Prosecution Insights
Last updated: November 04, 2026
Application No. 18/214,024

Systems and Methods for Programmatic Labeling of Training Data for Machine Learning Models via Clustering

Final Rejection §101§103§112
Filed
Jun 26, 2023
Priority
Jun 28, 2022 — provisional 63/356,407
Examiner
HADDAD, MAJD MAHER
Art Unit
2125
Tech Center
2100 — Computer Architecture & Software
Assignee
Snorkel AI Inc.
OA Round
2 (Final)
78%
Grant Probability
Favorable
3-4
OA Rounds
0m
Est. Remaining
99%
With Interview

Examiner Intelligence

Grants 78% — above average
78%
Career Allowance Rate
7 granted / 9 resolved
+22.8% vs TC avg
Strong +89% interview lift
Without
With
+88.9%
Interview Lift
resolved cases with interview
Typical timeline
3y 3m
Avg Prosecution
20 currently pending
Career history
40
Total Applications
across all art units

Statute-Specific Performance

§101
35.1%
-4.9% vs TC avg
§103
42.6%
+2.6% vs TC avg
§102
2.7%
-37.3% vs TC avg
§112
16.9%
-23.1% vs TC avg
Black line = Tech Center average estimate • Based on career data from 9 resolved cases

Office Action

§101 §103 §112
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 . This action is in response to the amendment and remarks filed September 4th, 2026. In the amendment, claims 1, 8-9, 14-15, and 20 were amended, no claims were cancelled or added. As such, claims 1, 8-9, 14-15, and 20 are pending. Response to Arguments Applicant’s arguments, see Pages 9-10, filed September 4th, 2026, with respect to the drawing objections and the rejected claims 1-9, 14-15, and 20 under the double patenting and 112(b) and (d) rejections have been fully considered and are persuasive. The drawing objections and the rejected claims 1-9, 14-15, and 20 under double patenting and 112(b) and (d) have been withdrawn. See updated 112(b) rejections for claims 5, 12, and 18. Applicant’s arguments with respect to the 101 and 103 rejections are not persuasive for the following reasons: 35 U.S.C 101: Applicant argues that the claims do not recite a mental process at Step 2A Prong 1 because the claims recite a specific technical process involving generating real-valued representations, clustering, training classifiers, and determining most likely clusters, none of which can practically be performed in the human mind (Page 10 of Remarks). The Examiner respectfully disagrees. As set forth in the updated rejection, the generating, forming/clustering, representing, determining, and assigning steps each fall within the mental process grouping because they involve observation, evaluation, and judgment that can be performed in the human mind or with pen and paper. Applicant argues that at Step 2A Prong 2 the claims integrate any abstract idea into a practical application since the specification describes technical problems with alternative approaches such as centroid or distance-based approaches problems of dimensionality and underperformance (Page 10 of Remarks). The Examiner respectfully disagrees. The alleged improvement described in the specification is not reflected in the claim language which recites the analysis, clustering, and labeling operations generically and does not recite any specific technique that addresses the dimensionality problem. An improvement described only in the specification but not captured in the claim cannot integrate the exception. The additional elements of training a classifier, using the datapoint as input, storing the classifier, and training a model are mere instructions to apply the judicial exception, insignificant extra-solution activity, and generic data gathering and output. Applicant argues that the ordered combination at Step 2B was not well-understood, routine, or conventional (WURC), and that no evidence has been shown that the specific technical steps or their combination was WURC (Page 10 of Remarks). The Examiner respectfully disagrees. As set forth in the updated Step 2B analysis, the training of a classifier using in-cluster datapoints as positive training examples and out of cluster datapoints as negative training examples is a well-understood, routine, and conventional supervised classification operation, supported by Mohamed (US 20130336579 A1) at paragraphs 3 and 19, satisfying the requirement of MPEP 2106.05(d) and Berkheimer. The remaining additional elements are recited at a high level of generality and amount to applying the exception, gathering data, storing data, and outputting data. Therefore, the claim as a whole remains directed to the abstract idea without significantly more. 35 U.S.C 103: Applicant argues that the cited references do not teach the amended limitations requiring that each trained classifier generate a prediction indicating a likelihood that the new datapoint belongs to the associated cluster or group, and that the most likely cluster or group be determined based on the predictions from the plurality of trained classifiers (Page 11 of Remarks). The Examiner respectfully disagrees. As set forth in the rejection, Rifkin runs all N trained binary classifiers on each new example and selects the classifier that outputs the largest (most positive) value, and teaches that a classifier's output is larger when the example is more likely to belong to its class (Figure 1 caption). This reads on the claimed per-classifier membership likelihood and the selection of the most likely cluster from the predictions of the plurality of classifiers, with Rifkin applied in combination with Xie's clusters as set forth in the rejection. Therefore, the argument is not persuasive and claims 1-20 remain rejected under 35 U.S.C. 103. Claim Objections Claims 5, 9-14, 18, and 20 are objected to because of the following informalities: Claim 9: "generate one or more real-valued representation" should read "generate one or more real-valued representations." Claims 14 and 20: "a plurality of real-value representations" should "a plurality of real-valued representations." Claims 5, 12, and 18: "determining a most likely cluster or group to which the new datapoint should be assigned further comprises...” should read "determining a most likely cluster or group the new datapoint should be assigned further comprises...” Appropriate correction is required. Claim Rejections - 35 USC § 112 The following is a quotation of 35 U.S.C. 112(b): (b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention. The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph: The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention. Claims 5, 12, and 18 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. Claims 5, 12, and 18 recite the limitation "… the trained classifier having the highest level of certainty in its output". There is insufficient antecedent basis for this limitation in the claim. 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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. Claim 1 Step 1: The claim recites a method; therefore, it is directed to the statutory category of a process. Step2A Prong 1: The claim recites, inter alia: generating one or more real-valued representations for each datapoint in a dataset: This limitation is seen as a mental process because it involves analyzing information and creating representations of a datapoint by expressing it numerically, which can be performed in the human mind. based on a similarity between the generated representations, forming one or more groups or clusters of datapoints: This limitation recites a mental process because it involves grouping representations into clusters, which can be performed in the human mind or by pen and paper. representing each formed group or cluster by a unique identifier: This limitation recites a mental process because it assigns a label to each group of clusters, which is mentally performable. for each new datapoint… determining a most likely cluster or group to which the new datapoint is assigned based on the predictions…: This is a mental process because it involves the determination of which group the new datapoint belongs to. assigning a label to the new datapoint based on the identifier of the cluster or group to which the new datapoint is assigned…: This is a mental process because it applies a label to an item based on a comparison, which can be performed in the human mind. Step2A Prong 2: This judicial exception is not integrated into a practical application because the additional elements are as follows: for each group or cluster, training a classifier of a plurality of classifiers to classify a datapoint as either inside or outside the group or cluster, wherein datapoints within the group or cluster are used as positive training examples and datapoints outside the group or cluster are used as negative training examples: Insignificant extra-solution as the limitation amounts to necessary data outputting (MPEP 2106.05(g)(3)). storing each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier: This limitation is merely a post-solution step of storing the data—a nominal addition to the claim that does not meaningfully limit the claim. The method storing is recited at a high level of generality. Simply implementing the abstract idea in a generic method is not a practical application of the abstract idea. Therefore, storing step is an insignificant extra-solution activity. See MPEP 2106.05(g). using the new datapoint as input to each trained classifier: Mere data gathering recited at a high level of generality, and thus is an insignificant extra-solution activity (MPEP 2106.05(g)). wherein each trained classifier generates a prediction indicating a likelihood that the new datapoint belongs to the associated cluster or group… the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). and using a plurality of new datapoints and the new datapoints' assigned labels to train a machine learning model: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows: for each group or cluster, training a classifier of a plurality of classifiers to classify a datapoint as either inside or outside the group or cluster, wherein datapoints within the group or cluster are used as positive training examples and datapoints outside the group or cluster are used as negative training examples: This falls under Well-Understood, Routine, Conventional activity — see MPEP 2106.05(d). Training a classifier using in-cluster datapoints as positive training examples and out of cluster datapoints as negative training examples is a conventional supervised classification operation, as seen in Mohamed (US 20130336579 A1, 2012). Mohamed treats this as a known technique, describing in paragraph 3 that "[c]onventional appearance based methods of object recognition typically use both positive and negative training." Mohamed further describes in paragraph 19 that "[t]he feature vectors may be used for training according to a standard classifier training algorithm such as Support Vector Machine, or Viola Jones, etc to produce a model file 130." storing each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier: This limitation is merely a post-solution step of storing the data—a nominal addition to the claim that does not meaningfully limit the claim. The method storing is recited at a high level of generality. Simply implementing the abstract idea in a generic method is not a practical application of the abstract idea. Therefore, storing step is an insignificant extra-solution activity. See MPEP 2106.05(g). using the new datapoint as input to each trained classifier: The additional element of “using” does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of receiving steps amounts to no more than mere data gathering. This element amounts to receiving data over a network and is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II (i). This cannot provide an inventive concept. wherein each trained classifier generates a prediction indicating a likelihood that the new datapoint belongs to the associated cluster or group… the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea and cannot provide inventive concept (MPEP 2106.05(f)). wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea and cannot provide inventive concept (MPEP 2106.05(f)). and using a plurality of new datapoints and the new datapoints' assigned labels to train a machine learning model: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea and cannot provide inventive concept (MPEP 2106.05(f)). The elements in combination as an ordered whole still do not amount to significantly more than the judicial exception (i.e., the abstract ideas of mental processes for generating representations, grouping data based on similarity, and assigning identifiers). The claim merely describes a process of analyzing data, forming clusters based on similarity comparisons, labeling those clusters, and assigning new datapoints to the most likely cluster which constitute abstract idea of data evaluation and organization. The recitation of training a machine learning model using classifiers, receiving datapoints as input, and storing data merely indicates a technological environment in which the abstract ideas are applied, without improving the functioning of a computer or any machine learning technology itself. Therefore, the claim as a whole remains focused on the abstract idea and fails Step 2B of the eligibility analysis. Claim 2 Step 1: A process, as above. Step2A Prong 1: The claim recites, inter alia: the real-valued representation for each datapoint in a dataset is generated by an embedding process: This limitation is a mental process because it involves generating a representation of data and expressing characteristics of the datapoint numerically, which is mentally performable. Step 2A Prong Two and Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible. Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible. Claim 3 Step 1: A process, as above. Step2A Prong 1: The claim recites, inter alia: the embedding process is a text embedding process: This limitation is a mental process because it involves analyzing textual information and representing it using numbers. Step 2A Prong Two and Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible. Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible. Claim 4 Step 1: A process, as above. Step2A Prong 1: The claim recites, inter alia: the unique identifier is based on one or more attributes of a datapoint or datapoints in the cluster or group: This limitation is a mental process because it involves evaluating characteristics of information and assigning a label based on the characteristics of the group. Step 2A Prong Two and Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible. Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible. Claim 5 Step 1: A process, as above. Step2A Prong 1: The claim recites, inter alia: determining a most likely cluster or group to which the new datapoint should be assigned further comprises determining the cluster or group associated with the trained classifier having the highest level of certainty in its output: This limitation is a mental process because it involves the determination of which group a datapoint should be assigned to and determining which group has the highest level of certainty, which is mentally performable. Step 2A Prong Two and Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible. Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible. Claim 6 Step 1: A process, as above. Step2A Prong 1: The claim recites, inter alia: the similarity between the generated representations is determined based on a metric: This limitation recites a mathematical concept because it uses a math equation to calculate the similarity between representations. Step 2A Prong Two and Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible. Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible. Claim 7 Step 1: A process, as above. Step2A Prong 1: The claim recites, inter alia: the metric is one of Manhattan distance, Euclidean distance, or Cosine distance: This limitation recites a mathematical concept because it uses a math equation to calculate the similarity between representations. Step 2A Prong Two and Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible. Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible. Claim 8 Step 1: A process, as above. Step2A Prong 1: This claim does not recite an additional abstract idea, but the claim depends on claim 1, which recites an abstract idea. Step 2A Prong Two: This judicial exception is not integrated into a practical application because the additional elements are as follows: the one or more real-valued representations for each datapoint in the dataset comprise a plurality of real-valued representations for each datapoint in the dataset: The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)). Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows: the one or more real-valued representations for each datapoint in the dataset comprise a plurality of real-valued representations for each datapoint in the dataset: The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself which cannot provide inventive concept (MPEP 2106.05(h)). Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible. Claim 9 Step 1: The claim recites a system; therefore, it is directed to the statutory category of a machine. Step2A Prong 1: The claim recites, inter alia: generate one or more real-valued representations for each datapoint in a dataset: This limitation is seen as a mental process because it involves analyzing information and creating representations of a datapoint by expressing it numerically, which can be performed in the human mind. based on a similarity between the generated representations, form one or more groups or clusters of datapoints: This limitation recites a mental process because it involves grouping representations into clusters, which can be performed in the human mind or by pen and paper. represent each formed group or cluster by a unique identifier: This limitation recites a mental process because it assigns a label to each group of clusters, which is mentally performable. for each new datapoint… determine a most likely cluster or group to which the new datapoint is assigned based on the predictions…: This is a mental process because it involves the determination of which group the new datapoint belongs to. assign a label to the new datapoint based on the identifier of the cluster or group to which the new datapoint is assigned…: This is a mental process because it applies a label to an item based on a comparison, which can be performed in the human mind. Step2A Prong 2: This judicial exception is not integrated into a practical application because the additional elements are as follows: [a] system, comprising: one or more electronic processors configured to execute a set of computer-executable instructions; and one or more non-transitory electronic data storage media containing the set of computer- executable instructions, wherein when executed, the instructions cause the one or more electronic processors: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). for each group or cluster, training a classifier of a plurality of classifiers to classify a datapoint as either inside or outside the group or cluster, wherein datapoints within the group or cluster are used as positive training examples and datapoints outside the group or cluster are used as negative training examples: Insignificant extra-solution as the limitation amounts to necessary data outputting (MPEP 2106.05(g)(3)). storing each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier: This limitation is merely a post-solution step of storing the data—a nominal addition to the claim that does not meaningfully limit the claim. The method storing is recited at a high level of generality. Simply implementing the abstract idea in a generic method is not a practical application of the abstract idea. Therefore, storing step is an insignificant extra-solution activity. See MPEP 2106.05(g). using the new datapoint as input to each trained classifier: Mere data gathering recited at a high level of generality, and thus is an insignificant extra-solution activity (MPEP 2106.05(g)). wherein each trained classifier generates a prediction indicating a likelihood that the new datapoint belongs to the associated cluster or group… the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). and using a plurality of new datapoints and the new datapoints' assigned labels to train a machine learning model: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows: [a] system, comprising: one or more electronic processors configured to execute a set of computer-executable instructions; and one or more non-transitory electronic data storage media containing the set of computer- executable instructions, wherein when executed, the instructions cause the one or more electronic processor: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea and cannot provide inventive concept (MPEP 2106.05(f)). for each group or cluster, training a classifier of a plurality of classifiers to classify a datapoint as either inside or outside the group or cluster, wherein datapoints within the group or cluster are used as positive training examples and datapoints outside the group or cluster are used as negative training examples: This falls under Well-Understood, Routine, Conventional activity — see MPEP 2106.05(d). Training a classifier using in-cluster datapoints as positive training examples and out of cluster datapoints as negative training examples is a conventional supervised classification operation, as seen in Mohamed (US 20130336579 A1, 2012). Mohamed treats this as a known technique, describing in paragraph 3 that "[c]onventional appearance based methods of object recognition typically use both positive and negative training." Mohamed further describes in paragraph 19 that "[t]he feature vectors may be used for training according to a standard classifier training algorithm such as Support Vector Machine, or Viola Jones, etc to produce a model file 130." storing each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier: This limitation is merely a post-solution step of storing the data—a nominal addition to the claim that does not meaningfully limit the claim. The method storing is recited at a high level of generality. Simply implementing the abstract idea in a generic method is not a practical application of the abstract idea. Therefore, storing step is an insignificant extra-solution activity. See MPEP 2106.05(g). using the new datapoint as input to each trained classifier: The additional element of “using” does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of receiving steps amounts to no more than mere data gathering. This element amounts to receiving data over a network and is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II (i). This cannot provide an inventive concept. wherein each trained classifier generates a prediction indicating a likelihood that the new datapoint belongs to the associated cluster or group… the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea and cannot provide inventive concept (MPEP 2106.05(f)). wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea and cannot provide inventive concept (MPEP 2106.05(f)). and using a plurality of new datapoints and the new datapoints' assigned labels to train a machine learning model: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea and cannot provide inventive concept (MPEP 2106.05(f)). Claim 10 recites similar limitations to claim 2. Therefore claim 10 is rejected using the same rationale as claim 2. Claim 11 recites similar limitations to claim 4. Therefore claim 11 is rejected using the same rationale as claim 4. Claim 12 recites similar limitations to claim 5. Therefore claim 12 is rejected using the same rationale as claim 5. Claim 13 recites similar limitations to claims 6 and 7. Therefore claim 13 is rejected using the same rationale as claims 6 and 7. Claim 14 recites similar limitations to claim 8. Therefore claim 14 is rejected using the same rationale as claim 8. Claim 15 Step 1: The claim recites a non-transitory computer medium; therefore, it is directed to the statutory category of an article of manufacture. Step2A Prong 1: The claim recites, inter alia: generate one or more real-valued representations for each datapoint in a dataset: This limitation is seen as a mental process because it involves analyzing information and creating representations of a datapoint by expressing it numerically, which can be performed in the human mind. based on a similarity between the generated representations, form one or more groups or clusters of datapoints: This limitation recites a mental process because it involves grouping representations into clusters, which can be performed in the human mind or by pen and paper. represent each formed group or cluster by a unique identifier: This limitation recites a mental process because it assigns a label to each group of clusters, which is mentally performable. for each new datapoint… determine a most likely cluster or group to which the new datapoint is assigned based on the predictions…: This is a mental process because it involves the determination of which group the new datapoint belongs to. assign a label to the new datapoint based on the identifier of the cluster or group to which the new datapoint is assigned…: This is a mental process because it applies a label to an item based on a comparison, which can be performed in the human mind. Step2A Prong 2: This judicial exception is not integrated into a practical application because the additional elements are as follows: [o]ne or more non-transitory computer-readable media comprising a set of computer-executable instructions that when executed by one or more programmed electronic processors, cause the processors to: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). for each group or cluster, train a classifier of a plurality of classifiers to classify a datapoint as either inside or outside the group or cluster, wherein datapoints within the group or cluster are used as positive training examples and datapoints outside the group or cluster are used as negative training examples: Insignificant extra-solution as the limitation amounts to necessary data outputting (MPEP 2106.05(g)(3)). store each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier: This limitation is merely a post-solution step of storing the data—a nominal addition to the claim that does not meaningfully limit the claim. The method storing is recited at a high level of generality. Simply implementing the abstract idea in a generic method is not a practical application of the abstract idea. Therefore, storing step is an insignificant extra-solution activity. See MPEP 2106.05(g). using the new datapoint as input to each trained classifier: Mere data gathering recited at a high level of generality, and thus is an insignificant extra-solution activity (MPEP 2106.05(g)). wherein each trained classifier generates a prediction indicating a likelihood that the new datapoint belongs to the associated cluster or group… the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). and using a plurality of new datapoints and the new datapoints' assigned labels to train a machine learning model: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows: [o]ne or more non-transitory computer-readable media comprising a set of computer-executable instructions that when executed by one or more programmed electronic processors, cause the processors to: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea and cannot provide inventive concept (MPEP 2106.05(f)). for each group or cluster, training a classifier of a plurality of classifiers to classify a datapoint as either inside or outside the group or cluster, wherein datapoints within the group or cluster are used as positive training examples and datapoints outside the group or cluster are used as negative training examples: This falls under Well-Understood, Routine, Conventional activity — see MPEP 2106.05(d). Training a classifier using in-cluster datapoints as positive training examples and out of cluster datapoints as negative training examples is a conventional supervised classification operation, as seen in Mohamed (US 20130336579 A1, 2012). Mohamed treats this as a known technique, describing in paragraph 3 that "[c]onventional appearance based methods of object recognition typically use both positive and negative training." Mohamed further describes in paragraph 19 that "[t]he feature vectors may be used for training according to a standard classifier training algorithm such as Support Vector Machine, or Viola Jones, etc to produce a model file 130." store each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier: This limitation is merely a post-solution step of storing the data—a nominal addition to the claim that does not meaningfully limit the claim. The method storing is recited at a high level of generality. Simply implementing the abstract idea in a generic method is not a practical application of the abstract idea. Therefore, storing step is an insignificant extra-solution activity. See MPEP 2106.05(g). using the new datapoint as input to each trained classifier: The additional element of “using” does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of receiving steps amounts to no more than mere data gathering. This element amounts to receiving data over a network and is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II (i). This cannot provide an inventive concept. wherein each trained classifier generates a prediction indicating a likelihood that the new datapoint belongs to the associated cluster or group… the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea and cannot provide inventive concept (MPEP 2106.05(f)). wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea and cannot provide inventive concept (MPEP 2106.05(f)). and using a plurality of new datapoints and the new datapoints' assigned labels to train a machine learning model: Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea and cannot provide inventive concept (MPEP 2106.05(f)). Claim 16 recites similar limitations to claim 2. Therefore, claim 16 is rejected using the same rationale as claim 2. Claim 17 recites similar limitations to claim 4. Therefore, claim 17 is rejected using the same rationale as claim 4. Claim 18 recites similar limitations to claim 5. Therefore, claim 18 is rejected using the same rationale as claim 5. Claim 19 recites similar limitations to claims 6 and 7. Therefore, claim 19 is rejected using the same rationale as claims 6 and 7. Claim 20 recites similar limitations to claim 8. Therefore, claim 20 is rejected using the same rationale as claim 8. 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 (i.e., changing from AIA to pre-AIA ) 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. The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows: 1. Determining the scope and contents of the prior art. 2. Ascertaining the differences between the prior art and the claims at issue. 3. Resolving the level of ordinary skill in the pertinent art. 4. Considering objective evidence present in the application indicating obviousness or nonobviousness. This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention. Claims 1-3, and 5-6 are rejected under 35 U.S.C. 103 as being unpatentable over Xie (“Unsupervised Deep Embedding for Clustering Analysis”, 2016) in view of Rifkin (“In Defense of One-Vs-All Classification”, 2004), in view of Wang (CN-110020147-A), and further in view of Ratner (“Snorkel: Rapid Training Data Creation with Weak Supervision”, 2017). Regarding claim 1, Xie teaches based on a similarity between the generated representations, forming one or more groups or clusters of datapoints (Page 3 Section 3.1, “In the first step, we compute a soft assignment between the embedded points and the cluster centroids... we use the Student’s t-distribution as a kernel to measure the similarity between embedded point zi and centroid µj”, Page 2 Section 3, “The proposed algorithm (DEC) clusters data by simultaneously learning a set of k cluster centers {µj ∈ Z}k j=1 in the feature space Z…” Xie teaches the soft assignment of embedded points to centroids based on the similarity in the embedding space to group similar points into clusters.); representing each formed group or cluster by a unique identifier (Page 2 Section 3, “Consider the problem of clustering a set of n points {xi ∈ X}n i=1 into k clusters, each represented by a centroid µj,j = 1,...,k. Instead of clustering directly in the data space X, we propose to first transform the data with a non-linear mapping fθ : X → Z, where θ are learnable parameters and Z is the latent feature space… The proposed algorithm (DEC) clusters data by simultaneously learning a set of k cluster centers {µj ∈ Z}k j=1 in the feature space Z and the parameters θ of the DNN that maps data points into Z.”, Page 5 Section 4.2, “For all algorithms we set the number of clusters to the number of ground-truth categories… ci is the cluster assignment produced by the algorithm, and m ranges over all possible one-to-one mappings between clusters and labels.” Each cluster in Xie is represented by its own centroid and cluster index ci. That per cluster index corresponds to the claimed unique identifier of the group or cluster.); assigning a label to the new datapoint based on the identifier of the cluster or group to which the new datapoint is assigned… (Page 3 Section 3.1, “In the first step, we compute a soft assignment between the embedded points and the cluster centroids... embedding, α are the degrees of freedom of the Student’s t distribution and qij can be interpreted as the probability of assigning sample i to cluster j (i.e., a soft assignment).”, Page 5 Section 4.2, “For all algorithms we set the number of clusters to the number of ground-truth categories… ci is the cluster assignment produced by the algorithm, and m ranges over all possible one-to-one mappings between clusters and labels.” Xie’s Deep Embedding Clustering algorithm assigns each datapoint to a specific cluster index c_i, which serves as its label.); and using a plurality of new datapoints and the new datapoints' assigned labels to train a machine learning model (Page 2 Section 3, “Consider the problem of clustering a set of n points {xi ∈X}n i=1 into k clusters, each represented by a centroid µj,j = 1,...,k… The proposed algorithm (DEC) clusters data by simultaneously learning a set of k cluster centers {µj ∈ Z}k j=1 in the feature space Z and the parameters θ of the DNN that maps data points into Z.”, Page 3 Section 3.1.2, “Specifically, our model is trained by matching the soft assignment to the target distribution… we take an initial classifier and an unlabeled dataset, then label the dataset with the classifier in order to train on its own high confidence predictions.” Xie labels the dataset with its own cluster predictions and trains the model on those self-generated labels. Training on the datapoints together with their assigned cluster labels reads on the claimed training step.). Xie does not teach for each group or cluster, training a classifier to classify a datapoint as either inside or outside the group or cluster; storing each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier; for each new datapoint, using the new datapoint as input to each trained classifier and determining a most likely cluster or group to which the new datapoint is assigned. Rifkin, in the same field of endeavor, teaches for each group or cluster, training a classifier of a plurality of classifiers to classify a datapoint as either inside or outside the group or cluster, wherein datapoints within the group or cluster are used as positive training examples and datapoints outside the group or cluster are used as negative training examples (Page 2 Introduction, “One of the simplest multiclass classification schemes built on top of real-valued binary classifiers is to train N different binary classifiers, each one trained to distinguish the examples in a single class from the examples in all remaining classes. When it is desired to classify a new example, the N classifiers are run, and the classifier which outputs the largest (most positive) value is chosen.”, Page 7 Section 3.1.1, “given a single data point x belonging to class i, in the one-vs-all scheme we pay a penalty for machine i if fi(x) < 1, and for all other classes j we pay a penalty if fj(x) > −1.”, Page 5 Section 2, “if a point xi is in the positive class (yi = 1)…” Rifkin trains one binary classifier for each class. Each classifier is trained using the examples of its own single class as the positive set and the examples of all remaining classes as the negative set.) storing each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier (Page 2 Introduction, “…train N different binary classifiers, each one trained to distinguish the examples in a single class from the examples in all remaining classes.”, Page 11 Section 3.1.2, “On the test set, the one-vs-all system has an error rate…”, Page 9 Section 3.1, “There is an additional problem with the claim that the single-machine approach requires fewer support vectors, which is that in the OVA (or AVA) case, it is computationally easy to “reuse” support vectors that appear in multiple machines, leading to a large reduction in the total computational costs.” Each binary classifier is trained for a specific class and is labeled with that class. The classifier and its corresponding class are saved when the model is run to predict new data points. Rifkin trains N binary classifiers each corresponding to a specific class, evaluates them on a test set, and reuses the svm classifiers. If the classifiers are reused, then the learned parameters were stored in memory to predict new data points.); for each new datapoint, using the new datapoint as input to each trained classifier, wherein each trained classifier generates a prediction indicating a likelihood that the new datapoint belongs to the associated cluster or group and determining a most likely cluster or group to which the new datapoint is assigned based on the predictions from the plurality of trained classifiers (Page 2 Introduction, “One of the simplest multiclass classification schemes built on top of real-valued binary classifiers… When it is desired to classify a new example, the N classifiers are run, and the classifier which outputs the largest (most positive) value is chosen. This scheme will be referred to as the “one-vs-all…”, Page 13 Fig. 1 Caption, “if class i is more likely than class j at point x, then, asymptotically fi(x) > fj(x).” Rifkin runs each new example through all N real valued classifiers, and each classifier outputs a value that is larger when the example is more likely to belong to that classifier's class. Choosing the class of the classifier with the largest output reads on determining the most likely cluster based on the predictions from the plurality of trained classifiers.) Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie’s teaching of generating and clustering embedded representations of datapoints into groups identified by labels with Rifkin’s teaching of training each class using a separate classifier to see which class a new example belongs to in order to provide a more efficient mechanism for assigning new datapoints to the most likely cluster/class (Introduction of Rifkin). Xie in view of Rifkin does not explicitly teach a method of training a machine learning model, comprising: generating a real-valued representation for each datapoint in a dataset. Wang, in the same field of endeavor, teaches [a] method of training a machine learning model, comprising: generating one or more real-valued representations for each datapoint in a dataset (Page 5 Paragraph 3, “S31, using word2vec (a word representation as tool real-valued vector) of the first intermediate data to calculate the vector of each of the words;”) Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie and Rifkin’s teaching with Wang’s real-valued representation for each datapoint in order to enhance the model’s learning and prediction capability (Page 7 Paragraph 3 of Wang) Xie in view of Rifkin and Wang does not teach wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers. Ratner, in the same field of endeavor, teaches wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers (Page 2 Introduction of Ratner, "we developed data programming as a paradigm for addressing both of these challenges by modeling multiple label sources without access to ground truth, and generating probabilistic training labels representing the lineage of the individual labels.", Page 3 Fig. 2 and its Caption, "Snorkel applies the LFs over unlabeled data and learns a generative model to combine the LFs' outputs into probabilistic labels. (3) Snorkel uses these labels to train a discriminative classification model, such as a deep neural network.", Page 4 Section 2.1 of Ratner, "Weak classifiers: Classifiers that are insufficient for our task… can be used as labeling functions.", Section 2.2 of Ratner, "We then use the predictions… as probabilistic training labels.", See Fig. 2, PNG media_image1.png 431 1146 media_image1.png Greyscale ). Ratner applies multiple labeling functions to unlabeled data and combines their outputs with a generative model into probabilistic training labels that are then used to train a downstream model. Combining the outputs of those classifier labeling functions into the training labels corresponds to the assigned labels comprising programmatic labels generated from the predictions of the plurality of trained classifiers. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie in view of Rifkin and Wang's predictions from the plurality of per cluster classifiers with Ratner's data programming that combines predictions into programmatic training labels in order to rapidly create large training sets and train models without hand labeling any training data (Introduction of Ratner). Regarding claim 2, Xie teaches the real-valued representation for each datapoint in a dataset is generated by an embedding process (Page 3 Section 3.1.1, “where zi = fθ(xi) ∈ Z corresponds to xi ∈ X after embedding, α…”, Page 2 Introduction, “Consider the problem of clustering a set of n points {xi ∈ X}n i=1 into k clusters, each represented by a centroid µj,j = 1,...,k. Instead of clustering directly in the data space X, we propose to first transform the data with a non linear mapping fθ : X → Z, where θ are learnable parameters and Z is the latent feature space.” Xie generates a latent representation for each datapoint by embedding it into a feature space using a deep neural network.). Regarding claim 3, Xie teaches the embedding process is a text embedding process (Page 1 Introduction, “Our experiments show significant improvements over state of-the-art clustering methods in terms of both accuracy and running time on image and textual datasets.”, Page 4 Section 4.1, “We evaluate the proposed method (DEC) on one text dataset… We then computed tf-idf features on the 2000 most frequently occurring word stems.” Xie explicitly teaches that the embedding process is performed on both images and texts and specifically uses tf-idf to embed the text datapoints.). Regarding claim 5, Xie does not teach determining a most likely cluster or group to which the new datapoint should be assigned further comprises determining the cluster or group associated with the trained classifier having the highest level of certainty in its output. Rifkin, in the same field of endeavor, teaches determining a most likely cluster or group to which the new datapoint should be assigned further comprises determining the cluster or group associated with the trained classifier having the highest level of certainty in its output (Page 2 Introduction, “to train N different binary classifiers, each one trained to distinguish the examples in a single class from the examples in all remaining classes. When it is desired to classify a new example, the N classifiers are run, and the classifier which outputs the largest (most positive) value is chosen.” Rifkin teaches the determination of which class a datapoint belongs to by inputting the new datapoint example to each classifier to determine which classifier has the highest output (certainty) that the data point belongs to.). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie’s clustering framework with Rifkin’s teaching of selecting the classifier with the largest (most positive) output in order to provide a decision rule for determining the most likely cluster a datapoint belongs to to improve the accuracy of the machine learning model (Section 3.1.5 of Rifkin). Regarding claim 6, Xie teaches the similarity between the generated representations is determined based on a metric (Page 3 Section 3.1.1, “we use the Student’s t-distribution as a kernel to measure the similarity between embedded point zi and centroid µj”). Claims 4, 7, 9-13, and 15-19 are rejected under 35 U.S.C. 103 as being unpatentable Xie (“Unsupervised Deep Embedding for Clustering Analysis”, 2016) in view of Rifkin (“In Defense of One-Vs-All Classification”, 2004), in view of Wang (CN-110020147-A), further in view of Ratner (“Snorkel: Rapid Training Data Creation with Weak Supervision”, 2017), and further in view of Haghighat (US 20240370740 A1). Regarding claim 4, Xie does not teach the unique identifier is based on one or more attributes of a datapoint or datapoints in the cluster or group. Haghighat, in the same field of endeavor, teaches the unique identifier is based on one or more attributes of a datapoint or datapoints in the cluster or group (Paragraph 32, “the CNN 22 can evaluate the proximity of subsequent new vectors… using the Euclidean distance between the evaluated new vector 40 and the unregistered new vectors… If the evaluated new vector 40 is closest to a previously-added new vector 40, with a corresponding confidence level, then the two new vectors 40 can be considered as corresponding with a new class and may be used to establish a new cluster 42 associated with that new class… The new class can then be registered, including by associating a generic or preliminary identifier with the class and establishing a covariance matrix of the new vectors 40 in the new cluster 42.” Haghighat teaches that feature vectors representing datapoint attributes are evaluated using the Euclidean distance to determine proximity to establish a new cluster which creates a new class and associated a preliminary identifier with it. The identifier is created based on the attributes of the datapoints forming the cluster.). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie’s clustering framework with Haghighat’s teaching of creating a new class that’s associating an identifier with a cluster based on the feature vectors of datapoints in order to improve the organization of clustered data (Paragraph 31 of Haghighat). Regarding claim 7, Xie does not teach the metric is one of Manhattan distance, Euclidean distance, or Cosine distance. Haghighat, in the same field of endeavor, teaches the metric is one of Manhattan distance, Euclidean distance, or Cosine distance (Paragraph 32, “the CNN 22 can evaluate the proximity of subsequent new vectors… using the Euclidean distance between the evaluated new vector 40 and the unregistered new vectors… If the evaluated new vector 40 is closest to a previously-added new vector 40, with a corresponding confidence level, then the two new vectors 40 can be considered as corresponding with a new class and may be used to establish a new cluster 42 associated with that new class… The new class can then be registered, including by associating a generic or preliminary identifier with the class and establishing a covariance matrix of the new vectors 40 in the new cluster 42.” Haghighat teaches that feature vectors representing datapoint attributes are evaluated using the Euclidean distance to determine proximity to establish a new cluster which creates a new class and associated a preliminary identifier with it. The identifier is created based on the attributes of the datapoints forming the cluster.). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie’s clustering framework with Haghighat’s teaching of creating a new class that’s associating an identifier with a cluster based on the Euclidean distance of feature vectors in order to improve the accuracy and organization of clustered data (Paragraphs 31 and 32 of Haghighat). Regarding claim 9, Xie teaches based on a similarity between the generated representations, form one or more groups or clusters of datapoints (Page 3 Section 3.1, “In the first step, we compute a soft assignment between the embedded points and the cluster centroids... we use the Student’s t-distribution as a kernel to measure the similarity between embedded point zi and centroid µj”, Page 2 Section 3, “The proposed algorithm (DEC) clusters data by simultaneously learning a set of k cluster centers {µj ∈ Z}k j=1 in the feature space Z…” Xie teaches the soft assignment of embedded points to centroids based on the similarity in the embedding space to group similar points into clusters.); representing each formed group or cluster by a unique identifier (Page 2 Section 3, “Consider the problem of clustering a set of n points {xi ∈ X}n i=1 into k clusters, each represented by a centroid µj,j = 1,...,k. Instead of clustering directly in the data space X, we propose to first transform the data with a non-linear mapping fθ : X → Z, where θ are learnable parameters and Z is the latent feature space… The proposed algorithm (DEC) clusters data by simultaneously learning a set of k cluster centers {µj ∈ Z}k j=1 in the feature space Z and the parameters θ of the DNN that maps data points into Z.”, Page 5 Section 4.2, “For all algorithms we set the number of clusters to the number of ground-truth categories… ci is the cluster assignment produced by the algorithm, and m ranges over all possible one-to-one mappings between clusters and labels.” Each cluster in Xie is represented by its own centroid and cluster index ci. That per cluster index corresponds to the claimed unique identifier of the group or cluster.); assign a label to the new datapoint based on the identifier of the cluster or group to which the new datapoint is assigned… (Page 3 Section 3.1, “In the first step, we compute a soft assignment between the embedded points and the cluster centroids... embedding, α are the degrees of freedom of the Student’s t distribution and qij can be interpreted as the probability of assigning sample i to cluster j (i.e., a soft assignment).”, Page 5 Section 4.2, “For all algorithms we set the number of clusters to the number of ground-truth categories… ci is the cluster assignment produced by the algorithm, and m ranges over all possible one-to-one mappings between clusters and labels.” Xie’s Deep Embedding Clustering algorithm assigns each datapoint to a specific cluster index c_i, which serves as its label.); and use a plurality of new datapoints and the new datapoints' assigned labels to train a machine learning model (Page 2 Section 3, “Consider the problem of clustering a set of n points {xi ∈X}n i=1 into k clusters, each represented by a centroid µj,j = 1,...,k… The proposed algorithm (DEC) clusters data by simultaneously learning a set of k cluster centers {µj ∈ Z}k j=1 in the feature space Z and the parameters θ of the DNN that maps data points into Z.”, Page 3 Section 3.1.2, “Specifically, our model is trained by matching the soft assignment to the target distribution… we take an initial classifier and an unlabeled dataset, then label the dataset with the classifier in order to train on its own high confidence predictions.” Xie labels the dataset with its own cluster predictions and trains the model on those self-generated labels. Training on the datapoints together with their assigned cluster labels reads on the claimed training step.). Xie does not teach for each group or cluster, training a classifier to classify a datapoint as either inside or outside the group or cluster; storing each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier; for each new datapoint, using the new datapoint as input to each trained classifier and determining a most likely cluster or group to which the new datapoint is assigned. Rifkin, in the same field of endeavor, teaches for each group or cluster, train a classifier of a plurality of classifiers to classify a datapoint as either inside or outside the group or cluster, wherein datapoints within the group or cluster are used as positive training examples and datapoints outside the group or cluster are used as negative training examples (Page 2 Introduction, “One of the simplest multiclass classification schemes built on top of real-valued binary classifiers is to train N different binary classifiers, each one trained to distinguish the examples in a single class from the examples in all remaining classes. When it is desired to classify a new example, the N classifiers are run, and the classifier which outputs the largest (most positive) value is chosen.”, Page 7 Section 3.1.1, “given a single data point x belonging to class i, in the one-vs-all scheme we pay a penalty for machine i if fi(x) < 1, and for all other classes j we pay a penalty if fj(x) > −1.”, Page 5 Section 2, “if a point xi is in the positive class (yi = 1)…” Rifkin trains one binary classifier for each class. Each classifier is trained using the examples of its own single class as the positive set and the examples of all remaining classes as the negative set.) store each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier (Page 2 Introduction, “…train N different binary classifiers, each one trained to distinguish the examples in a single class from the examples in all remaining classes.”, Page 11 Section 3.1.2, “On the test set, the one-vs-all system has an error rate…”, Page 9 Section 3.1, “There is an additional problem with the claim that the single-machine approach requires fewer support vectors, which is that in the OVA (or AVA) case, it is computationally easy to “reuse” support vectors that appear in multiple machines, leading to a large reduction in the total computational costs.” Each binary classifier is trained for a specific class and is labeled with that class. The classifier and its corresponding class are saved when the model is run to predict new data points. Rifkin trains N binary classifiers each corresponding to a specific class, evaluates them on a test set, and reuses the svm classifiers. If the classifiers are reused, then the learned parameters were stored in memory to predict new data points.); for each new datapoint, use the new datapoint as input to each trained classifier, wherein each trained classifier generates a prediction indicating a likelihood that the new datapoint belongs to the associated cluster or group and determining a most likely cluster or group to which the new datapoint is assigned based on the predictions from the plurality of trained classifiers (Page 2 Introduction, “One of the simplest multiclass classification schemes built on top of real-valued binary classifiers… When it is desired to classify a new example, the N classifiers are run, and the classifier which outputs the largest (most positive) value is chosen. This scheme will be referred to as the “one-vs-all…”, Page 13 Fig. 1 Caption, “if class i is more likely than class j at point x, then, asymptotically fi(x) > fj(x).” Rifkin runs each new example through all N real valued classifiers, and each classifier outputs a value that is larger when the example is more likely to belong to that classifier's class. Choosing the class of the classifier with the largest output reads on determining the most likely cluster based on the predictions from the plurality of trained classifiers.) Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie’s teaching of generating and clustering embedded representations of datapoints into groups identified by labels with Rifkin’s teaching of training each class using a separate classifier to see which class a new example belongs to in order to provide a more efficient mechanism for assigning new datapoints to the most likely cluster/class (Introduction of Rifkin). Xie in view of Rifkin does not explicitly teach a method of training a machine learning model, comprising: generating a real-valued representation for each datapoint in a dataset. Wang, in the same field of endeavor, teaches generate one or more real-valued representations for each datapoint in a dataset (Page 5 Paragraph 3, “S31, using word2vec (a word representation as tool real-valued vector) of the first intermediate data to calculate the vector of each of the words;”) Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie and Rifkin’s teaching with Wang’s real-valued representation for each datapoint in order to enhance the model’s learning and prediction capability (Page 7 Paragraph 3 of Wang) Xie in view of Rifkin and Wang does not teach wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers. Ratner, in the same field of endeavor, teaches wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers (Page 2 Introduction of Ratner, "we developed data programming as a paradigm for addressing both of these challenges by modeling multiple label sources without access to ground truth, and generating probabilistic training labels representing the lineage of the individual labels.", Page 3 Fig. 2 and its Caption, "Snorkel applies the LFs over unlabeled data and learns a generative model to combine the LFs' outputs into probabilistic labels. (3) Snorkel uses these labels to train a discriminative classification model, such as a deep neural network.", Page 4 Section 2.1 of Ratner, "Weak classifiers: Classifiers that are insufficient for our task… can be used as labeling functions.", Section 2.2 of Ratner, "We then use the predictions… as probabilistic training labels.", See Fig. 2, PNG media_image1.png 431 1146 media_image1.png Greyscale ). Ratner applies multiple labeling functions to unlabeled data and combines their outputs with a generative model into probabilistic training labels that are then used to train a downstream model. Combining the outputs of those classifier labeling functions into the training labels corresponds to the assigned labels comprising programmatic labels generated from the predictions of the plurality of trained classifiers. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie in view of Rifkin and Wang's predictions from the plurality of per cluster classifiers with Ratner's data programming that combines predictions into programmatic training labels in order to rapidly create large training sets and train models without hand labeling any training data (Introduction of Ratner). Xie in view of Rifkin in further view of Wang does not appear to teach a system, comprising: one or more electronic processors configured to execute a set of computer-executable instructions; and one or more non-transitory electronic data storage media containing the set of computer- executable instructions, wherein when executed, the instructions cause the one or more electronic processors to… Haghighat, in the same field of endeavor, teaches [a] system, comprising: one or more electronic processors configured to execute a set of computer-executable instructions; and one or more non-transitory electronic data storage media containing the set of computer- executable instructions, wherein when executed, the instructions cause the one or more electronic processors to (Paragraph 24 of Hagighat, “In one implementation, the controller 18 can be a microprocessor that executes a program, stored in the memory 20, for operation of the oven 10. Alternatively, the controller 18 can be an application-specific integrated circuit (“ASIC”). The memory 20 can be packaged with the controller 18 (i.e. a “system-on-chip” configuration) or can be connected with the controller 18 by associated circuitry and/or wiring.”) Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie in view of Rifkin in further view of Wang’s clustering framework with Haghighat’s teaching of a system comprising a processor and non-transitory memory to execute instructions in order to implement Xie’s method in a computing environment that executes the clustering and classification steps (Paragraph 24 of Hasghighat). Claim 10 recites similar limitations to claim 2. Therefore claim 10 is rejected using the same rationale as claim 2. Claim 11 recites similar limitations to claim 4. Therefore claim 11 is rejected using the same rationale as claim 4. Claim 12 recites similar limitations to claim 5. Therefore claim 12 is rejected using the same rationale as claim 5. Claim 13 recites similar limitations to claims 6 and 7. Therefore claim 13 is rejected using the same rationale as claims 6 and 7. Regarding claim 15, Xie teaches based on a similarity between the generated representations, form one or more groups or clusters of datapoints (Page 3 Section 3.1, “In the first step, we compute a soft assignment between the embedded points and the cluster centroids... we use the Student’s t-distribution as a kernel to measure the similarity between embedded point zi and centroid µj”, Page 2 Section 3, “The proposed algorithm (DEC) clusters data by simultaneously learning a set of k cluster centers {µj ∈ Z}k j=1 in the feature space Z…” Xie teaches the soft assignment of embedded points to centroids based on the similarity in the embedding space to group similar points into clusters.); representing each formed group or cluster by a unique identifier (Page 2 Section 3, “Consider the problem of clustering a set of n points {xi ∈ X}n i=1 into k clusters, each represented by a centroid µj,j = 1,...,k. Instead of clustering directly in the data space X, we propose to first transform the data with a non-linear mapping fθ : X → Z, where θ are learnable parameters and Z is the latent feature space… The proposed algorithm (DEC) clusters data by simultaneously learning a set of k cluster centers {µj ∈ Z}k j=1 in the feature space Z and the parameters θ of the DNN that maps data points into Z.”, Page 5 Section 4.2, “For all algorithms we set the number of clusters to the number of ground-truth categories… ci is the cluster assignment produced by the algorithm, and m ranges over all possible one-to-one mappings between clusters and labels.” Each cluster in Xie is represented by its own centroid and cluster index ci. That per cluster index corresponds to the claimed unique identifier of the group or cluster.); assign a label to the new datapoint based on the identifier of the cluster or group to which the new datapoint is assigned… (Page 3 Section 3.1, “In the first step, we compute a soft assignment between the embedded points and the cluster centroids... embedding, α are the degrees of freedom of the Student’s t distribution and qij can be interpreted as the probability of assigning sample i to cluster j (i.e., a soft assignment).”, Page 5 Section 4.2, “For all algorithms we set the number of clusters to the number of ground-truth categories… ci is the cluster assignment produced by the algorithm, and m ranges over all possible one-to-one mappings between clusters and labels.” Xie’s Deep Embedding Clustering algorithm assigns each datapoint to a specific cluster index c_i, which serves as its label.); and use a plurality of new datapoints and the new datapoints' assigned labels to train a machine learning model (Page 2 Section 3, “Consider the problem of clustering a set of n points {xi ∈X}n i=1 into k clusters, each represented by a centroid µj,j = 1,...,k… The proposed algorithm (DEC) clusters data by simultaneously learning a set of k cluster centers {µj ∈ Z}k j=1 in the feature space Z and the parameters θ of the DNN that maps data points into Z.”, Page 3 Section 3.1.2, “Specifically, our model is trained by matching the soft assignment to the target distribution… we take an initial classifier and an unlabeled dataset, then label the dataset with the classifier in order to train on its own high confidence predictions.” Xie labels the dataset with its own cluster predictions and trains the model on those self-generated labels. Training on the datapoints together with their assigned cluster labels reads on the claimed training step.). Xie does not teach for each group or cluster, training a classifier to classify a datapoint as either inside or outside the group or cluster; storing each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier; for each new datapoint, using the new datapoint as input to each trained classifier and determining a most likely cluster or group to which the new datapoint is assigned. Rifkin, in the same field of endeavor, teaches for each group or cluster, train a classifier of a plurality of classifiers to classify a datapoint as either inside or outside the group or cluster, wherein datapoints within the group or cluster are used as positive training examples and datapoints outside the group or cluster are used as negative training examples (Page 2 Introduction, “One of the simplest multiclass classification schemes built on top of real-valued binary classifiers is to train N different binary classifiers, each one trained to distinguish the examples in a single class from the examples in all remaining classes. When it is desired to classify a new example, the N classifiers are run, and the classifier which outputs the largest (most positive) value is chosen.”, Page 7 Section 3.1.1, “given a single data point x belonging to class i, in the one-vs-all scheme we pay a penalty for machine i if fi(x) < 1, and for all other classes j we pay a penalty if fj(x) > −1.”, Page 5 Section 2, “if a point xi is in the positive class (yi = 1)…” Rifkin trains one binary classifier for each class. Each classifier is trained using the examples of its own single class as the positive set and the examples of all remaining classes as the negative set.) store each trained classifier and associating the stored trained classifier with the cluster or group's unique identifier (Page 2 Introduction, “…train N different binary classifiers, each one trained to distinguish the examples in a single class from the examples in all remaining classes.”, Page 11 Section 3.1.2, “On the test set, the one-vs-all system has an error rate…”, Page 9 Section 3.1, “There is an additional problem with the claim that the single-machine approach requires fewer support vectors, which is that in the OVA (or AVA) case, it is computationally easy to “reuse” support vectors that appear in multiple machines, leading to a large reduction in the total computational costs.” Each binary classifier is trained for a specific class and is labeled with that class. The classifier and its corresponding class are saved when the model is run to predict new data points. Rifkin trains N binary classifiers each corresponding to a specific class, evaluates them on a test set, and reuses the svm classifiers. If the classifiers are reused, then the learned parameters were stored in memory to predict new data points.); for each new datapoint, use the new datapoint as input to each trained classifier, wherein each trained classifier generates a prediction indicating a likelihood that the new datapoint belongs to the associated cluster or group and determining a most likely cluster or group to which the new datapoint is assigned based on the predictions from the plurality of trained classifiers (Page 2 Introduction, “One of the simplest multiclass classification schemes built on top of real-valued binary classifiers… When it is desired to classify a new example, the N classifiers are run, and the classifier which outputs the largest (most positive) value is chosen. This scheme will be referred to as the “one-vs-all…”, Page 13 Fig. 1 Caption, “if class i is more likely than class j at point x, then, asymptotically fi(x) > fj(x).” Rifkin runs each new example through all N real valued classifiers, and each classifier outputs a value that is larger when the example is more likely to belong to that classifier's class. Choosing the class of the classifier with the largest output reads on determining the most likely cluster based on the predictions from the plurality of trained classifiers.) Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie’s teaching of generating and clustering embedded representations of datapoints into groups identified by labels with Rifkin’s teaching of training each class using a separate classifier to see which class a new example belongs to in order to provide a more efficient mechanism for assigning new datapoints to the most likely cluster/class (Introduction of Rifkin). Xie in view of Rifkin does not explicitly teach a method of training a machine learning model, comprising: generating a real-valued representation for each datapoint in a dataset. Wang, in the same field of endeavor, teaches generate one or more real-valued representations for each datapoint in a dataset (Page 5 Paragraph 3, “S31, using word2vec (a word representation as tool real-valued vector) of the first intermediate data to calculate the vector of each of the words;”) Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie and Rifkin’s teaching with Wang’s real-valued representation for each datapoint in order to enhance the model’s learning and prediction capability (Page 7 Paragraph 3 of Wang) Xie in view of Rifkin and Wang does not teach wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers. Ratner, in the same field of endeavor, teaches wherein the assigned labels comprise programmatic labels generated from the predictions of the plurality of trained classifiers (Page 2 Introduction of Ratner, "we developed data programming as a paradigm for addressing both of these challenges by modeling multiple label sources without access to ground truth, and generating probabilistic training labels representing the lineage of the individual labels.", Page 3 Fig. 2 and its Caption, "Snorkel applies the LFs over unlabeled data and learns a generative model to combine the LFs' outputs into probabilistic labels. (3) Snorkel uses these labels to train a discriminative classification model, such as a deep neural network.", Page 4 Section 2.1 of Ratner, "Weak classifiers: Classifiers that are insufficient for our task… can be used as labeling functions.", Section 2.2 of Ratner, "We then use the predictions… as probabilistic training labels.", See Fig. 2, PNG media_image1.png 431 1146 media_image1.png Greyscale ). Ratner applies multiple labeling functions to unlabeled data and combines their outputs with a generative model into probabilistic training labels that are then used to train a downstream model. Combining the outputs of those classifier labeling functions into the training labels corresponds to the assigned labels comprising programmatic labels generated from the predictions of the plurality of trained classifiers. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie in view of Rifkin and Wang's predictions from the plurality of per cluster classifiers with Ratner's data programming that combines predictions into programmatic training labels in order to rapidly create large training sets and train models without hand labeling any training data (Introduction of Ratner). Xie in view of Rifkin in further view of Wang does not appear to teach one or more non-transitory computer-readable media comprising a set of computer-executable instructions that when executed by one or more programmed electronic processors. Haghighat, in the same field of endeavor, teaches [o]ne or more non-transitory computer-readable media comprising a set of computer-executable instructions that when executed by one or more programmed electronic processors, cause the processors to… (Paragraph 24 of Hagighat, “In one implementation, the controller 18 can be a microprocessor that executes a program, stored in the memory 20, for operation of the oven 10. Alternatively, the controller 18 can be an application-specific integrated circuit (“ASIC”). The memory 20 can be packaged with the controller 18 (i.e. a “system-on-chip” configuration) or can be connected with the controller 18 by associated circuitry and/or wiring.”) Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie’s clustering framework with Haghighat’s teaching of a system comprising a processor and non-transitory memory to execute instructions in order to implement Xie’s method in a computing environment that executes the clustering and classification steps (Paragraph 24 of Hasghighat). Claim 16 recites similar limitations to claim 2. Therefore, claim 16 is rejected using the same rationale as claim 2. Claim 17 recites similar limitations to claim 4. Therefore, claim 17 is rejected using the same rationale as claim 4. Claim 18 recites similar limitations to claim 5. Therefore, claim 18 is rejected using the same rationale as claim 5. Claim 19 recites similar limitations to claims 6 and 7. Therefore, claim 19 is rejected using the same rationale as claims 6 and 7. Claims 8, 14, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Xie (“Unsupervised Deep Embedding for Clustering Analysis”, 2016) in view of Rifkin (“In Defense of One-Vs-All Classification”, 2004), in view of Wang (CN-110020147-A), further in view of Ratner (“Snorkel: Rapid Training Data Creation with Weak Supervision”, 2017), and further in view of Romano (US 11741133 B1). Regarding claim 8, Xie in view of Rifkin, Wang, and Ratner does not teach the one or more real-valued representations for each datapoint in the dataset comprise a plurality of real-valued representations for each datapoint in the dataset. Romano, in the same field of endeavor, teaches the one or more real-valued representations for each datapoint in the dataset comprise a plurality of real-valued representations for each datapoint in the dataset (Col. 4 Lines 64-67 of Romano, “…breaking down the first asset file into a first plurality of asset rows; for each asset row of the first plurality of asset rows: extracting one or more features from the asset row, encoding each of the one or more features, and embedding each encoded feature with a generational time stamp;”, Claim 1, “extracting one or more features from the asset row, encoding each of the one or more features, and embedding each encoded feature with a generational time stamp” Romano extracts multiple features from each data row and embeds each feature as its own encoded vector, producing several real valued representations per data point. Generating a plurality of embedded feature representations for each asset row reads on the claimed plurality of real-valued representations for each datapoint, where an asset row corresponds to the claimed datapoint.). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to combine Xie in view of Rifkin, Wang, and Ratner with Romano's generation of multiple feature representations per datapoint in order to enhance the datapoint representations and improve classification performance (Paragraph 3 of Romano). Claim 14 recites similar limitations to claim 8. Therefore claim 14 is rejected using the same rationale as claim 8. Claim 20 recites similar limitations to claim 8. Therefore, claim 20 is rejected using the same rationale as claim 8. 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 MAJD MAHER HADDAD whose telephone number is (571)272-2265. The examiner can normally be reached Mon-Friday 8-5 pm. 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, Kamran Afshar, can be reached at (571) 272-7796. 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. /M.M.H./Examiner, Art Unit 2125 /KAMRAN AFSHAR/Supervisory Patent Examiner, Art Unit 2125
Read full office action

Prosecution Timeline

Jun 26, 2023
Application Filed
Mar 06, 2026
Non-Final Rejection mailed — §101, §103, §112
Sep 04, 2026
Response Filed
Oct 02, 2026
Final Rejection mailed — §101, §103, §112 (current)

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12737630
FIRST NETWORK NODE AND METHOD PERFORMED THEREIN FOR HANDLING DATA IN A COMMUNICATION NETWORK
3y 11m to grant Granted Sep 15, 2026
Patent 12705535
Systems and Methods for Grouping Records Associated with Like Media Items
3y 6m to grant Granted Aug 11, 2026
Study what changed to get past this examiner. Based on 2 most recent grants.

Strategy Recommendation AI-generated — please review before filing

Get a prosecution strategy drawn from examiner precedents, rejection analysis, and claim mapping.
Typically takes 5-10 seconds — AI-generated, attorney review required before filing

Prosecution Projections

3-4
Expected OA Rounds
78%
Grant Probability
99%
With Interview (+88.9%)
3y 3m (~0m remaining)
Median Time to Grant
Moderate
PTA Risk
Based on 9 resolved cases by this examiner. Grant probability derived from career allowance rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month