Notice of Pre-AIA or AIA Status
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
2. This action is in response to the original filing on 06/14/2024. Claims 1-17 are pending and have been considered below.
Information Disclosure Statement
3. The information disclosure statement (IDS(s)) submitted on 05/19/2025 is/are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections – 35 USC § 103
4. 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 of this title, 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.
5. Claims 1-4, 9, 10, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Jiang et al. (Delving into Sample Loss Curve to Embrace Noisy and Imbalanced Data, arXiv, published 2021, pages 1-9) in view of Sheikholeslami et al. (The Impact of Importance-Aware Dataset Partitioning on Data-Parallel Training of Deep Neural Networks, SPRINGER, published 09 June 2023, pages 74-89), and further in view of Li et al. (U.S. Patent Application Pub. No. US 20210089883 A1).
Claim 1: Jiang teaches a method for training a neural network to detect label errors (i.e. Corrupted labels and class imbalance are commonly encountered in practically collected training data, which easily leads to over-fitting of deep neural networks (DNNs). In the probing stage, we train the network on the whole biased training data without intervention, and record the loss curve of each sample as an additional attribute; In the allocating stage, we feed the resulting attribute to a newly designed curve perception network, named CurveNet, to learn to identify the bias type of each sample and assign proper weights through meta-learning adaptively; abs, page 1), comprising:
training a neural network on a training dataset (i.e. In light of this, we propose to take advantage of the in formative training loss curve to distinguish clean samples of tail class from noisy samples, and generate proper sample weights accordingly. To this end, we propose a novel probe and-allocate training strategy: In the probing stage, we train a classifier with cyclical learning rate on the entire biased training data, and record the loss curve of each sample; abs, page 2), wherein the neural network produces one or more training loss data samples for each epoch of a plurality of training epochs (i.e. we delve into the loss curves throughout the whole training process and find distinguishable trends and characteristics between the noisy sample and clean tail sample … Gathering all of the loss value li,t of the ith sample together as a one-dimensional vector Li = [li,0,li,1,··· ,li,T], where T represents the number of training epoch; Overall Structure, CurveNet, pages 1-4), wherein training includes:
recording training loss of each data sample in every epoch (i.e. we train the network on the whole biased training data without intervention, and record the loss curve of each sample as an additional attribute… Gathering all of the loss value li,t of the ith sample together as a one-dimensional vector Li = [li,0,li,1,··· ,li,T], where T represents the number of training epoch; abs, CurveNet, pages 1-4); and
creating a loss (i.e. Gathering all of the loss value li,t of the ith sample together as a one-dimensional vector Li = [li,0,li,1,··· ,li,T], where T represents the number of training epoch. The normalized loss vectors can be denoted as I; CurveNet, page 4), wherein the loss is constructed as: (i.e. Gathering all of the loss value li,t of the ith sample together as a one-dimensional vector Li = [li,0,li,1,··· ,li,T], where T represents the number of training epoch. The normalized loss vectors can be denoted as I; CurveNet, page 4); and
forming a refined neural network by refining the neural network (i.e. Figure 2, the whole structure is composed of two stages of probing-stage as main-network for classification and allocating-stage for parameters refinement. Similar to the re-weighting network mentioned in Meta-weight-net, the allocating-stage adopts the meta-learning idea and allows the weighted loss values to guide the training of the classification network, giving the classifier more emphasis on hard positive samples while being robust to noise; Overall Structure, page 3), where refining includes applying a algorithm to the loss, the algorithm configured to separate samples into either a category of clean labels or noisy labels (i.e. Figure 5. It is clearly observable that in all the classes with different amount of samples our method distinguishes noisy and clean samples well; Image Classification on CIFAR10, pages 5-6).
Jiang does not explicitly teach a loss matrix, wherein the loss matrix is constructed as: |training loss samples|×|epochs|; applying a clustering algorithm to the loss, the clustering algorithm configured to separate samples.
However, Sheikholeslami teaches creating a loss matrix, wherein the loss matrix is constructed as: |training loss samples|×|epochs| (i.e. during warmup training, we collect the loss values (the result of the forward pass) of each example across Ewarmup epochs. At the end of warmup training, we will have a matrix such as in Fig. 3. In this matrix, each row corresponds to a single example, and each column corresponds to an epoch. Hence, an element ai,j in the matrix is the loss value of example i in epoch j; Section 3.2, pages 79-80).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the invention of Jiang to include the feature of Sheikholeslami. One would have been motivated to make this modification because implementing the example epoch loss matrix in order to efficiently store, access, and process the loss values associated with each training sample across multiple training epochs.
However, Li teaches applying a clustering algorithm to the loss (i.e. a Gaussian Mixture Model (GMM) is dynamically fit on per-sample loss distribution to divide the training samples; para. [0027]), the clustering algorithm configured to separate samples into either a category of clean labels or noisy labels (i.e. the per-sample loss of the first network with first set of parameters θ(1) may be modeled using a mixture model (e.g., a GMM model) to obtain clean probability W(2) for the second network with second set of parameters θ(2). For further example, the per-sample loss of the second network with second set of parameters θ(2) may be modeled using a mixture model (e.g., a GMM model) to obtain clean probability W(1) for the first network with first set of parameters θ(1). In some examples, a two-component GMM is fitted to the loss l(θ) (e.g., with a confidence penalty for asymmetric noise) using the Expectation-Maximization algorithm … Referring to FIG. 14, AUC curves illustrate that the dataset division in DivideMix helps to eliminate label noise. AUC curves 1402, 1404, and 1406 illustrate clean/noisy classification on CIFAR-10 training samples with 20% label noise, 50% label noise, and 80% label noise respectively; para. [0027, 0034, 0064]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang and Sheikholeslami to include the feature of Li. One would have been motivated to make this modification because it provides a technique for separating training samples into mostly clean and mostly noisy groups. This helps filter noisy label errors and avoids confirmation bias.
Claim 2: Jiang, Sheikholeslami, and Li teach the method of claim 1. Jiang further teaches comprising using a noise generator to create the training dataset (i.e. We construct biased training dataset with varying noisy and imbalance ratios by manually adjusting the sample number of each class and adding corrupted labels to the clean and balanced dataset such as CIFAR10 and CIFAR100; page 5).
Claim 3: Jiang, Sheikholeslami, and Li teach the method of claim 1. Jiang does not explicitly teach removing noisy labels before a subsequent period of training.
However, Li further teaches comprising removing noisy labels before a subsequent period of training (i.e. discards those noisy labels for those noisy samples, and leverages those noisy samples as unlabeled data to regularize the neural network model from overfitting and improve generalization performance; para. [0026, 0036-0039]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang and Sheikholeslami to include the feature of Li. One would have been motivated to make this modification because it provides a technique for separating training samples into mostly clean and mostly noisy groups. This helps filter noisy label errors and avoids confirmation bias.
Claim 4: Jiang, Sheikholeslami, and Li teach the method of claim 3. Jiang does not explicitly teach dynamically replacing noisy labels with a prediction of a neural network during a period of training.
However, Li further teaches dynamically replacing noisy labels with a prediction of a neural network during a period of training (i.e. during the co-guessing process, for unlabeled samples, the ensemble of both networks are used to make reliable guesses for labels of those unlabeled samples; para. [0028, 0044, 0047, 0049]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang and Sheikholeslami to include the feature of Li. One would have been motivated to make this modification because it provides a technique for separating training samples into mostly clean and mostly noisy groups. This helps filter noisy label errors and avoids confirmation bias.
Claims 9, 10, and 17 are similar in scope to Claims 1, 2 and are rejected under a similar rationale.
6. Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Jiang in view of Sheikholeslami, Li, and further in view of Sallee et al. (U.S. Patent Application Pub. No. US 20190354857 A1).
Claim 5: Jiang, Sheikholeslami, and Li teach the method of claim 3. Jiang does not explicitly teach statically replacing noisy labels with a prediction of a neural network, updated before a period of training by using the training of a previous round of the neural network.
However, Salle teaches statically replacing noisy labels with a prediction of a neural network (i.e. The pseudolabels use the existing partially trained model to supply alternate labels rather than requiring any architectural changes to the model … Pseudolabels can be used in place of assigned but noisy labels; para. [0023, 0034]), updated before a period of training (i.e. The ML trainer 106 receives the class vector 108, the data 102, a label 104 in a first number of epochs of ML training and a pseudolabel 116 in a subsequent epoch after the first number of epochs and produces updated model parameters 110; para. [0028]) by using the training of a previous round of the neural network (i.e. Pseudolabels can be part of a semi-supervised approach to ML training. Pseudolabels are training labels assigned by a partially trained model, or by a model trained with a subset of labeled data … The pseudolabel generator 112 can, based on the label 104 and the class vector 108, determine a pseudolabel 116 to be associated with the data 102 in a next training epoch; para. [0023, 0030]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, and Li to include the feature of Salle. One would have been motivated to make this modification because it improves robustness to label noise.
7. Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Jiang in view of Sheikholeslami, Li, and further in view of Kumar et al. (U.S. Patent Application Pub. No. US 20230024955 A1).
Claim 6: Jiang, Sheikholeslami, and Li teach the method of claim 1. Jiang does not explicitly teach using the neural network to classify received data as defective.
However, Kumar teaches using the refined neural network to classify received data as defective (i.e. an image may be applied to a first neural network trained to determine whether the image includes an exposure defect … to classify images into well-exposed and defective images as an output; para. [0004, 0023, 0037]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, and Li to include the feature of Kumar. One would have been motivated to make this modification because it provides an automated and consistent technique for identifying defective received data and permits deployment of the trained model to defect detection.
8. Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Jiang in view of Sheikholeslami, Li, Kumar, and further in view of Lin et al. (U.S. Patent Application Pub. No. US 20200200687 A1).
Claim 7: Jiang, Sheikholeslami, Li, and Kumar teach the method of claim 6. Jiang does not explicitly teach wherein classifying the received data as defecting includes using the neural network to classify an image of a label as a label defect.
However, Kumar further teaches wherein classifying the received data as defecting includes using the refined neural network to classify an image (i.e. an image may be applied to a first neural network trained to determine whether the image includes an exposure defect … to classify images into well-exposed and defective images as an output; para. [0004, 0023, 0037]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, and Li to include the feature of Kumar. One would have been motivated to make this modification because it provides an automated and consistent technique for identifying defective received data and permits deployment of the trained model to defect detection.
However, Lin teaches wherein classifying the received data as defecting includes using the refined network to classify an image of a label as a label defect (i.e. acquiring image information of a label on a material to be inspected by a visual inspection mechanism, analyzing and judging the image information, determining whether the image information corresponds to a defective label; para. [0024, 0060, 0061]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, Li, and Kumar to include the feature of Lin. One would have been motivated to make this modification because it provides an efficient method for an automated inspection of labels for defects.
9. Claim 8 is rejected under 35 U.S.C. 103 as being unpatentable over Jiang in view of Sheikholeslami, Li, Kumar, and further in view of Arik et al. (U.S. Patent Application Pub. No. US 20210034977 A1).
Claim 8: Jiang, Sheikholeslami, Li, and Kumar teach the method of claim 6. Jiang does not explicitly teach wherein classifying the received data as defecting includes using the neural network to classify data in a tabular dataset as defective.
However, Kumar further teaches wherein classifying the received data as defecting includes using the refined neural network to classify data as defective (i.e. an image may be applied to a first neural network trained to determine whether the image includes an exposure defect … to classify images into well-exposed and defective images as an output; para. [0004, 0023, 0037]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, and Li to include the feature of Kumar. One would have been motivated to make this modification because it provides an automated and consistent technique for identifying defective received data and permits deployment of the trained model to defect detection.
However, Arik teaches wherein classifying the received data as defecting includes using the refined neural network to classify data in a tabular dataset (i.e. TabNet is a neural network designed to learn in a decision tree-like manner. In other words, TabNet aims to offer interpretability and sparse feature selection. TabNet inputs raw tabular data without any feature pre-processing and is trained using gradient descent-based optimization to learn flexible representations and enable flexible integration into end-to-end learning … the encoder 202 is capable of performing either classification or regression; para. [0021, 0031]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, Li, and Kumar to include the feature of Lin. One would have been motivated to make this modification because it provides defect classification technique to tabular input data, thereby extending the defect classification technique to another known data representation.
10. Claims 11 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over Jiang in view of Sheikholeslami, Li, and further in view of Sun et al. (U.S. Patent Application Pub. No. US 20220166461 A1).
Claim 11: Jiang, Sheikholeslami, and Li teach the system of claim 9. Jiang does not explicitly teach wherein the network is transmitted to a remote processing unit for use in classifying data as including an error or being free of errors.
However, Sun teaches wherein the refined neural network is transmitted to a remote processing unit for use in classifying data as including an error or being free of errors (i.e. The collected signal data is uploaded, for example, periodically, to a centralized processor for training the trained neural network model 300, where the trained neural network model 400 resides, such as along a network remote from the vehicle 100, the local computer 110 of the vehicle 100, the external computer 142, or the cloud … The trained neural network model is deployed, for example, as a software package (that may be downloadable, for instance when the vehicle is parked and connected to an external wired or wireless network connection), for installation on the local vehicle 100 computer, e.g., processors 110, or external computer 142 (e.g., processors) … The trained model analyzes the signal data, and determines whether the corresponding cable has a fault, and if there is a detected fault, identifying the fault, including classifying, the type of the cable fault (cable fault type) among known types of cable faults; para. [0057, 0077, 0079]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, and Li to include the feature of Sun. One would have been motivated to make this modification because the neural network trained at a remote location can be distributed over a network to another computer and perform classification on locally received data.
Claim 12: Jiang, Sheikholeslami, Li, and Sun teach the system of claim 11. Jiang does not explicitly teach wherein the neural network is transmitted to a remote processing unit for use in classifying images as clean labels or noisy labels.
Li further teaches wherein the refined neural network for use in classifying images as clean labels or noisy labels (i.e. the model's output softmax probability for class c, D=(X,Y)={(χi,yi)}i=1 N denotes the training data, χi is a sample (e.g., an image), and yi ∈ {0,1}c is the one-hot label over C classes, θ denotes the model parameters; para. [0027, 0030, 0041, 0064]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang and Sheikholeslami to include the feature of Li. One would have been motivated to make this modification because it provides a technique for separating training samples into mostly clean and mostly noisy groups. This helps filter noisy label errors and avoids confirmation bias.
Sun further teaches wherein the refined neural network is transmitted to a remote processing unit for use in classifying (i.e. The collected signal data is uploaded, for example, periodically, to a centralized processor for training the trained neural network model 300, where the trained neural network model 400 resides, such as along a network remote from the vehicle 100, the local computer 110 of the vehicle 100, the external computer 142, or the cloud … The trained neural network model is deployed, for example, as a software package (that may be downloadable, for instance when the vehicle is parked and connected to an external wired or wireless network connection), for installation on the local vehicle 100 computer, e.g., processors 110, or external computer 142 (e.g., processors) … The trained model analyzes the signal data, and determines whether the corresponding cable has a fault, and if there is a detected fault, identifying the fault, including classifying, the type of the cable fault (cable fault type) among known types of cable faults; para. [0057, 0077, 0079]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, and Li to include the feature of Sun. One would have been motivated to make this modification because the neural network trained at a remote location can be distributed over a network to another computer and perform classification on locally received data.
11. Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Jiang in view of Sheikholeslami, Li, Sun, and further in view of Qiu et al. (U.S. Patent Application Pub. No. US 20220284301 A1).
Claim 13: Jiang, Sheikholeslami, Li, and Sun teach the system of claim 11. Jiang does not explicitly teach wherein the neural network is transmitted to a remote processing unit for use in classifying data in a tabular dataset as including an error or being free of errors.
Sun further teaches wherein the refined neural network is transmitted to a remote processing unit for use in classifying data as including an error or being free of errors (i.e. The collected signal data is uploaded, for example, periodically, to a centralized processor for training the trained neural network model 300, where the trained neural network model 400 resides, such as along a network remote from the vehicle 100, the local computer 110 of the vehicle 100, the external computer 142, or the cloud … The trained neural network model is deployed, for example, as a software package (that may be downloadable, for instance when the vehicle is parked and connected to an external wired or wireless network connection), for installation on the local vehicle 100 computer, e.g., processors 110, or external computer 142 (e.g., processors) … The trained model analyzes the signal data, and determines whether the corresponding cable has a fault, and if there is a detected fault, identifying the fault, including classifying, the type of the cable fault (cable fault type) among known types of cable faults; para. [0057, 0077, 0079]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, and Li to include the feature of Sun. One would have been motivated to make this modification because the neural network trained at a remote location can be distributed over a network to a another computer and perform classification on locally received data.
However, Qiu teaches wherein the refined neural network for use in classifying data in a tabular dataset as including an error or being free of errors (i.e. The anomaly detector may thus be trained to be applied to data timeseries as data samples and may thus identify whether a data timeseries is considered normal or abnormal … Typically, in such tabular data, the columns may define attributes while the rows define value of the attributes for respective data samples, or vice versa (e.g., the function of columns and rows may be switched); para. [0037, 0039, 0040]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, Li, and Sun to include the feature of Qiu. One would have been motivated to make this modification because it extends the error detection functionality to a structured data format.
12. Claims 14 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Jiang in view of Sheikholeslami, Li, and further in view of Ge et al. (U.S. Patent Application Pub. No. US 20220189005 A1).
Claim 14: Jiang, Sheikholeslami, and Li teach the system of claim 9. Jiang does not explicitly teach comprising a requestor computing device, the requestor computing device configured to send a request to the at least one processing unit to classify data as including an error or being free of errors.
However, Ge teaches comprising a requestor computing device (i.e. a client computer having a graphical user interface or a web interface through which a user can interact with an implementation of the subject matter described herein; para. [0045]), the requestor computing device configured to send a request to the at least one processing unit to classify data (i.e. the method further includes receiving the user input indicative of a user request to quantify at least one defect in the image of the target object … a user can provide the user input 222 requesting a binary determination associated with a defect (e.g., whether the defect(s) of a certain type (or having certain characteristics) is present or absent in the target object); para. [0007, 0023]) as including an error or being free of errors (i.e. The binary discriminator code is configured to output an affirmative inference result indicative of presence of the at least one defect or a negative inference result indicative of absence of the at least one defect; para. [0005]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, and Li to include the feature of Ge. One would have been motivated to make this modification because it permits remote users or applications to access and utilize the trained classification functionality without requiring the classification operation to be initiated locally at the training system.
Claim 15: Jiang, Sheikholeslami, Li, and Ge teach the system of claim 14. Jiang does not explicitly teach wherein the requestor computing device is configured to send a request to the at least one processing unit to classify an image of a label as a clean label or a noisy label.
Li further teaches the at least one processing unit to classify an image of a label as a clean label or a noisy label (i.e. During the co-divide process, for each network, a Gaussian Mixture Model (GMM) is dynamically fit on per-sample loss distribution to divide the training samples into a labeled set (e.g., including samples that are mostly clean/less likely to be noisy) and an unlabeled set (e.g., including samples that are highly likely to be noisy). The divided data (including the labeled set and unlabeled set) from one network is then used to train the other network. By using the co-divide process, the two networks are kept diverged, and may be used to filter different types of error and avoid confirmation bias in self-training; para. [0027, 0030, 0041, 0064]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang and Sheikholeslami to include the feature of Li. One would have been motivated to make this modification because it provides a technique for separating training samples into mostly clean and mostly noisy groups. This helps filter noisy label errors and avoids confirmation bias.
However, Ge further teaches wherein the requestor computing device is configured to send a request to the at least one processing unit to classify (i.e. the method further includes receiving the user input indicative of a user request to quantify at least one defect in the image of the target object … a user can provide the user input 222 requesting a binary determination associated with a defect (e.g., whether the defect(s) of a certain type (or having certain characteristics) is present or absent in the target object); para. [0007, 0023]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, and Li to include the feature of Ge. One would have been motivated to make this modification because it permits remote users or applications to access and utilize the trained classification functionality without requiring the classification operation to be initiated locally at the training system.
13. Claim 16 is rejected under 35 U.S.C. 103 as being unpatentable over Jiang in view of Sheikholeslami, Li, Ge, and further in view of Qiu et al. (U.S. Patent Application Pub. No. US 20220284301 A1).
Claim 16: Jiang, Sheikholeslami, Li, and Ge teach the method of claim 14. Jiang does not explicitly teach wherein the requestor computing device is configured to send a request to the at least one processing unit to classify data in a tabular dataset as including an error or begin free of errors.
However, Ge further teaches wherein the requestor computing device is configured to send a request to the at least one processing unit to classify data (i.e. the method further includes receiving the user input indicative of a user request to quantify at least one defect in the image of the target object … a user can provide the user input 222 requesting a binary determination associated with a defect (e.g., whether the defect(s) of a certain type (or having certain characteristics) is present or absent in the target object); para. [0007, 0023]) as including an error or begin free of errors (i.e. The binary discriminator code is configured to output an affirmative inference result indicative of presence of the at least one defect or a negative inference result indicative of absence of the at least one defect; para. [0005]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, and Li to include the feature of Ge. One would have been motivated to make this modification because it permits remote users or applications to access and utilize the trained classification functionality without requiring the classification operation to be initiated locally at the training system.
However, Qiu teaches the at least one processing unit to classify data in a tabular dataset as including an error or begin free of errors (i.e. The anomaly detector may thus be trained to be applied to data timeseries as data samples and may thus identify whether a data timeseries is considered normal or abnormal … Typically, in such tabular data, the columns may define attributes while the rows define value of the attributes for respective data samples, or vice versa (e.g., the function of columns and rows may be switched); para. [0037, 0039, 0040]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Jiang, Sheikholeslami, Li, Ge to include the feature of Qiu. One would have been motivated to make this modification because it extends the error detection functionality to a structured data format.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure.
Bai et al. (Pub. No. US 20220058922 A1), a trajectory error prediction machine learning model to predict a respective trajectory error for each respective future trajectory determined for each respective object state data item in the object state dataset; wherein the trajectory error prediction machine learning model is trained by predicting a training trajectory error for each at least one historical object state data in a historical epoch object state dataset to learn correlations between state and trajectory error based on trajectory parameters.
It is noted that any citation to specific pages, columns, lines, or figures in the prior art references and any interpretation of the references should not be considered to be limiting in any way. A reference is relevant for all it contains and may be relied upon for all that it would have reasonably suggested to one having ordinary skill in the art. In re Heck, 699 F.2d 1331, 1332-33, 216 U.S.P.Q. 1038, 1039 (Fed. Cir. 1983) (quoting In re Lemelson, 397 F.2d 1006, 1009, 158 U.S.P.Q. 275, 277 (C.C.P.A. 1968)).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TAN TRAN whose telephone number is (303)297-4266. The examiner can normally be reached on Monday - Thursday - 8:00 am - 5:00 pm MT.
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, Matt Ell can be reached on 571-270-3264. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/TAN H TRAN/Primary Examiner, Art Unit 2141
Prosecution Insights