paragraph 00154, “an interpreter may decide to refine one or more regions, which may be accomplished by selecting a region, automatically selecting the training data (e.g., as labeled) and adjusting and/or augmenting the training data to retrain a specific ML model for that region, which can then be executed to process the corresponding seismic data, which may generate a refined earth model for that region”).
Li does not teach, but Denli does teach
performing self-supervised learning on a first dataset to initially train the machine learning model (Denli, paragraph 0052, “the initial or first stage may comprise unsupervised (or self-supervised) machine learning with a larger dataset that is configured for pre-training”).
Li and Denli are considered analogous to the claimed invention because they both deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li to use self-supervised learning from Denli. One of the ordinary skill in the art would have known that both unsupervised learning and self-supervised learning serve the same purpose of learning features from unlabeled data. Therefore, substituting Li’s unsupervised learning with Denli’s self-supervised learning would have yield the predictable result of learning feature representation from unlabeled seismic data. (See MPEP 2141 (III)(B) Simple substitution of one known element for another to obtain predictable results).
Claim(s) 12, 15, 29, and 32 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li and Denli in view of He et al. (Masked Autoencoders Are Scalable Vision Learners) (hereafter referred to as He).
Regarding claim 12, Li and Denli teach the method of claim 1, He teaches
the machine learning model is a masked autoencoder network (He, abstract, “This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels”).
Li, Denli, and He are considered analogous to the claimed invention because they all deal with self-supervised learning on data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li and Denli to use a masked autoencoder from He. One of the ordinary skill in the art would have known to apply the technique of masking random data to train a machine learning model. He teaches that “Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pre training and shows promising scaling behavior” (He, abstract). (See MPEP 2141 (III)(G) Some Teaching, Suggestion, or Motivation in the Prior Art That Would Have Led One of Ordinary Skill To Modify the Prior Art Reference or To Combine Prior Art Reference Teachings To Arrive at the Claimed Invention).
Regarding claim 15, Li and Denli teach the method of claim 1 and 12, He teaches
training data to the masked autoencoder network is masked in rectangles or cuboids (He, Figure 1, “During pre-training, a large random subset of image patches (e.g., 75%) is masked out. The encoder is applied to the small subset of visible patches. Mask tokens are introduced after the encoder, and the full set of encoded patches and mask tokens is processed by a small decoder that reconstructs the original image in pixels. After pre-training, the decoder is discarded and the encoder is applied to uncorrupted images (full sets of patches) for recognition tasks”.
PNG
media_image1.png
192
329
media_image1.png
Greyscale
Examiner notes that the patches being masked are in a rectangular shape).
Regarding claim 29, Li and Denli teach the system of claim 18, He teaches
the machine learning model is a masked autoencoder network (He, abstract, “This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels”).
Li, Denli, and He are considered analogous to the claimed invention because they all deal with self-supervised learning on data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li and Denli to use a masked autoencoder from He. One of the ordinary skill in the art would have known to apply the technique of masking random data to train a machine learning model. He teaches that “Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pre training and shows promising scaling behavior” (He, abstract). (See MPEP 2141 (III)(G) Some Teaching, Suggestion, or Motivation in the Prior Art That Would Have Led One of Ordinary Skill To Modify the Prior Art Reference or To Combine Prior Art Reference Teachings To Arrive at the Claimed Invention).
Regarding claim 32, Li and Denli teach the system of claim 18 and 29, He teaches
training data to the masked autoencoder network is masked in rectangles or cuboids (He, Figure 1, “During pre-training, a large random subset of image patches (e.g., 75%) is masked out. The encoder is applied to the small subset of visible patches. Mask tokens are introduced after the encoder, and the full set of encoded patches and mask tokens is processed by a small decoder that reconstructs the original image in pixels. After pre-training, the decoder is discarded and the encoder is applied to uncorrupted images (full sets of patches) for recognition tasks”.
PNG
media_image1.png
192
329
media_image1.png
Greyscale
Examiner notes that the patches being masked are in a rectangular shape).
Claim(s) 13 and 30 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li, Denli, and He in view of Sun et al. (WO 2022079550) (hereafter referred to as Sun).
Regarding claim 13, Li, Denli, and He teach the method of claims 1 and 12, Sun teaches
the masked autoencoder network is configured to receive two dimensional input, and wherein the two dimensions comprise time and channel (Sun, paragraph 0032, “BERT was developed for pretraining sentences in NLP. Thus, this neural network needs to be modified to be applicable to waveform data 302, which includes plural traces 210-i, as illustrated in Figure 2. Because each trace of the traces 210-i is associated with a unique distance that characterizes the distance between the source 104 and a corresponding receiver 118 that recorded the trace 210-i, it is considered that there is a specific order in which the traces 210-i are arranged in Figure 2, i.e., first is the trace 210 recorded closest to the source 104, then the next trace is further away from the source 104, and so on until the last trace 212, which is the furthest from the source 104 is plotted in Figure 2.” And “Figure 2 shows an image 200 generated based on such recorded wave signals or traces 210-i (note that there are thousands of traces plotted in this figure, one next to the other, the first trace being 210 and the last one in the series of traces being 212). In the figure, the X-axis denotes the offsets (the distance between the source 104 and the receivers 118), while the Y-axis denotes the time of recording” (Sun, paragraph 0006). Examiner notes that the traces are mapped to channels and since there are a plurality of traces being inputted, there is a time dimension).
Li, Denli, He, and Sun are considered analogous to the claimed invention because they all deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li, Denli, and He to use a 2D seismic input from Sun. One of the ordinary skill in the art would have known to apply the technique of a 2D seismic dataset to train a seismic machine learning model. Sun teaches that “it is necessary to adapt BERT (which is a language analyzing neural networks) to pretraining of the waveform dataset, which can be achieved in this embodiment by replacing the mapping of a sentence to another sentence, with the mapping of one record of waveform data to itself, and the words in a sentence of the original BERT algorithm are replaced by traces of the waveform data 302.” (Sun, paragraph 0032). (See MPEP 2141 (III)(G) Some Teaching, Suggestion, or Motivation in the Prior Art That Would Have Led One of Ordinary Skill To Modify the Prior Art Reference or To Combine Prior Art Reference Teachings To Arrive at the Claimed Invention).
Regarding claim 30, Li, Denli, and He teach the system of claims 18 and 29, Sun teaches
the masked autoencoder network is configured to receive two dimensional input, and wherein the two dimensions comprise time and channel (Sun, paragraph 0032, “BERT was developed for pretraining sentences in NLP. Thus, this neural network needs to be modified to be applicable to waveform data 302, which includes plural traces 210-i, as illustrated in Figure 2. Because each trace of the traces 210-i is associated with a unique distance that characterizes the distance between the source 104 and a corresponding receiver 118 that recorded the trace 210-i, it is considered that there is a specific order in which the traces 210-i are arranged in Figure 2, i.e., first is the trace 210 recorded closest to the source 104, then the next trace is further away from the source 104, and so on until the last trace 212, which is the furthest from the source 104 is plotted in Figure 2.” And “Figure 2 shows an image 200 generated based on such recorded wave signals or traces 210-i (note that there are thousands of traces plotted in this figure, one next to the other, the first trace being 210 and the last one in the series of traces being 212). In the figure, the X-axis denotes the offsets (the distance between the source 104 and the receivers 118), while the Y-axis denotes the time of recording” (Sun, paragraph 0006). Examiner notes that the traces are mapped to channels and since there are a plurality of traces being inputted, there is a time dimension).
Li, Denli, He, and Sun are considered analogous to the claimed invention because they all deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li, Denli, and He to use a 2D seismic input from Sun. One of the ordinary skill in the art would have known to apply the technique of a 2D seismic dataset to train a seismic machine learning model. Sun teaches that “it is necessary to adapt BERT (which is a language analyzing neural networks) to pretraining of the waveform dataset, which can be achieved in this embodiment by replacing the mapping of a sentence to another sentence, with the mapping of one record of waveform data to itself, and the words in a sentence of the original BERT algorithm are replaced by traces of the waveform data 302.” (Sun, paragraph 0032). (See MPEP 2141 (III)(G) Some Teaching, Suggestion, or Motivation in the Prior Art That Would Have Led One of Ordinary Skill To Modify the Prior Art Reference or To Combine Prior Art Reference Teachings To Arrive at the Claimed Invention).
Claim(s) 14 and 31 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li, Denli, and He in view of Zhu et al. (Seismic Signal Denoising and Decomposition Using Deep Neural Network) (hereafter referred to as Zhu).
Regarding claim 14, Li, Denli, and He teach the method of claims 1 and 12, Zhu teaches
the masked autoencoder network is configured to receive three dimensional input, and wherein the three dimensions comprise time, channel, and frequency (Zhu, Figure 1,
PNG
media_image2.png
301
705
media_image2.png
Greyscale
Examiner notes that the input is in 3D with frequency bin, time points, and channels).
Li, Denli, He, and Zhu are considered analogous to the claimed invention because they all deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li, Denli, and He to use a 3D seismic data as input from Zhu. One of the ordinary skill in the art would have known to apply the known technique of using a 3D seismic data to train a seismic machine learning model. Therefore, applying Zhu’s technique would yield the predictable results of using 3D seismic data reduce exploration risk and increase accuracy. (See MPEP 2141 (III)(D) Applying a known technique to a known device ready for improvement to yield predicable results).
Regarding claim 31, Li, Denli, and He teach the system of claims 18 and 29, Zhu teaches
the masked autoencoder network is configured to receive three dimensional input, and wherein the three dimensions comprise time, channel, and frequency (Zhu, Figure 1,
PNG
media_image2.png
301
705
media_image2.png
Greyscale
Examiner notes that the input is in 3D with frequency bin, time points, and channels).
Li, Denli, He, and Zhu are considered analogous to the claimed invention because they all deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li, Denli, and He to use a 3D seismic data as input from Zhu. One of the ordinary skill in the art would have known to apply the known technique of using a 3D seismic data to train a seismic machine learning model. Therefore, applying Zhu’s technique would yield the predictable results of using 3D seismic data reduce exploration risk and increase accuracy. (See MPEP 2141 (III)(D) Applying a known technique to a known device ready for improvement to yield predicable results).
Claim(s) 17 and 34 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li and Denli in view of Jozinović et al. (Transfer learning: Improving neural network based prediction of earthquake ground shaking for an area with insufficient training data) (hereafter referred to as Jozinović).
Regarding claim 17, Li and Denli teach the method of claim 1, Jozinović teaches
region specific training comprises retraining a subset of layers of the machine learning model (Jozinović, section 4.2, “We achieved the best results when the weights of the first two convolutional layers are used from the pre-trained model with their learning rates set to zero, while the remaining layers are initialized using the Glorot uniform initializer”).
Li, Denli and Jozinović are considered analogous to the claimed invention because they all deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li and Denli to retrain some layers. Jozinović teaches that they “achieved the best results when the weights of the first two convolutional layers are used from the pre-trained model with their learning rates set to zero, while the remaining layers are initialized using the Glorot uniform initializer” and that “this has been done since the first two layers in this architecture extract seismogram features irrespective of the station specifics (geographical pattern, soil, topography, etc.), while the subsequent layers combine the extracted features from individual stations exploiting the network station pattern” (Jozinović, section 4.2). (See MPEP 2141 (III)(G) Some Teaching, Suggestion, or Motivation in the Prior Art That Would Have Led One of Ordinary Skill To Modify the Prior Art Reference or To Combine Prior Art Reference Teachings To Arrive at the Claimed Invention).
Regarding claim 34, Li and Denli teach the system of claim 1, Jozinović teaches
region specific training comprises retraining a subset of layers of the machine learning model (Jozinović, section 4.2, “We achieved the best results when the weights of the first two convolutional layers are used from the pre-trained model with their learning rates set to zero, while the remaining layers are initialized using the Glorot uniform initializer”).
Li, Denli and Jozinović are considered analogous to the claimed invention because they all deal with seismic data. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Li and Denli to retrain some layers. Jozinović teaches that they “achieved the best results when the weights of the first two convolutional layers are used from the pre-trained model with their learning rates set to zero, while the remaining layers are initialized using the Glorot uniform initializer” and that “this has been done since the first two layers in this architecture extract seismogram features irrespective of the station specifics (geographical pattern, soil, topography, etc.), while the subsequent layers combine the extracted features from individual stations exploiting the network station pattern” (Jozinović, section 4.2). (See MPEP 2141 (III)(G) Some Teaching, Suggestion, or Motivation in the Prior Art That Would Have Led One of Ordinary Skill To Modify the Prior Art Reference or To Combine Prior Art Reference Teachings To Arrive at the Claimed Invention).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure. Harsuko et al. (StorSeismic: A New Paradigm in Deep Learning for Seismic Processing) discloses a framework that processing seismic data by using BERT to store and train geometrical features of seismic data. Feichtenhofer el al. (Masked Autoencoders As Spatiotemporal Learners) discloses the use of Masked Autoencoders to spatiotemporal representation learning from videos.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to STEVEN VO whose telephone number is (571)272-9622. The examiner can normally be reached Monday - Friday from 7-3 pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Michelle Bechtold can be reached at (571) 431-0762. 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.
/S.V./Examiner, Art Unit 2148 /MICHELLE T BECHTOLD/Supervisory Patent Examiner, Art Unit 2148
Prosecution Insights