DETAILED ACTION
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 .
Claims 1-14 are presented for examination.
Response to Amendment
Applicant’s amendment has obviated the drawing objections. Therefore, those objections are withdrawn.
Priority
Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
Specification
The lengthy specification has not been checked to the extent necessary to determine the presence of all possible minor errors. Applicant’s cooperation is requested in correcting any errors of which applicant may become aware in the specification.
Claim Rejections - 35 USC § 101
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim 14 is rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claim(s) does/do not fall within at least one of the four categories of patent eligible subject matter because, under its broadest reasonable interpretation in light of the specification, it is directed to data per se. The claim is directed to a “data collection comprising [a] data structure … and [a] plurality of datasets” and recites no hardware on which the data are stored.
Claims 1-14 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The analysis of the claims will follow the 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (“2019 PEG”).
Claim 1
Step 1: The claim recites a method; therefore, it is directed to the statutory category of processes.
Step 2A Prong 1: The claim recites, inter alia:
[E]nabling an assessment of a plurality of datasets forming training data for a machine-learning algorithm, each dataset of the plurality of datasets including a respective input datapoint in an input space and an associated output datapoint in an output space, … each output datapoint indicating objects on a railroad track depicted by … respective two-dimensional image data: This limitation could encompass mentally assessing the datasets including the input and output datapoints.
[F]or each dataset of the plurality of datasets: determining a respective sequence of a predefined length, the respective sequence including further datasets progressively selected from the plurality of datasets based on a distance of the input datapoints thereof to the input datapoint of the respective dataset: This limitation could encompass mentally determining a sequence of datasets based on the distance of their inputs to an input of another dataset.
[F]or each dataset of the plurality of datasets: determining whether the input datapoint of the respective dataset and the input datapoints of each of the further datasets included in the respective sequence respectively fulfill a first neighborhood criterion that is defined in the input space: This limitation could encompass mentally determining whether the inputs of the datasets satisfy a neighborhood criterion.
[F]or each dataset of the plurality of datasets: determining whether the output datapoint of the respective dataset and the output datapoints of each of the further datasets included in the respective sequence respectively fulfill a second neighborhood criterion that is defined in the output space: This limitation could encompass mentally determining whether the outputs of the datasets satisfy a neighborhood criterion.
[F]or each dataset of the plurality of datasets and for each sequence entry of the respective sequence: determining a respective cumulative fulfillment ratio based on how many of the further datasets included in the sequence up to the respective entry fulfill both the first neighborhood criterion and the second neighborhood criterion: This limitation could encompass mentally determining what proportion of the datasets satisfy both neighborhood criteria.
[D]etermining a data structure, an array dimension of the data structure resolving the sequences determined for each one of the plurality of datasets, a further array dimension of the data structure resolving the cumulative fulfillment ratio, each entry of the data structure including a count of datapoints that are associated with the respective cumulative fulfillment ratio at the respective sequence entry defined by the position along the array dimension and the further array dimension: This limitation could encompass mentally arranging the fulfillment ratios by data structure in an array, or doing so with a pen and paper.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that the method is “computer-implemented”. However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f). The claim further recites that “each input datapoint correspond[s] to two-dimensional image data acquired using a camera”. This limitation is directed to the insignificant extra-solution activity of mere data gathering and output. MPEP § 2106.05(g).
Step 2B: The claim does not contain significantly more than the judicial exception. The acquisition of the data using a camera, in addition to being insignificant extra-solution activity, also recites the well-understood, routine, and conventional activity of storing and retrieving information in memory, MPEP § 2106.05(d)(II); Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015). Otherwise, the analysis at this step mirrors that of step 2A, prong 1. As an ordered whole, the claim is directed to a mentally performable process of assessing datasets. Nothing in the claim provides significantly more than this. As such, the claim is not patent eligible.
Claim 2
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “each entry of the data structure further comprises an identification of the datasets that are associated with the respective cumulative fulfillment ratio at the respective sequence entry defined by the position along the array dimension and the further array dimension.” Determining the data structure remains mentally performable under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. See claim 1 analysis.
Step 2B: The claim does not contain significantly more than the judicial exception. See claim 1 analysis.
Claim 3
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “an increment of the array dimension corresponds to a predetermined distance offset in the input space between adjacent input datapoints of the respective datasets in the sequences.” Determining the data structure remains mentally performable under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. See claim 1 analysis.
Step 2B: The claim does not contain significantly more than the judicial exception. See claim 1 analysis.
Claim 4
Step 1: The claim recites a method; therefore, it is directed to the statutory category of processes.
Step 2A Prong 1: The claim recites, inter alia:
[A]ssessing training data for training an algorithm, the training data having a plurality of datasets, each dataset of the plurality of datasets including a respective input datapoint in an input space and an associated output datapoint in an output space, the output datapoints of the plurality of datasets being ground-truth labels indicative of multiple classes to be predicted by the algorithm: This limitation could encompass mentally assessing the training data.
[D]etermining a data structure by the computer-implemented method according to claim 1: As shown in the analysis of claim 1, determining the data structure is mentally performable.
[O]n access to the data structure, assessing the training data: This limitation could encompass mentally assessing the training data.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites “accessing the data structure thus determined”. This limitation recites the insignificant extra-solution activity of mere data gathering and output. MPEP § 2106.05(g).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites “accessing the data structure thus determined”. This limitation recites the well-understood, routine, and conventional activity of storing and retrieving information in memory. MPEP § 2106.05(d)(II); Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015).
Claim 5
Step 1: A process, as above.
Step 2A Prong 1: The claim recites, inter alia, “determining a plot of the data structure, with a contrast of plot values of the plot being associated with the count of the datapoints, a first axis of the plot resolving the array dimension, and a second axis of the plot resolving the further array dimension”. This limitation could encompass mentally determining the plot with these parameters and/or drawing the plot using a pen and paper.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites “accessing the data structure …; and outputting the plot via a user interface.” These limitations recite the insignificant extra-solution activity of mere data gathering and output. MPEP § 2106.05(g).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites “accessing the data structure …; and outputting the plot via a user interface.” These limitations recite the well-understood, routine, and conventional activity of storing and retrieving information in memory. MPEP § 2106.05(d)(II); Versata, 793 F.3d at 1334, 115 USPQ2d at 1701.
Claim 6
Step 1: A process, as above.
Step 2A Prong 1: The claim recites, inter alia, “ identifying a subset of the plurality of datasets by selecting parts of the plot”. This limitation could encompass mentally identifying the subset.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites “presenting datasets in the subset to the user via the user interface.” This limitation recites the insignificant extra-solution activity of mere data gathering and output. MPEP § 2106.05(g).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites “presenting datasets in the subset to the user via the user interface.” This limitation recites the well-understood, routine, and conventional activity of storing and retrieving information in memory. MPEP § 2106.05(d)(II); Versata, 793 F.3d at 1334, 115 USPQ2d at 1701.
Claim 7
Step 1: A process, as above.
Step 2A Prong 1: The claim recites the same judicial exceptions as in claim 6.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites “highlighting the input datapoints or the output datapoints in a reduced-dimensionality plot of the input space or of the output space, respectively.” This limitation recites the insignificant extra-solution activity of mere data gathering and output. MPEP § 2106.05(g).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites “highlighting the input datapoints or the output datapoints in a reduced-dimensionality plot of the input space or of the output space, respectively.” This limitation recites the well-understood, routine, and conventional activity of storing and retrieving information in memory. MPEP § 2106.05(d)(II); Versata, 793 F.3d at 1334, 115 USPQ2d at 1701.
Claim 8
Step 1: A process, as above.
Step 2A Prong 1: The claim recites the same judicial exceptions as in claim 5.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites “obtaining a selection of a given dataset of the plurality of datasets; and highlighting in the plot an evolution of the respective cumulative fulfillment ratio of the given dataset for various positions along the first axis.” These limitations recite the insignificant extra-solution activity of mere data gathering and output. MPEP § 2106.05(g).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites “obtaining a selection of a given dataset of the plurality of datasets; and highlighting in the plot an evolution of the respective cumulative fulfillment ratio of the given dataset for various positions along the first axis.” These limitations recite the well-understood, routine, and conventional activity of storing and retrieving information in memory. MPEP § 2106.05(d)(II); Versata, 793 F.3d at 1334, 115 USPQ2d at 1701.
Claim 9
Step 1: A process, as above.
Step 2A Prong 1: The claim recites, inter alia, “solving inference tasks”. This limitation could encompass mentally solving the inference tasks.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites “training the algorithm based on the training data; and upon training the algorithm, using the algorithm”. However, these amount to mere instructions to apply the judicial exception using a generic computer programmed with a generic class of computer algorithm. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites “training the algorithm based on the training data; and upon training the algorithm, using the algorithm”. However, these amount to mere instructions to apply the judicial exception using a generic computer programmed with a generic class of computer algorithm. MPEP § 2106.05(f).
Claim 10
Step 1: The claim recites a method; therefore, it is directed to the statutory category of processes.
Step 2A Prong 1: The claim recites, inter alia:
[P]redicting … an inference output datapoint based on an inference input datapoint: This limitation could encompass mentally predicting the inference output based on an inference input.
[D]etermining a sequence of a predefined length, the respective sequence including further datasets progressively selected from a plurality of datasets based on a distance of the input datapoints thereof to the inference input datapoint: This limitation could encompass mentally determining the sequence including further datasets selected based on a distance to the inference input.
[D]etermining whether the inference input datapoint and the input datapoints of each of the further datasets included in the sequence respectively fulfill a first neighborhood criterion that is defined in the input space: This limitation could encompass mentally determining whether the datapoints satisfy a neighborhood criterion defined in the input space.
[D]etermining whether the inference output datapoint and the output datapoints of each of the further datasets included in the respective sequence respectively fulfill a second neighborhood criterion that is defined in the output space: This limitation could encompass mentally determining whether the datapoints satisfy a neighborhood criterion defined in the output space.
[F]or each sequence entry of the respective sequence: determining a cumulative fulfillment ratio based on how many of the further datasets included in the sequence up to the respective entry fulfill both the first neighborhood criterion and the second neighborhood criterion, thereby obtaining a trace of cumulative fulfillment ratios: This limitation could encompass mentally determining the fulfillment ratio based on how many datasets satisfy both criteria.
[P]erforming a comparison between the trace of the cumulative fulfillment ratios and the data structure determined via the computer-implemented method according to claim 1: Note that claim 1 is directed largely to mental processes as above. The remainder of the limitation could encompass mentally performing the comparison between the trace of cumulative fulfillment ratios and the data structure.
[B]ased on the comparison, selectively marking the inference output datapoint as reliable or unreliable: This limitation could encompass mentally marking the output as reliable or unreliable.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that the method is “computer-implemented” and performed in part “by a machine learning algorithm”. However, these amount to mere instructions to apply the judicial exception using a generic computer programmed with a generic class of computer algorithm. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites that the method is “computer-implemented” and performed in part “by a machine learning algorithm”. However, these amount to mere instructions to apply the judicial exception using a generic computer programmed with a generic class of computer algorithm. MPEP § 2106.05(f).
Claim 11
Step 1: A process, as above.
Step 2A Prong 1: The claim recites the same judicial exceptions as in claim 10.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that “the plurality of datasets form training data with which the machine-learning algorithm has been trained.” However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites that “the plurality of datasets form training data with which the machine-learning algorithm has been trained.” However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim 12
Step 1: The claim recites a computing device comprising a processor and a memory; therefore, the claim falls into the statutory category of machines.
Step 2A Prong 1: The claim recites the same judicial exceptions as in claim 1.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites a “computing device, comprising a processor and a memory, the processor being configured to load program code from the memory and execute the program code, wherein the processor is configured to execute the method … upon executing the program code.” However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites a “computing device, comprising a processor and a memory, the processor being configured to load program code from the memory and execute the program code, wherein the processor is configured to execute the method … upon executing the program code.” However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim 13
Step 1: The claim recites a computing device comprising a processor and a memory; therefore, the claim falls into the statutory category of machines.
Step 2A Prong 1: The claim recites the same judicial exceptions as in claim 10.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites a “computing device, comprising a processor and a memory, the processor being configured to load program code from the memory and execute the program code, wherein the processor is configured to execute the method … upon executing the program code.” However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites a “computing device, comprising a processor and a memory, the processor being configured to load program code from the memory and execute the program code, wherein the processor is configured to execute the method … upon executing the program code.” However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim 14
Step 1: For purposes of this rejection, it will be assumed that the claim falls into the statutory category of machines.
Step 2A Prong 1: The claim recites the same judicial exceptions as in claim 1.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites a “data collection comprising the data structure … and the plurality of datasets.” This limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites a “data collection comprising the data structure … and the plurality of datasets.” This limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim Rejections - 35 USC § 103
Claims 1-4, 9, 12, and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Fujitani et al. (US 20180197106) (“Fujitani”) in view of Chen et al. (US 20230162030) (“Chen”) and further in view of Brooks et al. (US 20200349498) (“Brooks”).
Regarding claim 1, Fujitani discloses “[a] computer-implemented method of enabling an assessment of a plurality of datasets forming training data for a machine-learning algorithm, each dataset of the plurality of datasets including a respective input datapoint in an input space and an associated output datapoint in an output space (apparatus may determine [enable an assessment of] a training data set by obtaining reference training data and target training data [plurality of datasets] comprising input data and output data – Fujitani, paragraphs 30-32), …, the method comprising:
for each dataset of the plurality of datasets: determining a respective sequence of a predefined length, the respective sequence including further datasets progressively selected from the plurality of datasets based on a distance of the input datapoints thereof to the input datapoint of the respective dataset (calculating section may calculate a degree of difference between target training data and each of a plurality of reference training data [number of reference training data to be evaluated = sequence of predefined length]; calculating section may calculate a degree of difference between two training data by calculating a sum or product of (i) Euclidean and/or Mahalanobis distance between the two input data; and (ii) Euclidean and/or Mahalanobis distance between two output data [note that this operation implicitly sorts the datapoints based on input distance and therefore progressively selects the datapoints based on this distance] – Fujitani, paragraphs 37-41);
for each dataset of the plurality of datasets: determining whether the input datapoint of the respective dataset and the input datapoints of each of the further datasets included in the respective sequence respectively fulfill a first neighborhood criterion that is defined in the input space (calculating section may calculate a degree of difference between target training data and each of a plurality of reference training data; calculating section may calculate a degree of difference between two training data by calculating a sum or product of (i) Euclidean and/or Mahalanobis distance between the two input data; and (ii) Euclidean and/or Mahalanobis distance between two output data – Fujitani, paragraphs 37-41; see also paragraph 90 (disclosing that the system determines whether the degree of difference between the target training data and the first reference training data is below a first threshold [first neighborhood criterion]; note that since the degree of difference includes a term calculating input difference, the comparison between the degree of difference and the first threshold is defined at least partially in the input space));
for each dataset of the plurality of datasets: determining whether the output datapoint of the respective dataset and the output datapoints of each of the further datasets included in the respective sequence respectively fulfill a second neighborhood criterion that is defined in the output space (calculating section may calculate a degree of difference between target training data and each of a plurality of reference training data; calculating section may calculate a degree of difference between two training data by calculating a sum or product of (i) Euclidean and/or Mahalanobis distance between the two input data; and (ii) Euclidean and/or Mahalanobis distance between two output data – Fujitani, paragraphs 37-41; see also paragraph 93 (disclosing that the system determines whether the degree of difference between the target training data and the farthest-from-target training data is below a second threshold [second neighborhood criterion] larger than the first threshold; note that since the degree of difference includes a term calculating output difference, the comparison between the degree of difference and the second threshold is defined at least partially in the output space)); [and]
for each dataset of the plurality of datasets and for each sequence entry of the respective sequence: determining … how many of the further datasets included in the sequence up to the respective entry fulfill both the first neighborhood criterion and the second neighborhood criterion (determining section may determine whether a degree of difference between the target training data and the first reference training data is below a first threshold; if not, the adding section generates a new training data set including the target training data; if so, the determining section determines whether a degree of difference between the target training data and the first reference training data is below a first threshold; if so [i.e., if both thresholds/criteria are satisfied], the target training data are added to the target training data set – Fujitani, paragraphs 90-94 and Fig. 7 [cardinality of the target training dataset after all data fulfilling both criteria are added = how many datasets fulfill both criteria]) ….”
Fujitani appears not to disclose explicitly the further limitations of the claim. However, Chen discloses that “each input datapoint correspond[s] to two-dimensional image data acquired using a camera (system comprises an image capturing device for capturing images of a scene including objects and a control unit configured to operate a neural network model for detecting objects in the scene; the image capturing device may be a camera – Chen, paragraphs 34-35 [note that images are two-dimensional]) ….”
Chen further discloses “determining a respective cumulative fulfillment ratio (method for selecting object samples for training of a neural network comprises determining an importance score for at least a portion of annotated object samples and defining a set of importance score thresholds [cumulative fulfillment ratios, “cumulative” because all data points that fulfill the higher threshold fulfill the lower threshold, so the lower threshold represents the cumulative count of data points that satisfy both that threshold and all higher thresholds] – Chen, paragraph 9) …; and
determining a data structure, an array dimension of the data structure resolving the sequences determined for each one of the plurality of datasets, a further array dimension of the data structure resolving the cumulative fulfillment ratio, each entry of the data structure including a count of datapoints that are associated with the respective cumulative fulfillment ratio at the respective sequence entry defined by the position along the array dimension and the further array dimension (Chen Fig. 7 and paragraph 83 disclose an array whose columns correspond to the importance thresholds [cumulative fulfillment ratios], whose rows correspond to the classes [i.e., sequences of data falling into that class], and whose entries are the count of datapoints of each class that fulfill each importance threshold [count of datapoints associated with the importance threshold/fulfillment ratio and the classes/sequences]).”
Chen and the instant application both relate to selecting machine learning training data and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Fujitani to determine a data structure having the claimed characteristics, as disclosed by Chen, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would provide a structured way to organize the data in a way that is both visually intelligible and allows for further analysis. See Chen, paragraphs 83-84.
Neither Fujitani nor Chen appears to disclose explicitly the further limitations of the claim. However, Brooks discloses that “each output datapoint indicat[es] objects on a railroad track depicted by the respective two-dimensional image data (first railcar [object on a railroad track] may be depicted in the frames generated during the first three seconds of the time period; the multi-object tracking algorithm can track [output] the presence of each individual railcar through the total 300 image frames [two-dimensional image data] generated – Brooks, paragraph 55; see also paragraph 28 (disclosing that the cars are on railroad tracks)) ….”
Brooks and the instant application both relate to image analysis of objects on railroad tracks and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Fujitani and Chen to use the system to track objects on railroad tracks, as disclosed by Brooks, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow the system to automate the inventory management process of objects traveling on the railroad tracks, thereby saving human effort. See Brooks, paragraphs 3-5.
Regarding claim 2, Fujitani, as modified by Chen/Brooks, discloses that “each entry of the data structure further comprises an identification of the datasets that are associated with the respective cumulative fulfillment ratio at the respective sequence entry defined by the position along the array dimension and the further array dimension (Chen Fig. 7 shows that the entries of the array are each associated with a class [i.e., an identification of which subset of the data is associated with each importance threshold/cumulative fulfillment ratio at each entry]).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the Fujitani/Brooks to identify the dataset associated with each entry of the array, as disclosed by Chen, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would provide a structured way to organize the data in a way that is both visually intelligible and allows for further analysis. See Chen, paragraphs 83-84.
Regarding claim 3, the rejection of claim 1 is incorporated. Fujitani further discloses “a predetermined distance offset in the input space between adjacent input datapoints of the respective datasets in the sequences (Fujitani Fig. 3 shows that the distance between datapoint T and closest datapoint A is given as predetermined distance offset DT-A; see also paragraphs 39-41 (disclosing that the distances between data points is determined at least in part based on the distance between the inputs)).”
Fujitani/Brooks appears not to disclose explicitly the further limitations of the claim. However, Chen discloses that “an increment of the array dimension corresponds to a predetermined … offset (Chen Fig. 7 shows that the importance thresholds are incremented by 0.1 [predetermined offset]) ….” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Fujitani/Brooks to provide for regular spacing between increments in an array dimension, as disclosed by Chen, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would provide a structured way to organize the data in a way that is both visually intelligible and allows for further analysis. See Chen, paragraphs 83-84.
Regarding claim 4, the rejection of claim 1 is incorporated. Fujitani further discloses “[a] method of assessing training data for training an algorithm, the training data having a plurality of datasets, each dataset of the plurality of datasets including a respective input datapoint in an input space and an associated output datapoint in an output space, the output datapoints of the plurality of datasets being ground-truth labels indicative of multiple classes to be predicted by the algorithm (apparatus may determine a training data set by obtaining reference training data and target training data [plurality of datasets] comprising input data and output data – Fujitani, paragraphs 30-32; output data of each reference training data may be represented by values or labels [indicative of classes to be predicted by the algorithm] – id. at paragraph 34) ….”
Fujitani/Brooks appears not to disclose explicitly the further limitations of the claim. However, Chen discloses “determining a data structure by the computer-implemented method according to claim 1 (see mapping of this element to Chen in the rejection of claim 1);
accessing the data structure thus determined (in order to reduce the effect of class imbalance, object samples are selected from each object class; the selected object samples fulfill a respective importance score threshold so that the variation in the number of samples from each object class is as small as possible [which requires accessing the data structure] – Chen, paragraph 82); [and]
on access to the data structure, assessing the training data (standard deviation of the number of counted object samples for each object class is calculated [calculation of standard deviation = assessment] – Chen, paragraph 84).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Fujitani/Brooks to assess the training data based on the data structure, as disclosed by Chen, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would provide a structured way to organize the data in a way that is both visually intelligible and allows for further analysis. See Chen, paragraphs 83-84.
Regarding claim 9, Fujitani, as modified by Chen/Brooks, discloses “upon assessing the training data, training the algorithm based on the training data (control unit is configured to operate a neural network model [algorithm] for detecting objects in the scene; for the neural network to be able to detect objects in the scene and classify them to be of a specific type, it is necessary that the neural network model has been trained for such detection on training data that represent each of a set of object classes – Chen, paragraph 53); and
upon training the algorithm, using the algorithm for solving inference tasks (control unit is configured to operate a neural network model for detecting objects [inference task] in the scene; for the neural network to be able to detect objects in the scene and classify them to be of a specific type, it is necessary that the neural network model has been trained for such detection on training data that represent each of a set of object classes – Chen, paragraph 53).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Fujitani/Brooks to use the training data to train an algorithm to perform an inference task, as disclosed by Chen, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would reduce human effort by allowing the task to be automated. See Chen, paragraph 53.
Regarding claim 12, Fujitani, as modified by Chen/Brooks, discloses “[a] computing device, comprising a processor and a memory, the processor being configured to load program code from the memory and execute the program code, wherein the processor is configured to execute the method according to claim 1 upon executing the program code (Fujitani Fig. 12 shows CPU 1200-12 and RAM 1200-14 and paragraph 123 discloses that the CPU operates according to programs stored in the RAM).”
Regarding claim 14, Fujitani, as modified by Chen/Brooks, discloses “[a] data collection comprising the data structure determined in accordance with claim 1 and the plurality of datasets (Chen Fig. 7 shows a data collection comprising the array/data structure and each class [dataset] associated with each entry of the array).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Fujitani/Brooks to organize the data into a data structure, as disclosed by Chen, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would provide a structured way to organize the data in a way that is both visually intelligible and allows for further analysis. See Chen, paragraphs 83-84.
Claims 5-6 are rejected under 35 U.S.C. 103 as being unpatentable over Fujitani in view of Chen and Brooks and further in view of Chrobak et al. (US 20230018914) (“Chrobak”).
Regarding claim 5, the rejection of claim 1 is incorporated. Chen further discloses “the count of the datapoints”, as shown in the rejection of claim 1. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Fujitani/Brooks to count the datapoints, as disclosed by Chen, for substantially the same reasons as given in the rejection of claim 1.
Neither Fujitani, Brooks, nor Chen appears to disclose explicitly the further limitations of the claim. However, Chrobak discloses “accessing the data structure by determining a plot of the data structure, with a contrast of plot values of the plot being associated with the count …, a first axis of the plot resolving the array dimension, and a second axis of the plot resolving the further array dimension (scatterplot of count of fish by distance from a camera subsystem shows a mix of lighter and darker regions in the plot reflecting the distribution of fish – Chrobak, paragraphs 37-41; see also Fig. 2 (showing that the scatterplots have two dimensions/axes corresponding to x- and y-directions of field of vision)); and
outputting the plot via a user interface (embodiments may be implemented on a computer having a display device for displaying information to the user [including the plots] – Chrobak, paragraph 72; see also paragraph 75 (disclosing that the client computer has a GUI)).”
Chrobak and the instant application both relate to plotting data and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Fujitani, Brooks, and Chen to plot the data using contrast to denote counts of objects, as disclosed by Chrobak, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would provide an informative visual through which a user can efficiently understand the data distribution. See Chrobak, paragraphs 37-41.
Regarding claim 6, Fuijitani, as modified by Chen, Brooks, and Chrobak, discloses that “assessing the training data comprises:
identifying a subset of the plurality of datasets by selecting parts of the plot (scatterplot of count of fish by distance from a camera subsystem shows a mix of lighter and darker regions in the plot reflecting the distribution of fish – Chrobak, paragraphs 37-41 [note that the entire scatterplot represents a subset of the dataset, albeit possibly an improper one]; see also Fig. 2 (showing that the scatterplots have two dimensions/axes corresponding to x- and y-directions of field of vision)); and
presenting datasets in the subset to the user via the user interface (embodiments may be implemented on a computer having a display device for displaying information to the user [including the plots]– Chrobak, paragraph 72; see also paragraph 75 (disclosing that the client computer has a GUI)).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Fujitani, Brooks, and Chen to display a subset of the dataset as a plot, as disclosed by Chrobak, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would provide an informative visual through which a user can efficiently understand the data distribution. See Chrobak, paragraphs 37-41.
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Fujitani in view of Chen, Brooks, and Chrobak and further in view of Pramod et al. (US 20230104757) (“Pramod”).
Regarding claim 7, neither Fujitani, Chen, Brooks, nor Chrobak appears to disclose explicitly the further limitations of the claim. However, Pramod discloses that “the step of presenting the datasets comprises highlighting the input datapoints or the output datapoints in a reduced-dimensionality plot of the input space or of the output space, respectively (output component may be communicatively coupled to display devices so as to permit the presentation of visualizations of output of subsystems of the detection system; for example, reduced dimension projections of classifier output [output datapoints] (e.g., cluster plots) [reduced-dimensionality plot of the output space] may be presented visually to a user – Pramod, paragraph 50).”
Pramod and the instant application both relate to visual plots of machine learning outputs and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Fujitani, Chen, Brooks, and Chrobak to plot the outputs in a reduced-dimensionality plot, as disclosed by Pramod, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would increase the intelligibility of the outputs of the model to a user. See Pramod, paragraph 50.
Claim 8 is rejected under 35 U.S.C. 103 as being unpatentable over Fujitani in view of Chen, Brooks, and Chrobak and further in view of Cogill et al. (US 20160267391) (“Cogill”).
Regarding claim 8, the rejection of claim 5 is incorporated. Chen further discloses a “cumulative fulfillment ratio”, as shown in the rejection of claim 1. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Fujitani/Chrobak/Brooks to provide a cumulative fulfillment ratio, as disclosed by Chen, for substantially the same reasons as given in the rejection of claim 1.
Neither Fujitani, Chrobak, Brooks, nor Chen appears to disclose explicitly the further limitations of the claim. However, Cogill discloses “obtaining a selection of a given dataset of the plurality of datasets (one may represent the possible evolutions of the state of each resource over time [each resource/time dataset = one dataset of a plurality] by plotting several sample paths in a plot, where the horizontal axis is the time and the vertical axis is the state or the value of some function thereof – Cogill, paragraph 47); and
highlighting in the plot an evolution of the respective [property] of the given dataset for various positions along the first axis (one may represent the possible evolutions of the state of each resource over time by plotting several sample paths in a plot, where the horizontal axis [first axis] is the time and the vertical axis is the state [property of the dataset] or the value of some function thereof – Cogill, paragraph 47 [note that “highlighting” is being interpreted as being equivalent to “displaying”]).”
Cogill and the instant application both relate to plotting evolutions of data and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Fujitani, Chen, Brooks, and Chrobak to plot the evolution of a property of a dataset, as disclosed by Cogill, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow the user to understand how the data are changing over time. See Cogill, paragraph 47.
Claims 10-11 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Fujitani in view of Chen and Brooks and further in view of Amar (US 11640470) (“Amar”).
Regarding claim 10, the rejection of claim 1 is incorporated. Fujitani further discloses “[a] computer-implemented method of supervising inference tasks provided by a machine-learning algorithm, the method comprising: …
determining a sequence of a predefined length, the respective sequence including further datasets progressively selected from a plurality of datasets based on a distance of the input datapoints thereof to the inference input datapoint (calculating section may calculate a degree of difference between target training data and each of a plurality of reference training data [number of reference training data to be evaluated = sequence of predefined length]; calculating section may calculate a degree of difference between two training data by calculating a sum or product of (i) Euclidean and/or Mahalanobis distance between the two input data; and (ii) Euclidean and/or Mahalanobis distance between two output data [note that this operation implicitly sorts the datapoints based on input distance and therefore progressively selects the datapoints based on this distance] – Fujitani, paragraphs 37-41);
determining whether the inference input datapoint and the input datapoints of each of the further datasets included in the sequence respectively fulfill a first neighborhood criterion that is defined in the input space (calculating section may calculate a degree of difference between target training data and each of a plurality of reference training data; calculating section may calculate a degree of difference between two training data by calculating a sum or product of (i) Euclidean and/or Mahalanobis distance between the two input data; and (ii) Euclidean and/or Mahalanobis distance between two output data – Fujitani, paragraphs 37-41; see also paragraph 90 (disclosing that the system determines whether the degree of difference between the target training data and the first reference training data is below a first threshold [first neighborhood criterion]; note that since the degree of difference includes a term calculating input difference, the comparison between the degree of difference and the first threshold is defined at least partially in the input space));
determining whether the inference output datapoint and the output datapoints of each of the further datasets included in the respective sequence respectively fulfill a second neighborhood criterion that is defined in the output space (calculating section may calculate a degree of difference between target training data and each of a plurality of reference training data; calculating section may calculate a degree of difference between two training data by calculating a sum or product of (i) Euclidean and/or Mahalanobis distance between the two input data; and (ii) Euclidean and/or Mahalanobis distance between two output data – Fujitani, paragraphs 37-41; see also paragraph 93 (disclosing that the system determines whether the degree of difference between the target training data and the farthest-from-target training data is below a second threshold [second neighborhood criterion] larger than the first threshold; note that since the degree of difference includes a term calculating output difference, the comparison between the degree of difference and the second threshold is defined at least partially in the output space));
for each sequence entry of the respective sequence: determining … how many of the further datasets included in the sequence up to the respective entry fulfill both the first neighborhood criterion and the second neighborhood criterion (determining section may determine whether a degree of difference between the target training data and the first reference training data is below a first threshold; if not, the adding section generates a new training data set including the target training data; if so, the determining section determines whether a degree of difference between the target training data and the first reference training data is below a first threshold; if so [i.e., if both thresholds/criteria are satisfied], the target training data are added to the target training data set – Fujitani, paragraphs 90-94 and Fig. 7 [cardinality of the target training dataset after all data fulfilling both criteria are added = how many datasets fulfill both criteria]) ….”
Fujitani/Brooks appears not to disclose explicitly the further limitations of the claim. However, Chen discloses “predicting, by the machine-learning algorithm, an inference output datapoint based on an inference input datapoint (neural network is trained to predict the class and location of an object [inference output datapoint] in an image [inference input datapoint] – Chen, paragraph 17); …
determining a cumulative fulfillment ratio (method for selecting object samples for training of a neural network comprises determining an importance score for at least a portion of annotated object samples and defining a set of importance score thresholds [cumulative fulfillment ratios, “cumulative” because all data points that fulfill the higher threshold fulfill the lower threshold, so the lower threshold represents the cumulative count of data points that satisfy both that threshold and all higher thresholds] – Chen, paragraph 9) …; …
obtaining a trace of cumulative fulfillment ratios (see Chen Fig. 7 and note that the number of samples for each class above the performance threshold is cumulative of the number of samples for each class above the higher performance thresholds and that these counts collectively form a trace of cumulative fulfillment ratios) …; [and]
performing a comparison between the trace of the cumulative fulfillment ratios and the data structure determined via the computer-implemented method according to claim 1 (standard deviation of the number of counted object samples for each object class and each importance score threshold [implying that the importance score thresholds/cumulative fulfillment ratios, and by extension the counts that form the trace, are compared to classes in the data structure] – Chen, paragraph 84) ….” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Fujitani/Brooks to perform analysis based on a trace of ratios, as disclosed by Chen, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would provide a structured way to organize the data in a way that is both visually intelligible and allows for further analysis. See Chen, paragraphs 83-84.
Neither Fujitani, Brooks, nor Chen appears to disclose explicitly the further limitations of the claim. However, Amar discloses “based on the comparison, selectively marking the inference output datapoint as reliable or unreliable (if an entropy-based measure of reliability associated with the top-k number of outputs exceeds a threshold [i.e., based on a comparison], the top ranked output is accepted as being reliable – Amar, col. 18, ll. 32-55).”
Amar and the instant application both relate to determining the reliability of machine learning outputs and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Fujitani, Brooks, and Chen to mark the output as reliable or unreliable based on the comparison, as disclosed by Amar, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would increase the model quality by ensuring that its outputs are reliable. See Amar, col. 18, ll. 32-55.
Regarding claim 11, Fujitani, as modified by Chen, Brooks, and Amar, discloses that “the plurality of datasets form training data with which the machine-learning algorithm has been trained (Fujitani Fig. 11 shows that after all target training data are selected, a selected data set is formed and training is performed using the selected data set).”
Regarding claim 13, Fujitani, as modified by Chen, Brooks, and Amar, discloses “[a] computing device, comprising a processor and a memory, the processor being configured to load program code from the memory and execute the program code, wherein the processor is configured to execute the method according to claim 10 upon executing the program code (Fujitani Fig. 12 shows CPU 1200-12 and RAM 1200-14 and paragraph 123 discloses that the CPU operates according to programs stored in the RAM).”
Response to Arguments
Applicant's arguments filed July 23, 2026 (“Remarks”) have been fully considered but they are not persuasive.
Applicant argues, to the best of Examiner’s understanding given the limited intelligibility of the argument, that the amended claims are now eligible under 35 USC § 101 because they are directed to machine learning and are now directed to a railroad object detection use case. Remarks at 9-10. However, the mere recitation of machine learning is not of itself enough to confer eligibility on the claims. This is the central holding of the U.S. Court of Appeals for the Federal Circuit in Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025). Here, the claims as a whole are directed to a mentally performable process of determining whether input and output datapoints satisfy neighborhood criteria and using the results to create a data structure. The claims are not directed to an improvement in machine learning as such, but to the use of machine learning to implement this abstract idea. The machine learning is invoked purely as an intended use of the training data generated by this mentally performable algorithm. Moreover, confining the judicial exception to the use case of detecting objects on railroad tracks merely defines the field of use of the judicial exception and is not enough to confer eligibility. MPEP § 2106.05(h).
Regarding the rejections under 35 USC § 103, Applicant alleges that (a) Fujitani fails to disclose that a sequence of nearest neighbors is determined for each dataset because Fujitani processes one data point at a time; (b) Fujitani applies a threshold to a combined value that mixes input and output distance and thus cannot teach the claim’s recitation of two independently operable criteria; (c) Chen’s importance scores are not derived from a neighborhood analysis of the input or output spaces; (d) the rows of Chen’s table are categorical labels rather than positions in a nearest-neighbor sequence ordered by input-space distance; and (e) Fujitani is not combinable with Chen because the statement of motivation to combine is a “conclusory statement that is not based on the actual teachings of the references or the art as a whole”. Remarks at 11-16.
However, regarding (a), insofar as each input datapoint necessarily belongs to a dataset, and insofar as in the limit each datapoint may be regarded as its own dataset, the reference meets the claim. Fujitani discloses sorting datapoints based at least in part on input data, thereby determining a sequence of datasets of a predefined length (viz., 1) for each dataset (i.e., datapoint) of the plurality of datasets.
Regarding (b), it is unclear to what “combined value” Applicant is referring. Rather, Fujitani, at paragraphs 37-41, explicitly discloses that the calculating section calculates a sum or product of (i) a distance between input data, and (ii) a distance between output data. While the sum is perhaps a “combined value,” it is made up of distinct, individual terms, namely input distance and output distance, which correspond to the claimed “first neighborhood criterion” and “second neighborhood criterion,” respectively.
Regarding (c), this argument is a classic case of Applicant attacking references individually where the rejection is based on the combination. Examiner never represented that Chen performs a neighborhood analysis of input or output spaces. That is Fujitani’s role in the rejection. One cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986).
Regarding (d), Applicant is arguing that the reference fails to disclose features that are not claimed. The independent claim does not require that the data structure have any ordering whatsoever, let alone an ordering based on input-space distance. Nor does the claim require that the rows of the data structure correspond to positions in a nearest-neighbor sequence. Rather, the claim requires merely that the rows “resolv[e] the sequences determined for each one of the plurality of datasets”. Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. See In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993).
Regarding (e), it is noted that the motivation to combine need not come directly from the secondary reference. “The motivation to combine may be implicit and may be found in the knowledge of one of ordinary skill in the art, or, in some cases, from the nature of the problem to be solved.” MPEP § 2143(I)(G) (citing DyStar Textilfarben GmbH & Co. Deutschland KG v. C.H. Patrick Co., 464 F.3d 1356, 1366, 80 USPQ2d 1641, 1649 (Fed. Cir. 2006)). Here, an ordinary artisan would recognize that organizing the data generated by the method of Fujitani as a data structure or table as disclosed by Chen would have the benefits of making the data both more visually intelligible to a human user and more organized for subsequent retrieval and use.
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 RYAN C VAUGHN whose telephone number is (571)272-4849. The examiner can normally be reached M-R 7:00a-5:00p ET.
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.
/RYAN C VAUGHN/
Primary Examiner, Art Unit 2125
Prosecution Insights