DETAILED ACTION
This Action is responsive to claims filed 06/04/2025.
Notice of Pre-AIA or AIA Status
The present application is being examined under the pre-AIA first to invent provisions.
Status of the Claims
Claims 1-20 have been amended. Claims 1-20 are currently pending.
Response to Arguments
Applicant's arguments, see Pages 12-13, filed 06/04/2025 regarding the 35 U.S.C. 101 Rejection of claims 1-20 have been fully considered but they are not persuasive.
The Applicant argues the independent claims are analogous to Example 39. The Examiner respectfully disagrees with the Applicant. Applying transformations and creating datasets of digital facial images is not practically performed within the human mind or with the aid of pen and paper. On the contrary, selecting a set of samples, determining the uncertainty produced by a machine learning model, calculating a ranking score, selecting a subset of samples, and sorting the samples are all steps recited at a level of generality that does not preclude their execution by a human mind or one aided with pen and paper. See the updated 35 U.S.C. 101 Rejection below.
Applicant's arguments, see Pages 13-14, filed 06/04/2025 regarding the 35 U.S.C. 102(a)(2) Rejection of claims 1-4, 7-18, and 20 have been fully considered but they are not persuasive.
The Examiner respectfully disagrees with the Applicant that the cited reference Gokalp does not teach the newly-amended limitations of the independent claims. Gokalp, in at least Columns 6-8, 15, and 45, reads on utilizing a web interface (which would reasonably utilize known URL technology) to present a ranked list, track when a training epoch has concluded with a client, and receiving a report of at the conclusion of an epoch. The amendments to the remaining dependent claims listed were not substantive. See the updated 35 U.S.C. 102(a)(2) Rejection below.
Applicant’s arguments with respect to claim(s) 6 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
By way of dependency, the 35 U.S.C. 102(a)(2) Rejection of claim 6 has been withdrawn. However, new grounds of rejection have been made under 35 U.S.C. 103. See the updated 103 Rejection below.
Applicant's arguments, see Pages 15, filed 06/04/2025 regarding the 35 U.S.C. 103 Rejection of claim 5 have been fully considered but they are not persuasive.
The amendments to claim 5 are not substantive enough to differentiate them from the previously filed claims. In the absence of other arguments, and given that the 102(a)(2) Rejection of claim 1 is to be maintained, the 35 U.S.C. 103 Rejection of Claim 5 is also maintained. See the updated 103 Rejection below.
Claim Rejections - 35 USC § 101
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more; and because the claims as a whole, considering all claim elements both individually and in combination, do not amount to significantly more than the abstract idea, see Alice Corporation Pty. Ltd. v. CLS Bank International, et al, 573 U.S. (2014). In determining whether the claims are subject matter eligible, the Examiner applies the 2019 USPTO Patent Eligibility Guidelines. (2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50, Jan. 7, 2019.)
Step 1:
Claims 1-15 recite a system for training a machine learning model, which falls under the statutory category of a machine. Claims 16-19 recite a method for training a machine learning model, which falls under the statutory category of a process. Claim 20 recites a non-transitory computer readable medium embodying programmed instructions executed by a processor, wherein the instructions direct the processor to implement a method for training a machine learning model, which falls under the statutory category of a manufacture.
Step 2A – Prong 1:
Claim 1 recites an abstract idea, law of nature, or natural phenomenon. The limitations of “selecting a set of samples that are unclassified,”, “determining an uncertainty of the labels predicted for the samples by the machine learning model,”, “calculating a ranking score for each of the samples in the set based at least on the uncertainty for a corresponding prediction for a label,”, “selecting a subset of the samples that have more than a threshold ranking score,”, “sorting the samples of the subset into a ranked list,” under the broadest reasonable interpretation, cover a mental process including an observation, evaluation, judgment or opinion that could be performed in the human mind or with the aid of pencil and paper.
Step 2A – Prong 2:
The additional elements of claim 1 do not integrate the abstract idea into a judicial exception. The claim recites the additional elements “a system”, “a processor”, “memory”, “computer program code”, “a web interface”, “clients”, “tracking logic”, and “a report” are recognized as generic computer components recited at a high level of generality. Although they have and execute instructions to perform the abstract idea itself, this also does not serve to integrate the abstract idea into a practical application as it merely amounts to instructions to "apply it." (See MPEP 2106.04(d)(2) indicating mere instructions to apply an abstract idea does not amount to integrating the abstract idea into a practical application).
The additional elements of “a machine learning model”, “training data”, “labels”, “a ranking score”, and “uncertainty for a corresponding prediction for a label” are recognized as non-generic computer components, but are recited at a high level of generality and are found to generally link the abstract idea to a particular technological environment or field of use (See MPEP 2106.05(h)).
The additional elements recited in the limitation “utilize training data to train the machine learning model across multiple epochs;”, “operating the machine learning model to predict labels for the samples,”, “utilizing a web interface to submit the samples of the subset as labeling tasks to labeling clients for replacement labels selected in an order based on the ranked list…”, “implementing tracking logic to track a status of the labeling tasks sent to the labeling clients during the epoch to generate the replacement labels;”, and “train the machine learning model in another epoch using the subset of the samples and the corresponding replacement labels as the additional training data.” are found to be mere instructions to apply the abstract idea (See MPEP 2106.04(d)(2) indicating mere instructions to apply an abstract idea does not amount to integrating the abstract idea into a practical application).
The additional elements recited in the limitations “receive a request to train the machine learning model;”, “receive the replacement labels from the client;”, and “receive a report that the epoch of training has completed for the machine learning model“ are found to be insignificant extra solution data retrieval or transmittal steps (See MPEP 2106.05(g)).
Step 2B:
The only limitation on the performance of the described method is a limitation reciting “a system”, “a processor”, “memory”, “computer program code”, “a web interface”, “clients”, “tracking logic”, and “a report”. These elements are insufficient to transform a judicial exception to a patentable invention because the recited elements are considered insignificant extra-solution activity (generic computer system, processing resources, links the judicial exception to a particular, respective, technological environment). The claim thus recites computing components only at a high-level of generality such that it amounts to no more than mere instructions to apply the exception using generic computer components; mere instructions to apply an exception using a generic computer component cannot provide an inventive concept (see MPEP 2106.05(f)).
The additional elements of “a machine learning model”, “training data”, “labels”, “a ranking score”, and “an uncertainty for a corresponding prediction for a label” are recognized as non-generic computer components, but are recited at a high level of generality and are found to generally link the abstract idea to a particular technological environment or field of use (See MPEP 2106.05(h)).
The additional elements recited in the limitation “utilize training data to train the machine learning model across multiple epochs;”, “operating the machine learning model to predict labels for the samples,”, “utilizing a web interface to submit the samples of the subset as labeling tasks to labeling clients for replacement labels selected in an order based on the ranked list…”, “implementing tracking logic to track a status of the labeling tasks sent to the labeling clients during the epoch to generate the replacement labels;”, and “train the machine learning model in another epoch using the subset of the samples and the corresponding replacement labels as the additional training data.” is found to be mere instructions to apply the abstract idea (See MPEP 2106.04(f) indicating mere instructions to apply an abstract idea does not recite significantly more than the judicial exception).
In addition, the claimed “receive a request to train the machine learning model;”, “receive the replacement labels from the client;”, and “receive a report that the epoch of training has completed for the machine learning model“ are acknowledged to be well-understood, routine, conventional activity (see, e.g., court recognized WURC examples in MPEP 2106.05(d)(II)(i)).
Taken alone or in ordered combination, these additional elements do not amount to significantly more than the above-identified abstract idea. There is no indication that the combination of elements improves the functioning of a computer or improves any other technology. Their collective functions merely provide conventional computer implementation.
For the reasons above, claim 1 is rejected as being directed to non-patentable subject matter under §101. This rejection applies equally to independent claims 16 and 20.
Claim 16 recites similar limitations to claim 1, with the inclusion of additional elements “A method for training a machine learning model” (generic computer components and additional elements that generally link).
Claim 20 recites similar limitations to claim 1, with the inclusion of generic computer component additional element “A non-transitory computer readable medium embodying programmed instructions executed by a processor, wherein the instructions direct the processor to implement a method for training a machine learning model, the method comprising:” (generic computer components and additional elements that generally link).
Dependent Claims:
Claim 2 (claim 17) recites abstract idea mental process steps (“identify a first period of time for performing an epoch of training on the machine learning model; identify a second period of time for a labeling client to generate a replacement label for a sample; and dynamically select a number of the samples to include in the subset by dividing the first period of time by the second period of time to determine an expected number of the samples that a labeling client is capable of generating replacement labels for during the epoch.”).
Claim 3 (claim 18) recites an abstract idea mental process step (“modify the number of the samples to include in the subset based on a number of the labeling clients.”).
Claim 4 (claim 19) recites a mere data transmittal step (“flush data of the tracking logic at the end of the epoch.”).
Claim 5 recites mere data transmittal or manipulation steps (“submit the subset to the labeling clients by adding the samples from the subset to a buffer; and flush the buffer in response to the machine learning model completing the epoch of training.”).
Claim 6 recites mere data retrieval or transmittal steps (“transmitting an identifier for a first sample in the buffer to the labeling client via the web interface; receiving a replacement label for the first sample from the labeling client via the web interface: and transmitting the identifier for a next sample in the buffer to the labeling client via the web interface”).
Claim 7 recites mere extra solution activity steps (“create, for each of the labeling tasks, a uniform resource locator enabling the labeling clients to request information about the samples.”)
Claim 8 recites a mere data transmittal step (“submit the samples of the subset individually to the labeling clients”)
Claim 9 recites mere extra solution activity steps (“generate a graphical user interface for tracking at least one of performance or budget used during training of the machine learning model”).
Claim 10 recites an abstract idea mental process step (“selecting a predefined number of the samples having a highest amount of uncertainty”).
Claim 11 recites instructions to apply the abstract idea (“initiate the training and submission of the subset to the labeling clients for the replacement labels, in response to the request that defines classes for the labels, includes a pointer to the set of the samples, and defines an end condition for halting training of the machine learning model and halting labeling of the samples.”).
Claim 12 recites mere extra solution activity steps (“activate a model-optimization routine for the machine learning model in response to determining that a change in performance of the machine learning model across the epochs is less than a threshold amount.”).
Claim 13 recites abstract idea mental process steps (“determining a score for each object depicted within a sample; and aggregating the scores for the objects within the sample.”).
Claim 14 recites an abstract idea mental process step (“supplement the training data for the machine learning model with the subset of the samples, wherein the labels predicted by the machine learning model for the subset are replaced with corresponding replacement labels from the labeling client;”) and instructions to apply (“and train the machine learning model with the training data that was supplemented.”).
Claim 15 recites an abstract idea mental process step (“supplement prior, already labeled training data for the machine learning model with the subset of the samples, and utilize a whole training set comprising the subset and the prior training data to train the machine learning model.”).
Claim Rejections - 35 USC § 102
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim(s) 1-4 and 7-20 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Gokalp et al. (US 11,120,364 B1) hereinafter Gokalp.
In regards to claim 1: The present invention claims: “A system for training a machine learning model, the system comprising: at least one processor; and at least one memory including computer program code…” Gokalp teaches “Various embodiments of methods and apparatus for efficient training of machine learning models such as classifiers using an automated workflow comprising intelligently guided labeling feedback sessions are described.” (Column 1, Lines 49-52, See Column 51 for implementation on hardware)
“receive a request to train the machine learning model;”
“and utilize training data to train the machine learning model across multiple epochs…” Gokalp teaches “The service may support the automation of, among other parts of the workflow, the following steps in at least some embodiments…(f) continuous performance evaluation as training iterations proceed…” (Column 4, Lines 30-31).
“prepare additional training data concurrently with an epoch by: selecting a set of samples that are unclassified,” See Gokalp Fig. 26, 2607-2613 and Column 43 “At a high level, a given training iteration may comprise at least two categories of operations in various embodiments…new labeling candidates may be presented in the labeling sessions in batches…”
“operating the machine learning model to predict labels for the samples,” Gokalp Columns 3 and 4 detail the training of classifiers (Lines 67 and 1-9).
“determining an uncertainty of the labels predicted for the samples by the machine learning model,” Gokalp Column 6 details how a variance or other similar metric (“a predicted classification score to a class boundary…”), including a measure if multiple classifiers “agree” on a class. Mapping such variances or similar metrics to “an uncertainty.”
“calculating a ranking score for each of the samples in the set based at least on the uncertainty for a corresponding prediction for a label,” Gokalp teaches “In at least one embodiment, instead of or in addition to using such a variance metric to rank data items as candidates for labeling feedback, other metrics (such as proximity of a predicted classification score to a class boundary) may be used, and/or other types of active learning approaches may be employed.” (Column 6, Lines 32-34)
“selecting a subset of the samples that have more than a threshold ranking score,” Gokalp teaches “A variety of training completion criteria may be employed in different embodiments---e.g., in some embodiments, training may be considered complete after acceptable results are obtained on a set of diagnosis tests, or when values of selected metrics values meet threshold criteria,” and “A measure of variation among the predictions generated by the different models may be computed, and those data items whose variation measures meet a threshold criterion may be selected as candidates for labeling feedback in the next training iteration…” (Column 6, Lines 1-5 and Lines 15-19).
“sorting the samples of the subset into a ranked list,” See Gokalp Fig. 27, item 2710 and “For example, during a given session in one embodiment, a visualization data set may be presented via the interactive programmatic interface, in which information about several labeling candidate data items is included, with the data items arranged in an order based at least in part on a respective rank assigned to the data items with respect to estimated learning contribution and/or one or more other metrics.” (Column 45, Lines 46-53).
“utilizing a web interface to submit the samples of submitting the subset as labeling tasks to a client labeling clients for replacement labels selected in an order based on the ranked list” Gokalp teaches “In some embodiments, a pool of label providers may be selected, and each such label provider may be presented, via an interactive interface such as web page, with a respective set of data items for which labeling feedback is desired.” (Column 6, Lines 35-38). See Gokalp Fig. 27 2713 and 2716 and “During the feedback session, respective labels for one or more of the presented data items may be obtained, together with a filter criterion to be used to select additional data items for presentation via additional visualization data sets in some embodiments (element 2713). In at least some embodiments, the ranking and/or selection of individual data items for presentation via the interactive interface may thus be based not just on metrics generated and analyzed at the classification service, but also on filter criteria indicated by the label provider,” (Column 45, Lines 58-66).
“receive the replacement labels from the labeling clients… and train the machine learning model in another epoch using the subset of the samples and the corresponding replacement labels as the additional training data.” Gokalp teaches “The set of newly-labeled (or re-labeled) data items may be added to the training set to be used for one or more training iterations (e.g., the next training iteration) in various embodiments (element 2716). As more training iterations are performed, additional labels for data items that have been identified as likely to contribute to faster or more goal-directed learning may thus be used to gradually increase the training set size in various embodiments.” (Column 46, Lines 10-18).
“implementing tracking logic to track a status of the labeling tasks sent to the labeling clients during the epoch to generate the replacement labels;” Gokalp teaches “A variety of training completion criteria may be employed in different embodiments---e.g., in some embodiments, training may be considered complete after acceptable results are obtained on a set of diagnosis tests, or when values of selected metrics values meet threshold criteria, while in other embodiments, the training iterations may simply be terminated when a budget of resources is used up.” (Column 6, Lines 1-7)
“receive a report that the epoch of training has completed for the machine learning model;” Gokalp teaches “In at least some embodiments, in addition to a group of classifiers used to help identify the next set of labeling feedback candidates, a final (with respect to the current training iteration) classifier may also be trained in a given iteration, e.g., using all the labeled training data available, and the results obtained from the final-with-respect-to-the- current-iteration classifier on a test set may be used to evaluate whether quality-related training completion criteria have been met. Of course, training iterations may also be terminated for reasons other than achieving a desired level of classification quality with respect to various measures- e.g., training may be concluded when a budget of resources or time is exhausted in some embodiments, even if all the classification quality goals have not been reached.” (Column 15, lines 30-43). See also Figures 20 and 21, which reasonably read on a visual representation of a “report” when an epoch of training is complete.
In regards to claim 2: The present invention claims: “identify a first period of time for performing an epoch of training on the machine learning model;” See Gokalp Fig. 4, T0…T3… represent training iterations.
“identify a second period of time for a labeling client to generate a replacement label for a sample;” See Gokalp Fig. 4, f1…f10 represent points where the label providers give feedback.
“and dynamically select a number of the samples to include in the subset by dividing the first period of time by the second period of time to determine an expected number of the samples that a labeling client is capable of generating replacement labels for during the epoch.” Gokalp teaches “In at least some embodiments, as the training iterations proceed, the interactions with individual label providers may be analyzed, e.g., to determine which label providers are more proficient in identifying particular classes of data items, to determine the rate at which individual label providers are able to generate labels, and so on.” (Columns 7 and 8, Lines 65-67 and 1-4). Column 6, Lines 57-65 teach how the number of items selected for label feedback may be intentionally curtailed by the system.
In regards to claim 3: The present invention claims: “modify the number of the samples to include in the subset based on a number of the labeling clients.” Gokalp teaches “In some embodiments, a pool of label providers may be selected, and each such label provider may be presented, via an interactive interface such as web page, with a respective set of data items for which labeling feedback is desired.” (Column 6, Lines 35-38) (multiple clients) and lines 57-65 teach “In some embodiments, a number of candidate data items selected for labeling feedback, e.g., using a committee of models, may be subdivided into smaller subsets, e.g., comprising N data items each, and only an individual subset may be presented to a label provider for feedback at a time. For example, from among 200 data items for which labels are desired, 10 items may be presented to a given label provider at a time, so as not to overwhelm the label provider with too many items at once.” Column 41 also goes into how a particular provider’s subset may be customized.
In regards to claim 4: The present invention claims: “flush data of the tracking logic at the end of the epoch.” Gokalp teaches “Rules to determine when a given training iteration is to be considered complete (e.g., when a specified number of epochs is completed, when the difference in a metric value from a previous epoch falls below a threshold, when a specified amount of time has elapsed, when a specified amount of processing resources have been consumed, etc.) may be indicated via iteration completion criteria parameter 2326 in some embodiments.” (Column 40, Lines 15-22) Rules or metric values being different between epochs would reasonably read on “flushing” the tracking logic.
In regards to claim 7: The present invention claims: “create, for each of the labeling tasks, a uniform resource locator enabling the labeling clients to request information about the samples.” See above where Gokalp teaches using a web-based interface, which would reasonably include a URL. Gokalp also teaches “FIG. 13 illustrates an example interactive interface element that provides summarized information about a set of status indicators, according to at least some embodiments. Such an element may be included as part of web-based or other graphical interactive interface of a classification service in various embodiments, as also indicated in FIG. 7. An indication of the last update to the classifier ( e.g., when the training of the iteration-final classifier of the most recent training iteration was completed) may be provided in an update timestamp/history information element 1302 of the interface in the depicted embodiment. A history request button 1304 may be used to submit a request for historical information pertaining to one or more classification model metrics; examples of the kinds of historical information that may be presented in some embodiments are provided below, e.g., in the context of FIG. 14.” (Column 30, Lines 28-43).
In regards to claim 8: The present invention claims: “submit the samples of the subset individually to the labeling clients.” Gokalp Fig. 27, 2713 where one or more labels for individual data items are obtained. See also Fig. 4, where each batch may be one data item and therefore each request for a label and label would be one-by-one.
In regards to claim 9: The present invention claims: “generate a graphical user interface for tracking at least one of performance or budget used during training of the machine learning model.” See Gokalp Figures 19-21. See above where the metric may include a budget.
In regards to claim 10: The present invention claims: “selecting a predefined number of the samples having a highest amount of uncertainty.” See above where Gokalp teaches ranking the data items to receive label feedback for, and where the batches of requests may be constrained in volume (For example, Column 6, Lines 57-65 would always yield the top 10-ranked items).
In regards to claim 11: The present invention claims: “initiate the training and submission of the subset to the labeling clients for the replacement labels, in response to the request that defines classes for the labels, includes a pointer to the set of the samples, and defines an end condition for halting training of the machine learning model and halting labeling of the samples.” See Gokalp Figs. 26 and 27 for the initiation (2601/2701), acquisition of data samples (2710), and end condition check to halt the training in the classification system (2616/2719, Column 6, Lines 1-7). Gokalp also teaches various types of memory used (Column 6, and various instruction set architectures (ISAs) in Column 51. (Given the instant application provides no significant detail distinguishing a pointer from a typical memory pointer, and that the listed ISAs taught in Gokalp all include memory pointers, mapping the access of of said samples in some version of RAM to reasonably accommodate the use of memory pointers).
In regards to claim 12: The present invention claims: “activate a model-optimization routine for the machine learning model in response to determining that a change in performance of the machine learning model across the epochs is less than a threshold amount.” Gokalp teaches “The service may support the automation of, among other parts of the workflow, the following steps in at least some embodiments… ( d) optimized training iterations, in which the service may try a variety of models and/or hyper-parameter combinations and select the best among the models and hyperparameter combinations, ( e) justification-based debugging and analysis of the models, (f) continuous performance evaluation as training iterations proceed…” (Column 4, Lines 26-31).
In regards to claim 13: The present invention claims: “calculating of the ranking score for a sample comprises determining a score for each object depicted within a sample; and aggregating the scores for the objects within the sample.” Gokalp teaches “The order in which the data items are arranged may be based at least in part on an estimated rank, with respect to one or more metrics such as an estimated learning contribution, associated with including respective ones of the one or more data items in a training set for one or more training iterations…” See also “A final ordering of the set of data items to be presented as labeling feedback candidates may be performed by a ranker module 308, which may use the search query terms entered by users, the item labels generated by users, the inverted index, and/or the static rank obtained from the active learning subsystem as inputs in the depicted embodiment. As indicated in FIG. 3, the final ranking produced by the ranker 308 may be based on a combination of factors in at least some embodiments, including but not necessarily limited to the static rank 310 obtained from the active learning results, features extracted from terms of search queries 381, additional features extracted from the attributes of the data items, detections of potentially incorrect labels obtained from label providers, and the like. As such, the final ranking may be considered an example of dynamic ranking…” (mapping the ranker module accumulating the static rank and other metrics into a final ranking of one or more data items to “aggregating…”).
In regards to claim 14: The present invention claims: “supplement the training data for the machine learning model with the subset of the samples, wherein the labels predicted by the machine learning model for the subset are replaced with corresponding replacement labels from the labeling clients; and train the machine learning model with the training data that was supplemented.” See Gokalp Column 6, Lines 45-49 and Column 5, Lines 38-45. Gokalp reads on data being labeled or re-labeled and that data being fed back into the classifier in subsequent iterations.
In regards to claim 15: The present invention claims: “supplement prior, already labeled training data for the machine learning model with the subset of the samples, and utilize a whole training set comprising the subset and the prior training data to train the machine learning model.” See Gokalp Column 5, Lines 41-45 and Column 6, Lines 45-49. Gokalp reads on data being labeled or re-labeled and that data being fed back into the classifier in subsequent iterations.
In regards to claims 16-19: Claims 16-19 recite similar limitations to claims 1-4, with the exception of “A method for training a machine learning model, the method comprising:” therefore both sets of claims are similarly rejected.
In regards to claim 20: Claim 20 recite similar limitations to claims 1, with the exception of “A non-transitory computer readable medium embodying programmed instructions executed by a processor, wherein the instructions direct the processor to implement a method for training a machine learning model, the method comprising:” therefore both sets of claims are similarly rejected.
Claim Rejections - 35 USC § 103
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
The factual inquiries for establishing a background for determining obviousness under pre-AIA 35 U.S.C. 103(a) 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 under pre-AIA 35 U.S.C. 103(a), the examiner presumes that the subject matter of the various claims was commonly owned at the time any inventions covered therein were made absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and invention dates of each claim that was not commonly owned at the time a later invention was made in order for the examiner to consider the applicability of pre-AIA 35 U.S.C. 103(c) and potential pre-AIA 35 U.S.C. 102(e), (f) or (g) prior art under pre-AIA 35 U.S.C. 103(a).
Claims 5 and 6 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gokalp et al. (US 11,120,364 B1) hereinafter Gokalp and Park et al. (US 20220197834 A1) hereinafter Park.
In regards to claim 5: The present invention claims: “submit the subset to the labeling clients by adding the samples from the subset to a buffer; and flush the buffer in response to the machine learning model completing the epoch of training.” While Gokalp teaches the use of memory throughout their disclosure, including various types of RAM (Column 51, SRAM or SDRAM typically contains a buffer); that their embodiments may throttle or bottleneck the amount of data items for labeling, and how the user may set said batch size (Column 6-7, Lines 60-67 and 1-3); and each batch of samples and corresponding label feedback being different each training iteration (Fig. 4), Gokalp fails to explicitly teach “…adding the samples from the subset to a buffer; and flush the buffer…” However, Park, in the field of memory management for a machine learning model, teaches “The fetcher 100 may reduce a number of memory accesses by reusing data. The fetcher 100 may share memory bandwidth resources with one or more executers 150, thereby alleviating a memory bottleneck phenomenon. The reuse buffer 130 may be a space in which input data read from the loader 110 is stored. The input data may include the input feature map 101.
The buffer controller may calculate an address in which input data is to be stored, may write the input data on the calculated address, and may flush an assigned address. For example, the buffer controller may calculate an address of the reuse buffer 130 to be assigned, based on a load counter, a loading unit, a size of the reuse buffer 130, and a quantity of data to be shared. The buffer controller may calculate an address in which flushing is to be performed based on a send counter of each sender 140.” ([0051] and [0052])
Park highlights how the quantity of training data can require more advanced methods of memory management for machine learning models and training ([0003]-[0005]). It would have been obvious to one of ordinary skill in the art at the time of the applicant’s filing to incorporate similar memory management technology with the methods of Gokalp to efficiently manage training data and label requests.
In regards to claim 6: The present invention claims: “transmitting an identifier for a first sample in the buffer to the labeling client via the web interface; receiving a replacement label for the first sample from the labeling client via the web interface: and transmitting the identifier for a next sample in the buffer to the labeling client via the web interface” See above where Gokalp Fig. 27, 2713 reads on one or more labels for individual data items are obtained. See also Fig. 4, where each batch may be one data item and therefore each request for a label and label would be one-by-one. It would be reasonable to one of ordinary skill in the art when combining the methods of Gokalp with the memory buffer implementations of Park to include identifiers with samples or labels when communicating with a client.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GRIFFIN T BEAN whose telephone number is (703)756-1473. The examiner can normally be reached M - F 7:30 - 4:30.
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, Li Zhen can be reached at (571) 272-3768. 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.
/GRIFFIN TANNER BEAN/ Examiner, Art Unit 2121
/Li B. Zhen/ Supervisory Patent Examiner, Art Unit 2121