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 .
Claim Objections
Claim 1 objected to because of the following informalities: non-transitory computer-readable media was used here and is not present in any of the other claims. It should be non-transitory computer- readable medium. Appropriate correction is required.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim 1 is rejected under 103 as being unpatentable over Doan Huu et al (US 20250021866 A1) (“Huu”) in view of Mo et al (US 20250156744 A1) (“Mo”) and Westemeier et al (NPL: Generating Synthetic Time Series for Machine-Learning-Empowered Monitoring of Electric Motor Test Benches)
Regarding Claim 1, Huu teaches input the combined set of time series into a plurality of candidate forecasting algorithms, where each candidate forecasting algorithm generates forecast values from the combined set of time series; ([0023] The trained models which include time series forecasting models are being interpreted as the forecasting algorithms and the task to predict demand is being viewed as the forecast values)
Determine forecasting errors for the plurality of candidate forecasting algorithms based at least on the forecast values and the combined set of time series; ([0038] The validation data here is being interpreted as the forecast values. The data being used here is time series data so that is being interpreted as the combined time series and it is being used to determined error rates.)
Train a machine learning model to determine ranks of forecasting algorithms for forecasting a given time series based on the forecasting errors and the N characteristics; ([0019] They ML model is being trained to find the best performing model for the time series data. The time series data has attributes to describe the data which I am interpreting as characteristics and the error values for the TSF data is being interpreted as forecasting errors. With the goal of finding the model with the least error that is being interpreted as ranking the different algorithms.)
And select one algorithm of the candidate forecasting algorithms to generate predictions for the given time series based on at least the ranks ([0019] Selecting a model here is being interpreted as selecting a model with the least error (best accuracy) as selecting an algorithm for the time series base based on rankings
Huu does not teach One or more non transitory computer -readable media that include stored thereon computer-executable instructions that when executed by at least a processor of a computer case the computer to: create a time series generator by 1) accessing a first set of time series, 2) identifying N characteristics of each time series of the first set of time series, 3) training an auto-encoder with a bottleneck layer of N nodes based on a loss function that minimizes discrepancies between bottleneck layer activations and the N characteristics, and 4) setting the bottleneck layer as an input to the time series generator; determine gaps in an N-dimensional characteristic space for the first set of time series;
However, Westmeier does teach One or more non transitory computer-readable media that include stored thereon computer-executable instructions that when executed by at least a processor of a computer cause the computer to: create a time series generator ([Abstract] teaches generating synthetic time series data)
accessing a first set of time series ([Page 4- Section IV] teaches building an encoder-and-decoder structure with time series data.)
identifying N characteristics of each time series of the first set of time series ([Abstract] teaches Frechet InceptionTime Distance where they are extracting and verifying the synthesized time series data characteristics matches up)
training an auto-encoder with a bottleneck layer of N nodes based on a loss function that minimizes discrepancies between bottleneck layer activations and the N characteristics ([Page 5] teaches The Kullback-Leibner Divergence functions as the loss function with it being used between the actual latent distribution and the multivariant gaussian that is functioning as the bottleneck layer activations and the N characteristics. Figure 5 on page 5 shows a VAE figure which is an autoencoder)
and 4) setting the bottleneck layer as an input to the time series generator ([Page 5] the prior is being used to draw samples for data generation. The prior here is functioning as input to the time series generation)
determine gaps in an N-dimensional characteristic space for the first set of time series ([Page 5] The KL will be able to determine the gaps in between the time series data)
Huu and Westemeier are analogous art because they both relate to time series data
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Huu with the time series synthetic data generation of Westmeier. Doing so would allow for ML models to train properly without having the initial large amounts of needed time series data ([Westmeier -Abstract] “However, machine learning (ML) for process monitoring requires large amounts of training data, especially as the targeted fault states are scarce and yet diverse in their appearances. Therefore, we propose to use synthetic time series data to leverage the high cost of acquiring training data from experiments in real test benches.)
Huu and Westmeier doesn’t teach input one or more characteristics vectors that fill into the bottleneck layer to produce a gap-filling set of time series; combine the first set of time series with the gap-filling set of time series to generate a combined set of time series;
However, Mo does teach input one or more characteristics vectors that fill the gaps into the bottleneck layer to produce a gap-filling set of time series; ([0031] teaches using time series prediction techniques to fill gaps in multiple datasets by generating synthetic metric data. That synthetic metric data is being interpreted as vectors and with it being data at least one characteristic that can describe the data will be present teaching this limitation)
Combine the first set of time series with the gap-filling set of time series to generate a combined set of time series; ([0024] teaches filling gaps between datasets to create a continuous and connected dataset which would teach this limitation)
Huu, Westmeier and Mo are analogous art because they are all deal with time series
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Huu with the time series synthetic data generation of Westmeier, and the gap filling technique of Mo. Doing so would allow for synthetic data to be generation without the high cost needed to set up the environment ([Mo-0023] “[0023] Aspects of the present invention relate generally to generating synthetic metrics data and, more particularly, to generating high-fidelity synthetic metrics data for application performance management (APM) applications. Embodiments of the present invention provide a lightweight process for generating synthetic metrics data on demand. Embodiments of the present invention provide synthetic metrics data without a high-cost of setting up the environment.”)
Claim 2 is rejected under 103 as being unpatentable over Doan Huu et al (US 20250021866 A1) (“Huu”) in view of Mo et al (US 20250156744 A1) (“Mo”), Westmeier et al (NPL: Generating Synthetic Time Series for Machine-Learning-Empowered Monitoring of Electric Motor Test Benches), and Chung et al (US 20240185095 A1) (“Chung”)
Regarding Claim 2, Huu, Mo, and Westmeier teaches all the limitations of Claim 1
Huu does not teach the non-transitory computer-readable medium of claim 1, wherein the bottleneck layer comprises 22 nodes that correspond to 22 characteristics of the time series
However, Chung does teach the non-transitory computer-readable medium of claim 1, wherein the bottleneck layer comprises 22 nodes that correspond to 22 characteristics of the time series ([0050] establishes the relationship between nodes and characteristics and [0055] establishes the relationship between noes and the bottleneck layer. Since no established number was stated, it can be reasonably interpreted that the one or more nodes mentioned could be 22 nodes linked to 22 characteristics which teaches this limitation)
Huu, Westmeier, Mo and Chung are analogous art because they all deal with time series data.
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Huu with the time series synthetic data generation of Westmeier, the gap filling technique of Mo, and the nodes to characteristics pairing of Chung. Doing so would allow for proper forecasting of a product’s life cycle ([Chung-0004] “[0004] Recently, various demand prediction models based on machine learning or data mining have been developed. There is proposed a model for predicting a product life cycle curve using useful covariate information including product features and promotions based on Bayesian Functional Regression, or a prediction model based on deep learning and nonlinear neural network regression. However, most studies cannot be said to have sufficiently solved the challenges of uncertainty and complexity of new products.”)
Claims 3 and 6 are rejected under 103 as being unpatentable over Doan Huu et al (US 20250021866 A1) (“Huu”) in view of Mo et al (US 20250156744 A1) (“Mo”), Westmeier et al (NPL: Generating Synthetic Time Series for Machine-Learning-Empowered Monitoring of Electric Motor Test Benches), and Talagala et al (NPL: Meta-learning how to forecast time series) (“Talagala”)
Regarding Claim 3, Huu, Mo, and Westmeier teaches all the limitations of Claim 1
Huu, does not teach wherein time series in the first set of time series are synthesized to simulate a range of time series behaviors comprising outliers, multiple seasonalities, change-points, intermittency, and high-level effects
However, Talagala does teach wherein time series in the first set of time series are synthesized to simulate a range of time series behaviors comprising outliers, multiple seasonalities, change-points, intermittency, and high-level effects ([Figure 6] Teaches time series simulation over a different season, the outliers, change-points, intermittency, and high-level effects could also be seen in the graphs.)
Huu, Mo, Westmeier and Talagala are analogous art because they all deal with time series data.
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Huu with the time series synthetic data generation of Westmeier, the gap filling technique of Mo, and the forecasting technique of Talagala. Doing so would allow for someone to identify the best forecasting method for their time series ([Talagala-Abstract] “A crucial task in time series forecasting is the identification of the most suitable forecasting method. We present a general framework for forecast-model selection using meta-learning. A random forest is used to identify the best forecasting method using only time series features. The framework is evaluated using time series from the M1 and M3 competitions and is shown to yield accurate forecasts comparable to several benchmarks and other commonly used automated approaches of time series forecasting. A key advantage of our proposed framework is that the time-consuming process of building a classifier is handled in advance of the forecasting task at hand.”)
Regarding Claim 6, Huu, Mo, and Westmeier teaches all the limitations of Claim 1.
Talagala also teaches wherein the candidate forecasting algorithms comprise one or more of an autoregressive integrated moving average (ARIMA) model, an error-trend-seasonality (ETS) model, a deep learning model, an error feedback estimation (EFE) model, and a prophet model ([Table 3] teaches different forecasting algorithm models including an ARIM, and ETS model.
Claim 4 is rejected under 103 as being unpatentable over Doan Huu et al (US 20250021866 A1) (“Huu”) in view of Mo et al (US 20250156744 A1) (“Mo”), Westmeier et al (NPL: Generating Synthetic Time Series for Machine-Learning-Empowered Monitoring of Electric Motor Test Benches), and Chowdhury et al (NPL: Time Series Data Augmentation using Time-Warped Auto-Encoders) (“Chowdhury”)
Regarding Claim 4 Huu, Mo, and Westmeier teaches all the limitations of Claim 1.
Huu does not teach wherein the loss function is configured to minimize discrepancies between the bottleneck layer activations and the characteristics vectors be evaluating a combined error of a) a difference between output and input, and b) a difference between the bottleneck layer and the characteristic features of the first set of time series
However, Chowdhury does teach wherein the loss function is configured to minimize discrepancies between the bottleneck layer activations and the characteristics vectors be evaluating a combined error of a) difference between output and input, and b) a difference between the bottleneck layer and the characteristic features of the first set of time series ([Page 2] Details an auto encoder that takes in time series data that looks at the difference between the input and reconstructed time series which functions as the output. A bottleneck layer is present in the TiGA architecture and it can be reasonably interpretated that the time series data will have characteristic features. A loss function here is being used to assess the difference between the input and reconstructed time series which is the input and output. The input and output here since it taking in reconstructed time series data generated from the TiGA architecture which contains a bottleneck layer which is shown is figure 1 is being interpreted as calculating the input between the bottleneck layer and characteristic features of the first set of time series with the soft-DTW being used to calculate the reconstruction loss)
Huu, Mo, Westmeier and Chowdhury are analogous art because they all deal with time series data.
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Huu with the time series synthetic data generation of Westmeier, the gap filling technique of Mo, and the time series augmentation of Chowdhury. Doing so would allow for higher quality time series samples for the purpose of oversampling ([Chowdhury-Abstract] “It is of high importance for any data oversampling algorithm to produce synthetic data samples that are not only stemming from the same real data distribution, but also have enough variation to provide more learning opportunities to trained models. Early efforts on time series data augmentation either rely on generating new samples by interpolating between two close real data neighbors or using Generative Adversarial Networks. In this paper, we present Time Series Generation based on Auto-encoders (TiGA), a novel algorithm for time series generation using time-warped autoencoders. Our idea is to exploit the lossy transformation of autoencoders for the purpose of generating synthetic time series data samples. To the best of our knowledge, this is the first effort that leverages the latent features generated by autoencoders for the purpose of time series data oversampling. We evaluate our proposed approach on an open-source real-life solar flare prediction dataset. Results show that TiGA produces samples that are both quantitatively and qualitatively superior to current state-of-the-art methods.”)
Claim 5 is rejected under 103 as being unpatentable over Doan Huu et al (US 20250021866 A1) (“Huu”) in view of Mo et al (US 20250156744 A1) (“Mo”), Westmeier et al (NPL: Generating Synthetic Time Series for Machine-Learning-Empowered Monitoring of Electric Motor Test Benches), and Wang et al (NPL: Gap-Filling and Missing Information Recovery for Time Series of MODIS Data Using Deep Learning-Based Methods) (“Wang”)
Regarding Claim 5, Huu, Mo, and Westmeier teaches all the limitations of Claim 1.
Huu does not teach wherein using the bottleneck layer as an input layer allows generation of the combined set of time series from characteristic vectors that are selected to minimize gaps between two nearest points in the N-dimensional characteristics space.
However, Wang does teach wherein using the bottleneck layer as an input layer allows generation of the combined set of time series from characteristic vectors that are selected to minimize gaps between two nearest points in the N-dimensional characteristics space (Page 7 teaches filling the gaps of time series data with figure 3 showcasing the encoder and decoder and the concatenate in channels is being interpretated as the bottleneck layer. The gap filling being done is being interpreted as gap minimization.)
Huu, Mo, Westmeier and Wang are analogous art because they all deal with time series data.
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Huu with the time series synthetic data generation of Westmeier, the gap filling technique of Mo, and the gap filling for missing information of Wang. Doing so would allow for better time series information restoration ([Wang-Abstract] “The developed method could restore time series of images with detailed texture and generally performed better than the other comparative methods, especially with large or overlapped missing areas in time series. Our study provides an available method to gap-fill time series of remote sensing images and highlights the power of the deep learning methods in reconstructing remote sensing images”)
Claims 8-9, 11, 14-15, 18, 19 are rejected under 103 as being unpatentable over Talagala et al (NPL: Meta-learning how to forecast time series) (“Talagala”), in view of Chowdhury et al (NPL: Time Series Data Augmentation using Time-Warped Auto-Encoders) (“Chowdhury”) and Wang et al (NPL: Gap-Filling and Missing Information Recovery for Time Series of MODIS Data Using Deep Learning-Based Methods) (“Wang”)
Regarding Claim 8, Talagala teaches a computing system, comprising: at least one processor connected to at least one memory; a non-transitory computer readable medium including instructions stored thereon that when executed by at least the processor cause the processor to: analyze each time series in a training set of time series to yield a vector of N characteristics for each of the time series; ([Section 3.1] Teaches generating simulated time series data when the sample is deemed too small. The small sample sizes are being interpreted as being analyzed and the produced time series data to compensate is being interpreted as the vector of n characteristics for each time series)
Input the testing set of time series to each of a set of distinct forecasting algorithms, and calculate forecasting error for each algorithm based on performance for each time series ([Table 3] teaches using MASE which takes in forecast errors across all the algorithms)
Train a ranking function to assign a rank to each forecasting algorithm based on a provided vector of N characteristics; ([Table 3] teaches ranking forecasting algorithms. It is being interpreted since they are using time series data that N characteristics are present)
And automatically select one of the forecasting algorithms to monitor an additional time series based on processing N characteristics of the additional time series with the ranking function. ([Page 12] teaches selecting a model with the lowest forecast error measure as that being identified as the best forecasting method which teaches this limitation)
Talagala does not teach train an auto-encoder using a loss function that minimizes error: (1) between bottleneck layer activations in the auto-encoder and the vectors of characteristics for the time series, and (2) between an input layer and an output layer of the auto-encoder;
However, Chowdhury does teach train an auto encoder using a loss function that minimizes error: (1) between bottleneck layer activations in the auto-encoder and the vector of characteristics for the time series and (2) between an input layer and output layer of the auto-encoder (([Page 2-Methedology] Details an auto encoder that takes in time series data that looks at the difference between the input and reconstructed time series which functions as the output. A bottleneck layer is present in the TiGA architecture and it can be reasonably interpretated that the time series data will have characteristic features. A loss function is used to calculate the reconstruction loss after all these steps performed in the TiGA algorithm which contains an input layer and an output layer.)
generate a testing set of time series based at least in part on inputting the new vectors of N characteristics to the bottleneck layer of the trained auto-encoder ([Page 2-Methedology] teaches reconstructing a time series from an auto encoder by inputting the time series into the autoencoder and outputting the reconstructed time series and the TiGA architecture contains a bottleneck layer which would teach this limitation.
Talagala and Chowdhury are analogous art are analogous art because they both deal with time series data.
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Talagala with the time series augmentation of Chowdhury. Doing so would allow for higher quality time series samples for the purpose of oversampling ([Chowdhury-Abstract] “It is of high importance for any data oversampling algorithm to produce synthetic data samples that are not only stemming from the same real data distribution, but also have enough variation to provide more learning opportunities to trained models. Early efforts on time series data augmentation either rely on generating new samples by interpolating between two close real data neighbors or using Generative Adversarial Networks. In this paper, we present Time Series Generation based on Auto-encoders (TiGA), a novel algorithm for time series generation using time-warped autoencoders. Our idea is to exploit the lossy transformation of autoencoders for the purpose of generating synthetic time series data samples. To the best of our knowledge, this is the first effort that leverages the latent features generated by autoencoders for the purpose of time series data oversampling. We evaluate our proposed approach on an open-source real-life solar flare prediction dataset. Results show that TiGA produces samples that are both quantitatively and qualitatively superior to current state-of-the-art methods.”)
Talagala and Chowdhury do not teach generate one or more new vectors of N characteristics by minimizing gaps between neighboring points in an N-dimensional characteristic space;
However, Wang does teach generate one or more new vectors of N characteristics by minimizing gaps between neighboring points in an N-dimensional characteristic space; (Page 7 teaches filling the gaps of time series data. The gap filling being done is being interpreted as gap minimization.)
Talagala, Chowdhury, and Wang are analogous art are analogous art because they both deal with time series data.
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Talagala with the time series augmentation of Chowdhury. the gap filling for missing information technique of Wang. Doing so would allow for better time series information restoration ([Wang-Abstract] “The developed method could restore time series of images with detailed texture and generally performed better than the other comparative methods, especially with large or overlapped missing areas in time series. Our study provides an available method to gap-fill time series of remote sensing images and highlights the power of the deep learning methods in reconstructing remote sensing images”)
Regarding Claim 9, Talagala, Chowdhury, and Wang teaches all the limitations of Claim 8.
Talagala also teaches wherein the instructions further causes the processor to: forecast values of the additional time series with the selected forecasting algorithm; and generate an alert where the forecast values differ from actual values of the additional time series ([Table 3] Generates a table for the different algorithms based off of their MASE values which can function as a loss function. The results of the table as far are the performance of the different algorithms functions as alert.
Regarding Claim 11, Talagala, Chowdhury, and Wang teaches all the limitations of Claim
Talagala also teaches wherein each training time series and testing time series re of an equal length that is between 1000 and 2000 observations ([Page 14] Teaches times series being of equal lengths based on the series it is based on. It can be reasonably interpreted that those lengths can be between 1000 to 2000
Regarding Claim 14 Talagala, Chowdhury, and Wang teaches all the limitations of Claim
Talagala also teaches wherein the forecasting algorithms comprise one or more of an autoregressive integrated moving average (ARIMA) model, an error-trend-seasonality (ETS) model, a deep learning model, an error feedback estimation (EFE) model, and a prophet model ([Table 3] teaches different forecasting algorithm models including an ARIM, and ETS model.
Regarding Claim 15, it is rejected according to the rejection of Claim 8
Regarding Claim 18, it is rejected according to the rejection of Claim 9
Regarding Claim 19 it is rejected according to the rejection of Claim 11
Claim 10 is rejected under 103 as being unpatentable over Talagala et al (NPL: Meta-learning how to forecast time series) (“Talagala”), in view of Chowdhury et al (NPL: Time Series Data Augmentation using Time-Warped Auto-Encoders) (“Chowdhury”) and Wang et al (NPL: Gap-Filling and Missing Information Recovery for Time Series of MODIS Data Using Deep Learning-Based Methods) (“Wang”), and Chung et al (US 20240185095 A1) (“Chung”)
Regarding Claim 10, Talagala, Chowdhury, Wang teaches all the limitations of Claim 8.
Talagala does not teach wherein the bottleneck layer comprises N nodes that correspond to N characteristics in the vector of N characteristics
However, Chung does teach wherein the bottleneck layer comprises N nodes that correspond to N characteristics in the vector of N characteristics ([0050] establishes the relationship between nodes and characteristics which would teach this limitation)
Talagala, Chowdhury, Wang, and Chung are all analogous are because they deal with time series
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Talagala with the time series augmentation of Chowdhury. the gap filling for missing information technique of Wang, and the nodes to characteristics pairing of Chung. Doing so would allow for proper forecasting of a product’s life cycle ([Chung-0004] “[0004] Recently, various demand prediction models based on machine learning or data mining have been developed. There is proposed a model for predicting a product life cycle curve using useful covariate information including product features and promotions based on Bayesian Functional Regression, or a prediction model based on deep learning and nonlinear neural network regression. However, most studies cannot be said to have sufficiently solved the challenges of uncertainty and complexity of new products.”)
Claim 12 and 20 are rejected under 103 as being unpatentable over Talagala et al (NPL: Meta-learning how to forecast time series) (“Talagala”), in view of Chowdhury et al (NPL: Time Series Data Augmentation using Time-Warped Auto-Encoders) (“Chowdhury”), Wang et al (NPL: Gap-Filling and Missing Information Recovery for Time Series of MODIS Data Using Deep Learning-Based Methods) (“Wang”), and Nikain et al (US 20240054124 A1) (“Nikain”)
Regarding Claim 12, Talagala, Chowdhury, Wang teaches all the limitations of Claim 8.
Talagala does not teach wherein the analysis of each time series to yield a vector of N characteristics generates the vector to include a full set of catch-22 characteristics
However, Nikain does teach wherein the analysis of each time series to yield a vector of N characteristics generates the vector to include a full set of catch-22 characteristics ([0043] details time series vectors including catch 22 characteristics)
Talagala, Chowdhury, Wang, and Nikan are all analogous art because they all deal with time series
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Talagala with the time series augmentation of Chowdhury. the gap filling for missing information technique of Wang, and the catch-22 vector characteristics of Nikain. Doing so would allow for consistent data to be maintained ([Nikain-0002] “A major factor in data quality is redundant data that exists in multiple systems with inconsistent values. In a telecommunication network having a software-defined networking (SDN) architecture, different locations in the telecommunication network may be provisioned with a common resource pool of network function virtualization infrastructure (NFVI), and to the extent possible, routers, switches, edge caches, middle-boxes, and the like, may be instantiated and terminated from the common resource pool. However, along with the increasing ease of making infrastructure changes in the telecommunication network, ever increasing volumes of data are collected and stored in relation to network operations, such as inventory records, operational data, and so forth. As such, while mechanisms exists to attempt to maintain consistent data across various systems, opportunities still arise for inconsistencies due to various reasons.”)
Regarding Claim 20, it is rejected according to the rejection of Claim 12
Claim 13 and 16 is rejected under 103 as being unpatentable over Talagala et al (NPL: Meta-learning how to forecast time series) (“Talagala”), in view of Chowdhury et al (NPL: Time Series Data Augmentation using Time-Warped Auto-Encoders) (“Chowdhury”), Wang et al (NPL: Gap-Filling and Missing Information Recovery for Time Series of MODIS Data Using Deep Learning-Based Methods) (“Wang”), and Cerqueira et al (NPL: Model Selection for Time Series Forecasting An Empirical Analysis of Multiple Estimators) (“Cerqueira”)
Regarding Claim 13, Talagala, Chowdhury, and Wang teaches all the limitations of Claim 8.
Talagala also teaches wherein the instructions to automatically select one of the forecasting algorithms further cause the processor to: assign ranks to the forecasting algorithms with the trained ranking function based on a N dimensional vector for the additional time series ([Table 3] Teaches assigning ranks to different time series algorithms based on vector input which can be seen in section 3 of the methodology page 10. The table ranks can function as a way to choose the three top ranked forecasting algorithms.)
Talagala does not teach calculate the respective cross-validation errors of the top ranked algorithms with respect to the additional time series; and select, as the one of the forecasting algorithms, the top ranked algorithm with the least respective cross-validation error
However, Cerqueira does teach calculate the respective cross-validation errors of the top ranked algorithms with respect to the additional time series; and select, as the one of the forecasting algorithms, the top ranked algorithm with the least respective cross-validation error ([Section 4.4] Teaches testing 50 different algorithms and selecting one that maximizes the expected performance using cross validation techniques with the estimators being used for model selection for forecasting stated in the introduction Paragraph 4.
Talagala, Chowdhury, Wang, and Cerqueira are all analogous art because they all deal with time series
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Talagala with the time series augmentation of Chowdhury, the gap filling for missing information technique of Wang, and the cross validation ranking of Cerqueira. Doing so would allow for better rate at selecting the best performing model when it comes to forecasting ([Introduction] “The goal of this work is to study different estimators (e.g. K-fold cross-validation, Holdout) for model selection in time series forecasting tasks, in which the observations are not i.i.d.. Given a pool of alternative models, this work studies: (i) how often the best solution is picked (the one which maximizes forecasting performance on test data); and (ii) how much performance is lost when it does not. In other words, the goal is to analyze how different estimators rank the available predictive models by their performance in unseen observations. This work particularly emphasizes the top-ranked model, which is the one most probably selected for deployment. That is, for predicting future observations of the domain under study.”)
Regarding Claim 16, it is rejected according to the rejection of Claim 13
Claim 17 is rejected under 103 as being unpatentable over Talagala et al (NPL: Meta-learning how to forecast time series) (“Talagala”), in view of Chowdhury et al (NPL: Time Series Data Augmentation using Time-Warped Auto-Encoders) (“Chowdhury”), Wang et al (NPL: Gap-Filling and Missing Information Recovery for Time Series of MODIS Data Using Deep Learning-Based Methods) (“Wang”), and Badinski et al (US 20250067630 A1) (“Badinski”)
Regarding Claim 17, Talagala, Chowdhury, and Wang teaches all the limitations of Claim 15
Talagala does not teach wherein the minimizing gaps between neighboring points in an N-dimensional characteristics space is performed based on t-distributed stochastic neighbor embedding
However, Badinski does teach wherein the minimizing gaps between neighboring points in an N-dimensional characteristics space is performed based on t-distributed stochastic neighbor embedding ([0209-0210] teaches using t-distributed stochastic neighbor for dimension reduction with the method later being used to fill gaps in a data base.)
Talagala, Chowdhury, Wang, and Badinski are all analogous art because they all deal with time series
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Talagala with the time series augmentation of Chowdhury, the gap filling for missing information technique of Wang, and the gap filling of Badinski. Doing so would allow for better monitoring of the time series forecasting ([0002] “In batch processing or batch plants, the production of multiple products occurs with the same set of equipment or processing units, for example a chemical or biological reactor. Monitoring the production processes is important to optimize the production processes, recognize abnormalities in the production processes and the like. However, due to the variability of products produced with the plant, it can be difficult to monitor the production process.”)
Allowable Subject Matter
Claim 7 is allowed
Conclusion
The Prior arts are made of recorded and relied upon is considered to applicant’s disclosure
Michael Langford (20250139503) (2023-10-31) ([Abstract] “Methods and systems are described herein for minimizing development time in artificial intelligence models by automating model selection based on dataset fittings of time-series data prior to hyperparameter optimization. The systems and methods described herein aim to reduce the redundancies and improve the efficiencies of model selection, model training, and/or hyperparameter selection. The systems and methods achieve this by using information about the attributes of the time-series dataset that may be used to determine a model that may be most effective at fitting a given dataset. If a model is selected prior to hyperparameter optimization, the time and resources spent training, fitting, and/or tuning models that are not selected can be avoided.”)
Mukherjee et al (20250139494) (2023-11-01) ([Abstract] “A computer-implemented method for forecasting a future value of one or more elements of a time-series of data includes obtaining a time-series of data, obtaining a library having a plurality of selected loss functions, obtaining at least one Business Specification Rule (BSR), each BSR including a Context, a Metric and a Priority, for each selected loss function, generating input-associated perturbated outputs based on the BSRs and the time-series of data by training a deep learning artificial intelligence (DLAI) model to learn a set of learned weights to be given to each of the selected loss functions, deriving a custom composite loss function based on the sets of learned weights for the plurality of selected loss functions in the LFL, and using the custom composite loss function to train a final DLAI model on the time-series of data. The final DLAI model may then be used to forecast future outcomes.”
Jorn Brouwers (20250044271) (2022-12-12) ([Abstract] “A computer-implemented method of predicting quality of a food product sample after a mixing process, based on properties of the food product, comprises: building a hybrid model by: training an autoencoder in an unsupervised learning step using historical process data of food product samples; training a supervised model in a supervised learning step using the output of the autoencoder; and predicting the quality of the food product by inputting process data of current samples into the hybrid model and classifying the samples.”
Ramadas et al (20250037734) (2023-07-28) ([Abstract] “A device includes a memory configured to store one or more segments of time-series data. The device also includes one or more processors configured to generate, using a feature extractor, a latent-space representation of a segment of the time-series data. The one or more processors are also configured to provide one or more inputs to a classifier, the one or more inputs including at least one input based on the latent-space representation. The one or more processors are also configured to generate, based on output of the classifier, a processing control signal for the segment.”
Dang et al (20240220858) (2023-05-10) ([Abstract] “A prediction system may obtain data, via a network, from devices and process the data, using a first machine learning, to identify a plurality of signals. The prediction system may train a second machine learning model to analyze the plurality of signals to forecast a first forecasted time series and evaluate a first performance of the first forecasted time series. The prediction system may determine that the first performance does not satisfy a performance threshold and may refine the plurality of signals to obtain a refined plurality of signals. The prediction system may train a third machine learning model to analyze the refined plurality of signals to forecast a second forecasted time series and evaluate a second performance of the second forecasted time series. The prediction system may use the refined plurality of signals and the third machine learning model to predict a performance of a third forecasted time series.”
Yanchenko et al (20240211801) (2022-12-27) ([Abstract] “Mechanisms are provided for automatic identification of a reconciliation computer tool for producing coherent reconciled data from base data generated by a computer model. A machine learning training operation is executed on one or more performance prediction computer model(s) (PPCMs) based on first input features of at least one hierarchical dataset, and second input features of a plurality of different reconciliation computer tools. The PPCM(s) generate a prediction of performance of a corresponding reconciliation computer tool based on the first and second input features. Features are extracted from a runtime hierarchical dataset and input into the trained PPCM(s) which generate predictions of performance of a plurality of reconciliation computer tools based on the extracted features of the runtime hierarchical dataset. The reconciliation computer tools are ranked relative to one another based on the predictions of performance. An output is generated based on the ranking of the reconciliation computer tools.”
Mohammed Islam (20240130621) (2023-11-02) ([Abstract] “A measurement system may comprise an actively illuminated camera system, in some embodiments coupled to a time-of-flight sensor or an array of laser diodes beam split into a plurality of spatially separated lights. The camera system may capture two or three dimensional images, and the light source may comprise semiconductor diodes, such as light emitting diodes. The system includes a processor coupled to non-transitory computer readable medium and configured to use artificial intelligence to make one or more decisions. The processing may also involve artificial intelligence or machine learning techniques to analyze anomalous occurrences, or generative artificial intelligence to interface with a user or improve the performance of camera-based systems. Algorithms may also be used to improve the performance of generative artificial intelligence processing. The camera output may be fused with data from other sensors, and the camera may also capture information about the pose or gestures of a user.
Sari et al (20230186053) (2021-12-09) ([Abstract] “A device includes one or more processors configured to process a portion of time-series data using a trained encoder network to generate a dimensionally reduced encoding of the portion of the time-series data. The one or more processors are further configured to process the dimensionally reduced encoding using a trained decoder network to determine decoder output data. The one or more processors are also configured to set parameters of a predictive machine-learning model based on the decoder output data, wherein the predictive machine-learning model is configured to, based on the parameters, determine a predicted future value of the time-series data.”
Chen et al (20220261598) (2021-10-26) ([Abstract] “To rank time series forecasting in machine learning pipelines, time series data may be incrementally allocated from a time series data set for testing by candidate machine learning pipelines based on seasonality or a degree of temporal dependence of the time series data. Intermediate evaluation scores may be provided by each of the candidate machine learning pipelines following each time series data allocation. One or more machine learning pipelines may be automatically selected from a ranked list of the one or more candidate machine learning pipelines based on a projected learning curve generated from the intermediate evaluation scores.”)
Any inquiry concerning this communication or earlier communications from the examiner should be directed to URIAH V MOORE whose telephone number is (571)384-8341. The examiner can normally be reached Monday-Friday 8am-5pm.
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, Mariela Reyes can be reached at (571)270-1006. 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.
/Mariela Reyes/ Supervisory Patent Examiner, Art Unit 2142
Prosecution Insights