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 .
Response to Arguments
Applicant’s arguments, see page, filed 01/12/2026, with respect to the rejection(s) of claim(s) 1-15 under 35 U.S.C. § 103 have been considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made over Taheri et al. (US 20230324860 A1) in view of Dempster et al. (ROCKET: exceptionally fast and accurate time series classification using random convolutional kernels, 13 July 2020), and Rao et al. (Spatio-Temporal Functional Neural Networks, 20 November 2020).
Applicant's arguments filed 01/12/2026 page 7 paragraph 7, have been fully considered but they are not persuasive. As stated by the applicant specification paragraph [0040] "random convolutional kernels (i.e., kernels with random length, weights, bias, dilation, and padding). ... In the second step, the transformed features are used to train a linear classifier.", the applicant.
Applicant's arguments filed 1/12/2026 page 8 paragraph 2, have been fully considered but they are not persuasive. the rejection does not require that the cited references expressly recognize applicant's problem or recite the claimed feature verbatim. .
Applicant's arguments filed 1/12/2026 page 8 paragraph 3, have been fully considered but they are not persuasive. The cited reference is not unrelated or improperly combined. Dempster et al. .
In response to applicant’s argument Page 8 Paragraph 4 that there is no teaching, suggestion, or motivation to combine the references, the examiner recognizes that obviousness may be established by combining or modifying the teachings of the prior art to produce the claimed invention where there is some teaching, suggestion, or motivation to do so found either in the references themselves or in the knowledge generally available to one of ordinary skill in the art. See In re Fine, 837 F.2d 1071, 5 USPQ2d 1596 (Fed. Cir. 1988), In re Jones, 958 F.2d 347, 21 USPQ2d 1941 (Fed. Cir. 1992), and KSR International Co. v. Teleflex, Inc., 550 U.S. 398, 82 USPQ2d 1385 (2007). In this case, see Dempster et al. [5 Conclusion] "Rocket, achieves state-of-the-art accuracy with a fraction of the computational expense of existing state-of-the-art methods, and can scale to millions of time series. We also show that Rocket is significantly more accurate and, with one exception, fundamentally more scalable than several recently-proposed scalable methods for time series classification.".
Claim Rejections - 35 USC § 103
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(s) 1-17, and 19-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Taheri et al. (US 20230324860 A1) in view of Dempster et al. (ROCKET: exceptionally fast and accurate time series classification using random convolutional kernels, 13 July 2020), and Rao et al. (Spatio-Temporal Functional Neural Networks, 20 November 2020).
As to claim 1, Taheri teaches A method, comprising:
for receipt of time-series data indicative of energy consumption associated with a type of building of a plurality of different types of buildings and a climatic zone from a plurality of climatic zones: (see Taheri paragraph [0049], “receives a time-series input 620 and generates one or more time-series data points 650 representing an estimated energy consumption for the building”, and paragraph [0066] “The building types covered by the tests include office, hotel, leisure centre, farm, manufacturing and food producing facilities, located in the United Kingdom and the United States of America”, and paragraph [0042], “receive time-series data for the building comprising weather data…weather data may be retrieved (e.g., from a third-party source) using a network query using the location of the building as a parameter (e.g., the post or zip code of the building)”):
executing random convolutional kernel (RCK) on the time-series data to generate a classification group of the time-series data according to the type of building and the climatic zone; and ( see Dempster introduction [0004] “In contrast to learned convolutional kernels as used in typical convolutional neural networks, we show that it is effective to generate a large number of random convolutional kernels which, in combination, capture features relevant for time series classification—even though, in isolation, a single random convolutional kernel may only very approximately capture a relevant feature in a given time series.”, See Dempster Method [0001] “Rocket transforms time series using a large number of random convolutional kernels, i.e., kernels with random length, weights, bias, dilation, and padding. The transformed features are used to train a linear classifier.” Even though Dempster doesn’t teach classification by building type and climatic zone, a person of ordinary skill in the art would have the knowledge to apply it to perform the classification by building type and climatic zones.
executing a trained functional neural network (FNN) comprising a plurality of continuous layers having continuous neurons configured to map time-series data derived functions to a short-term energy consumption forecast using integral operations over bivariate parameter functions on the time-series data of the classification group to provide a short-term energy consumption forecast. (see Taheri paragraph [0039], “energy estimation system 410 receives time-series data 420 for a building as input. The building may be the building 300 of FIG. 3. The time-series data 420 may be received sample-by-sample over time, in mini-batches, or in one large batch. The time-series data 420 comprises a set of time samples that represent measurements associated with the building at successive points in time”, and paragraph [0040], “the energy estimation system 410 employs a dilated convolutional neural network architecture to generate estimated energy consumption values for the building. In FIG. 4, the energy estimation system 410 outputs one or more time-series data points 430 representing the estimated energy consumption.”)
Taheri disclosed receiving and processing time-series data indicative of building energy consumption from multiple building types and climatic zones. Taheri does not explicitly teach "executing random convolutional kernel (RCK) on the time-series data to generate a classification group of the time-series data according to the type of building and the climatic zone", and " functional neural network (FNN) comprising a plurality of continuous layers having continuous neurons configured to map time-series data derived functions"
However, Dempster teaches executing a random convolutional kernel (RCK) on the time-series data to generate a classification group of the time-series data (see section 3, pages 1460-1461, "Rocket transforms time series using a large number of random convolutional kernels, i.e., kernels with random length, weights, bias, dilation, and padding. The transformed features are used to train a linear
classifier.")
Also, Rao teaches a functional neural network that is comprising of multiple layer (see Rao [II. PRELIMINARIES] "The fundamental idea of FNN is to embed the FLM in Eq (4) into the fully connected neural network structure in deep learning [15]. In particular, the architecture of FNN is described as follows. The first layer of FNN consists of novel functional neurons.", and see Rao [II. PRELIMINARIES] "Fig. 2",
PNG
media_image1.png
318
344
media_image1.png
Greyscale
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the invention of Taheri to replace convolutional neural network with a Functional neural network instead of convolutional neural network to achieve accurate prediction (see section II. PRELIMINARIES of Rao) "FNNs have demonstrated improved performances over the alternative methods (e.g., RNN, LSTM) in temporal to scalar predictive modeling tasks..." Thus, the combination Taheri and Rao teaches "functional neural network (FNN) comprising a plurality of continuous layers having continuous neurons configured to map time-series data".
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the invention of Taheri to executing a random convolutional kernel (RCK) on the time-series data to generate a classification group of the time-series data because random kernels have very low computational requirements, making learning and classification extremely fast and achieves state-of-the-art accuracy with a fraction of the computational expense of existing state-of-the-art methods (see Section 5 Conclusion of Dempster). Thus, the combination of Taheri-Rao and Dempster teaches "executing random convolutional kernel (RCK) on the time-series data to generate a classification group of the time-series data according to the type of building and the climatic zone".
As to claim 2, Taheri as modified by Rao teaches wherein the FNN comprises a plurality of continuous layers trained to map time-series data derived functions related to the different types of buildings and the plurality of climatic zones to a short-term energy consumption forecast model configured to provide the short-term energy consumption forecast. (see Taheri paragraph [0063], “In the examples described above, the dilated convolutional neural network architecture comprises a plurality of parameters, in the form of the weights and/or biases of the neural network layers 622, 624, 626, 642 and 648. … The dilated convolutional neural network architecture may be trained by dividing a set of time series data covering an available period into a plurality of windowed data segments split into input and output sections.”, and paragraph [0064] “it may be preferred to train the dilated convolutional neural network architecture such that different buildings have different parameter sets, i.e. to train one set of parameters for the building using only data relating to that building. For example, test data indicated that the response of each building to weather and operational data varied considerably among groups of buildings”.
Taheri does not explicitly teach "the FNN comprises a plurality of continuous layers."
However, Rao teaches the FNN comprises a plurality of continuous layers (see Rao [II. PRELIMINARIES] "The fundamental idea of FNN is to embed the FLM in Eq (4) into the fully connected neural network structure in deep learning [15]. In particular, the architecture of FNN is described as follows. The first layer of FNN consists of novel functional neurons. … are supplied into subsequent layers of numerical neurons (e.g., the inputs and outputs are both scalar values) for further manipulations till the output layer that holds the response variable. … ")
As to claim 3, Taheri-Rao as modified by Dempster teaches wherein the RCK is configured to
generate the classification group according to the type of building and the climatic zone from a database of class labels used to generate different classes based on class labels (see Dempster Classifier [0001] “The transformed features are used to train a linear classifier. Rocket can, in principle, be used with any classifier. We have found that Rocket is very effective when used in conjunction with linear classifiers, which have the capacity to make use of a small amount of information from each of a large number of features.”, wherein the FNN is trained for each of the class labels (see Taheri paragraph [0064] “convolutional neural network architecture may be trained individually for each building”)
As to claim 4, Taheri teaches wherein the short-term energy consumption forecast is based on a selected time window from a plurality of time windows (see Taheri paragraph [0063], “time-series data for a number of years may be split into a plurality of windows (e.g. by setting an advancing temporal stride between each window), where predictions are then made in regard to a day of output data.”)
As to claim 5, Taheri as modified by Rao teaches wherein the FNN is trained across the plurality of time windows (See Taheri paragraph [0063], “The dilated convolutional neural network architecture may be trained by dividing a set of time series data covering an available period into a plurality of windowed data segments split into input and output sections. For example, time-series data for a number of years may be split into a plurality of windows (e.g. by setting an advancing temporal stride between each window), where predictions are then made in regard to a day of output data.”,
Taheri does not explicitly teach "FFN"
However, Rao teaches FNN "The architecture of a FNN with three functional neurons on the first layer and two numerical neurons on the second layer."
As to claim 6, Taheri and Rao teaches wherein the time-series data and the short-term energy consumption forecast are represented as periodic functions (See paragraph [0041], “yet still be able to capture different temporal patterns within the data, e.g. following a training operation. In one case, the number of convolution layers within the dilated convolutional neural network architecture may be set based on the sampling frequency of the input time-series data 420 for a particular building.”, see figure [7] also represents the periodic functions of the received time series data and prediction data. See paragraph [0066] “As may be seen in FIG. 7, each building has a different energy consumption profile. Each chart shows a first portion 702, 712, 722 of the time-series data that was used for training and a second portion 704, 714, 724 of the time-series data that was predicted by the trained energy estimation system. The original measured data 706, 716, 726 for the time period is also shown. These charts show that the trained energy estimation system accurately learns the complex temporal patterns to produce accurate energy consumption estimates, including the capture of both daily and seasonal patterns in energy consumption.”, and see Rao [II. PRELIMINARIES] "FNN can be trained with the gradient descent approach under certain assumptions [15], [16]. The forward propagation step can go through as follows. First, in the functional neurons, the integral … is approximated by the numerical integration techniques. The output value … can then calculated by the formula in Eq (5). The forward propagation calculation in subsequent numerical layers is straightforward. In the backward propagation step, the partial derivatives from the output layer up to the second hidden layer (i.e., the numerical layer after the functional neuron layer) can be easily calculated as in the classic neural networks. Whereas, it is essential to ensure that the partial derivatives of the values at the second layer … can be estimated using numerical approximations of the following quantity …")
PNG
media_image2.png
371
531
media_image2.png
Greyscale
As to claim 7, Taheri teaches wherein the time-series data comprises one or more of temperature time-series data, humidity time-series data, precipitation time-series data, or vehicle count time-series data (See Taheri paragraph [0042], “the dilated convolutional neural network architecture is configured to receive time-series data for the building comprising weather data. The weather data may comprise one or more of temperature (e.g., external temperature), humidity, wind properties (e.g., wind speed and/or direction) and solar irradiance”).
As to claim 8, this is directed to a computer-program embodiment that corresponds to method claim 1. See the rejection for claim 1 above, which also applies to claim 8. In addition, Taheri teaches a system having a memory, computer readable instructions, and one or more processors to perform the operations (see Taheri paragraph [0097], “In certain examples, a non-transitory computer-readable storage medium may be provided storing instructions that, when executed by a processor, cause the processor to perform a series of operations”
As to claim 9, this is directed to a computer-program embodiment that corresponds to method claim 2. See the rejection for claim 2 above, which also applies to claim 9.
As to claim 10, this is directed to a computer-program embodiment that corresponds to method claim 3. See the rejection for claim 3 above, which also applies to claim 10.
As to claim 11, this is directed to a computer-program embodiment that corresponds to method claim 4. See the rejection for claim 4 above, which also applies to claim 11.
As to claim 12, this is directed to a computer-program embodiment that corresponds to method claim 5. See the rejection for claim 5 above, which also applies to claim 12.
As to claim 13, this is directed to a computer-program embodiment that corresponds to method claim 6. See the rejection for claim 6 above, which also applies to claim 13.
As to claim 14, this is directed to a computer-program embodiment that corresponds to method claim 7. See the rejection for claim 7 above, which also applies to claim 14.
As to claim 15, this is directed to a system or a computing device claim that corresponds to method claim 1. See the rejection for claim 1 above, which also applies to claim 15. In addition, Taheri teaches a system having a memory, computer readable instructions, and one or more processors to perform the operations (see paragraph [0097], “In certain examples, a non-transitory computer-readable storage medium may be provided storing instructions that, when executed by a processor, cause the processor to perform a series of operations”
As to claim 16, Rao teaches wherein the FNN is configured to perform forward propagation and backward propagation to update the bivariate parameter functions until a stopping criterion is reached. (see Rao [II. PRELIMINARIES] "FNN can be trained with the gradient descent approach under certain assumptions [15], [16]. The forward propagation step can go through as follows. First, in the functional neurons, the integral … is approximated by the numerical integration techniques. The output value … can then calculated by the formula in Eq (5). The forward propagation calculation in subsequent numerical layers is straightforward. In the backward propagation step, the partial derivatives from the output layer up to the second hidden layer (i.e., the numerical layer after the functional neuron layer) can be easily calculated as in the classic neural networks. Whereas, it is essential to ensure that the partial derivatives of the values at the second layer … can be estimated using numerical approximations of the following quantity …")
As to claim 17, Dempster teaches wherein the RCK transforms the time-series data using kernels having randomly selected lengths from a set of predetermined values, random weights, random bias, random dilation, and random padding. (see [3 Method] "Rocket transforms time series using a large number of random convolutional kernels, i.e., kernels with random length, weights, bias, dilation, and padding.")
As to claim 19, Taheri as modified by Rao and Dempster teaches wherein the FNN is configured to incorporate temperature time-series data as a functional feature to improve the short-term energy consumption forecast. (see Rao [II. PRELIMINARIES] "FNNs have demonstrated improved performances over the alternative methods (e.g., RNN, LSTM) in temporal to scalar predictive modeling tasks… The fundamental idea of FNN is to embed the FLM in Eq (4) into the fully connected neural network structure in deep learning [15]. In particular, the architecture of FNN is described as follows. The first layer of FNN consists of novel functional neurons. The functional neurons take the functional covariates…")
Note: Functional Neural Network is known to receive time series data as a function to improve the prediction/forecasting.
As to claim 20, Taheri as modified by Rao and Dempster teaches the method of claim 1, wherein the executing the trained FNN comprises selecting a specialized FNN model corresponding to the classification group generated by the RCK, and supplying the time-series data to the selected specialized FNN model to obtain the short-term energy consumption forecast. (See Taheri paragraph [0064] "each building may have a separate set of parameters and a separate model. In other cases, a common set of parameters may be used, either for all buildings or for buildings of a common type and within a given group of buildings with shared building properties (e.g., office buildings of a given size).In tests, it was found that energy consumption patterns often change considerably between buildings, even between buildings of a common type. If a set of buildings demonstrate different energy consumption patterns within measured data, it may be preferred to train the dilated convolutional neural network architecture such that different buildings have different parameter sets, i.e. to train one set of parameters for the building using only data relating to that building."
Claim(s) 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Taheri et al. (US 20230324860 A1) in view of Dempster et al. (ROCKET: exceptionally fast and accurate time series classification using random convolutional kernels, 13 July 2020), and Rao et al. (Spatio-Temporal Functional Neural Networks, 20 November 2020), and Jain et al. (Integrated Approach for Short Term Load Forecasting using SVM and ANN, March 2009).
As to claim 18, Taheri as modified by Rao and Dempster teaches the energy forecast method as claimed in claim 1.
Taheri, Rao, and Dempster do not teach "wherein the short-term energy consumption forecast is configured to enable pre-incident planning of optimal power shut-off during a natural disaster."
However, Jain teaches
wherein the short-term energy consumption forecast is configured to enable pre-incident planning of optimal power shut-off during a natural disaster. (See Jain [1 Introduction] "The system operators use the load forecasting result as a basis of off-line network analysis to determine if the system might be vulnerable. If so, corrective actions should be prepared, such as load shedding, power purchases and bringing peaking units on line."
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the invention of Taheri-Rao-Dempster to use the results of the energy forecast method to determine pre-incident contingency plan like load shedding. Thus, the combination of Taheri-Rao-Dempster and Jain teaches "wherein the short-term energy consumption forecast is configured to enable pre-incident planning of optimal power shut-off during a natural disaster".
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 ABDULLAH K ABOUD whose telephone number is (571)272-0025. The examiner can normally be reached Mon-Fri 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, Li B Zhen, can be reached at (571) 272-3768. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/ABDULLAH KHALED ABOUD/ Examiner, Art Unit 2121
/Li B. Zhen/ Supervisory Patent Examiner, Art Unit 2121
Prosecution Insights