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 filed 03/10/2026 have been fully considered but they are not fully persuasive.
Regarding the 101 rejections, applicant’s arguments and amendments to the independent claims are persuasive and overcome the previous 101 rejections. Specifically, applicant’s amended limitations of determining future points of the time-series utilizing a forecasting deep neural network (DNN) to analyze the time-series over a forecast horizon comprising a plurality of forecasted time points, wherein the forecasting DNN creates separate but related forecasting functions for respective ones of the forecasted time points; determining an uncertainty of the future points utilizing an uncertainty DNN to analyze the time-series and the future points, wherein the uncertainty DNN receives as input (i) the historical or current observation and (ii) the future points, and wherein the uncertainty DNN outputs an estimate of a residual error associated with the future points as aleatoric uncertainty over the forecast horizon; and providing the future points of the time-series and the uncertainty to a decision module, and, based on the future points and the uncertainty, generating or causing performance of at least one predetermined network action in the communications network that includes any of provisioning network resources, deploying additional equipment, or proactively rerouting connections provides a technical improvement because the aleatoric uncertainty of forecasted predictions are used to update a communications network. See pg. 11 of “Remarks”: “As amended, Claims 1, 10, and 19 require receiving network performance monitoring (PM) time-series data and providing forecast and aleatoric uncertainty outputs to a decision module that causes performance of predetermined network actions. These actions include provisioning network resources, deploying equipment, proactively rerouting connections, and similar operational changes in a telecommunications network. See Spec. [0002], [0035]-[0037], [0046]-[0048]. The claims therefore do not merely "display" or "output" information. Rather, they use the forecast and uncertainty in a closed-loop forecasting system to enact concrete changes to the operation of a technical system (the network). See Spec. [0047]-[0048]. Under the Memo and the ARP Decision, this constitutes integration into a practical application because the computed results are used to control and improve operation of a real-world technical system. The focus of the claims, when considered as a whole, is not an abstract mathematical concept, but a network control workflow driven by residual-based aleatoric uncertainty.” Applicant’s amendments and corresponding arguments that the claimed invention provides a technical improvement to the field of communications networks are persuasive. Therefore, the 101 rejections are withdrawn.
Regarding the 103 rejections, applicant's arguments filed with respect to the prior art rejections have been fully considered but they are moot. Applicant has amended the claims to recite new combinations of limitations. Applicant's arguments are directed at the amendment. Please see below for new grounds of rejection, necessitated by Amendment.
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.
Claims 1-4, 6, 8-13, 15, and 17-20 are rejected under 35 U.S.C. 103 as being unpatentable over Alhussein, et al., US Pre-Grant Publication US20230216811A1 (“Alhussein”) in view of Amiri, et al., US Pre-Grant Publication US20210150305A1 (“Amiri”) and further in view of Kawashima, et al., Non-Patent Literature “The Aleatoric Uncertainty Estimation Using a Separate Formulation with Virtual Residuals” (“Kawashima”).
Regarding claim 1, Alhussein discloses:
A method comprising steps of: receiving a time-series that includes a historical or current observation from a communications network; (Alhussein, ⁋2-3, “The present disclosure pertains to the field of network traffic engineering and in particular to a method and apparatus for managing network traffic in a communication network [from a communications network;]. Management and guidance of network traffic is needed as network traffic varies with time [receiving a time-series]”, and Alhussein, ⁋4, “A traffic forecasting module predicts the (future) demand or network flow in the network based on network flow statistics collected from the network [that includes a historical or current observation].”).
determining future points of the time-series utilizing a forecasting deep neural network (DNN) to analyze the time-series over a forecast horizon comprising a plurality of forecasted points, (Alhussein, ⁋4, “A traffic forecasting module predicts the (future) demand or network flow in the network based on network flow statistics collected from the network [determining future points of the time-series utilizing a forecasting deep neural network (DNN) to analyze the time-series].”, and Alhussein, ⁋33, “Bayesian/probabilistic traffic forecasting models can be generated out of Bayesian neural networks [forecasting deep neural network (DNN)]”, and Alhussein, ⁋9, “the forecasting and planning horizon indicative of a time period for which the traffic forecasting model is applicable [over a forecast horizon comprising a plurality of forecasted points,].”).
determining an uncertainty of the future points utilizing an uncertainty DNN to analyze the time-series and the future points, wherein the uncertainty DNN receives as input (i) the historical or current observation and (ii) the future points, (Alhussein, ⁋33, “for managing and guiding network traffic in a network based on network traffic demand uncertainty [determining an uncertainty of the future points… and (ii) the future points,]. The traffic demand uncertainty may be extracted from the Bayesian/probabilistic traffic forecasting models or other suitable network traffic forecasting model. Bayesian/probabilistic traffic forecasting models can be generated out of Bayesian neural networks [utilizing an uncertainty DNN to analyze the time-series and the future points,]”, and Alhussein, ⁋54, “According to embodiments, at step 201, the inference block 210 (for example the Bayesian neural network or other suitable model) is continuously or at predetermined intervals or intermittently provided with one or more network statistics from the changing database 220. These network statistics can be provided as discrete time epochs, which can be defined as a period of time in past history of network operation that may be identified by one or more a notable events or one or more particular characteristics [wherein the uncertainty DNN receives as input (i) the historical or current observation]”).
and providing the future points of the time-series and the uncertainty to a decision module, and, based on the future points and the uncertainty, generating or causing performance of at least one predetermined network action in the communications network that includes any of provisioning network resources, deploying additional equipment, or proactively rerouting connections. (Alhussein, ⁋35, “According to embodiments, among two different types of the uncertainties for network traffic demand [providing the future points of the time-series], epistemic uncertainty (or model uncertainty) and aleatoric uncertainty (e.g. data uncertainty or inherent noise or random uncertainty) are utilized to guide TE frameworks [and the uncertainty]”, and Alhussein, ⁋56, “The controller node 230 [to a decision module,] is configured to re-configure or re-design the network topology, and set or assign the network resource reservations for the existing network services and requests. The controller node 230, at step 203, sets resource reservations based on a function of the extracted values (i.e. the predictive mean (μ), the epistemic uncertainty (model uncertainty) (σm 2), and the inherent noise or aleatoric uncertainty (σn 2)) [and, based on the future points and the uncertainty, generating or causing performance of at least one predetermined network action in the communications network that includes any of provisioning network resources, deploying additional equipment, or proactively rerouting connections.].”).
While Alhussein teaches a system for determining uncertainty for forecasted time series in a communications network, Alhussein does not explicitly teach:
wherein the forecasting DNN creates separate but related forecasting functions for respective ones of the forecasted time points;
and wherein the uncertainty DNN outputs an estimate of a residual error associated with the future points as aleatoric uncertainty over the forecast horizon;
Amiri teaches wherein the forecasting DNN creates separate but related forecasting functions for respective ones of the forecasted time points; (Amiri, ⁋50, “The neural network architecture of the forecasting module 24 creates separate but related forecasting functions for each forecasted time point [wherein the forecasting DNN creates separate but related forecasting functions for respective ones of the forecasted time points;]”).
Alhussein and Amiri are both in the same field of endeavor (i.e. forecasting). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Alhussein and Amiri to teach the above limitation(s). The motivation for doing so is that using separate forecasting functions for forecasted time points improves a forecasting model’s robustness (cf. Amiri, ⁋24, “However, the capacity of the DNN of the present disclosure can be increased by enabling it to devise a “separate” time waveform for each forecasted data point instead of providing a “common” time waveform for all the data points as is done in conventional systems. The forecasting model of the present disclosure is able to improve prediction performance, even on a dataset having complicated or only partially available periodic patterns.”).
While Alhussein in view of Amiri teaches a system for determining aleatoric uncertainty for forecasted time series in a communications network over a forecasting horizon, the combination does not explicitly teach:
and wherein the uncertainty DNN outputs an estimate of a residual error associated with the future points as aleatoric uncertainty over the forecast horizon;
Brando teaches and wherein the uncertainty DNN outputs an estimate of a residual error associated with the future points as aleatoric uncertainty over the forecast horizon; (Kawashima, pg. 3 col. 1, “A. Separate Formulation of Aleatoric Uncertainty In this section, we present an optimization framework for uncertainty estimation in regression problems [over the forecast horizon;]. We focus on training a network that predicts the signal value yi and its certainty wi [and wherein the uncertainty DNN]… Here, Lt(yi) is the loss for the target estimation, and Lu(wi) is that for the uncertainty estimation. By using this formulation, we can balance the target and uncertainty estimation terms by adjusting λ appropriately. We define Lt(yi) = ri(x;θ), and Lu(wi) = exp(wi(x;θ))˜ri(x;θ) − wi(x;θ), resulting in the loss function L(x,θ) = N 1 N {ri(x;θ) + λ{exp(wi(x;θ))˜ri(x;θ) − wi(x;θ)}}.”, and Kawashima, pg. 2 col. 2, “Eq. 2 can be re-written as Eq. 3 below, and the network outputs the signal estimation y and the certainty estimation w: L(x,θ) = 1 N N i=1 {exp(wi(x;θ))ri(x;θ) − wi(x;θ)}. (3) In this equation, we introduce two variables: ri is the absolute value of residual [outputs an estimate of a residual error associated with the future points]”).
Alhussein, in view of Amiri, and Kawashima are both in the same field of endeavor (i.e. regression learning). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Alhussein, in view of Amiri, and Kawashima to teach the above limitation(s). The motivation for doing so is that training with residuals reduced overfitting (cf. Kawashima, pg. 8 col. 1, “In this study, we presented a framework for error quantification based on heteroscedastic aleatoric uncertainty estimation in regression problems of deep learning. Existing methods cannot estimate the error accurately because of overfitting in the target estimation. We mitigated the effect of overfitting with virtual residuals and avoided underestimating the estimation errors.”).
Regarding claim 2, Alhussein in view of Amiri and Kawashima teaches the method of claim 1. Alhussein further teaches wherein the steps further include utilizing the uncertainty DNN to analyze the time-series and the future points. (Alhussein, ⁋33, “for managing and guiding network traffic in a network based on network traffic demand uncertainty [to analyze the time-series and the future points.]. The traffic demand uncertainty may be extracted from the Bayesian/probabilistic traffic forecasting models or other suitable network traffic forecasting model. Bayesian/probabilistic traffic forecasting models can be generated out of Bayesian neural networks [wherein the steps further include utilizing the uncertainty DNN]”).
Regarding claim 3, Alhussein in view of Amiri and Kawashima teaches the method of claim 1. Alhussein further teaches wherein the steps further include performing the determining steps concurrently. (Alhussein, ⁋33, for managing and guiding network traffic in a network based on network traffic demand uncertainty. The traffic demand uncertainty may be extracted from the Bayesian/probabilistic traffic forecasting models or other suitable network traffic forecasting model. Bayesian/probabilistic traffic forecasting models can be generated out of Bayesian neural networks; the Bayesian forecasting model is interpreted as being the forecasting and uncertainty neural network (i.e. wherein the steps further include performing the determining steps concurrently.)”).
Regarding claim 4, Alhussein in view of Amiri and Kawashima teaches the method of claim 1. Alhussein further teaches wherein the forecasting DNN and the uncertainty DNN include various components including any of dense layers, long short-term memory (LSTM) layers, pooling layers, and convolutional layers. (Alhussein, ⁋97, “The vector computation unit 607 may be mainly used for computation at a non-convolutional layer or fully-connected layers (FC, fully connected layers) of a neural network, for example, pooling (pooling), batch normalization (batch normalization), or local response normalization (local response normalization) [various components including any of dense layers, long short-term memory (LSTM) layers, pooling layers, and convolutional layers.].”, and Alhussein, ⁋101, “Operations of the layers of the Bayesian neural network may be performed by the operation circuit 603 or the vector computation unit 607 [wherein the forecasting DNN and the uncertainty DNN include].”).
Regarding claim 6, Alhussein in view of Amiri and Kawashima teaches the method of claim 1. Alhussein further teaches wherein the uncertainty includes a range over time. (Alhussein, ⁋69, “determines that the uncertainty of the traffic forecasting model for a certain time period [wherein the uncertainty includes a range over time.]”).
Regarding claim 8, Alhussein in view of Amiri and Kawashima teaches the method of claim 1. Alhussein further teaches wherein the uncertainty is based on noise including any of a variance of the noise, a probability the noise is higher than a threshold, and a sign of the noise. (Alhussein, ⁋55, “Bayesian neural network or other suitable model as would be readily understood) resembles a Bayesian or probabilistic traffic forecasting model 215 that is capable of extracting the predictive mean (μ), the model uncertainty (σm 2), and the inherent noise or aleatoric uncertainty (σn 2) [noise] associated with the forecasting model.”, and Alhussein, ⁋59, “the traffic forecasting module to mitigate instances where an overall weighted uncertainty measure exceeds a threshold [wherein the uncertainty is based on noise including any of a variance of the noise, a probability the noise is higher than a threshold, and a sign of the noise.]. The overall weighted uncertainty measure can be defined as provided in Equation 2. γ1σm 2+γ2σn 2 (2)”).
Regarding claim 9, Alhussein in view of Amiri and Kawashima teaches the method of claim 1. Alhussein further discloses wherein steps further include responsive to the uncertainty satisfying a criterion, automatically initiating the predetermined network action by at least one of proactively re-routing connections in the network, provisioning additional network resources, or instructing deployment of virtualized network functions (VNFs). (Alhussein, ⁋35, “According to embodiments, among two different types of the uncertainties for network traffic demand, epistemic uncertainty (or model uncertainty) and aleatoric uncertainty (e.g. data uncertainty or inherent noise or random uncertainty) are utilized to guide TE frameworks [wherein steps further include responsive to the uncertainty satisfying a criterion,]”, and Alhussein, ⁋56, “The controller node 230 is configured to re-configure or re-design the network topology, and set or assign the network resource reservations for the existing network services and requests. The controller node 230, at step 203, sets resource reservations based on a function of the extracted values (i.e. the predictive mean (μ), the epistemic uncertainty (model uncertainty) (σm 2), and the inherent noise or aleatoric uncertainty (σn 2)) [automatically initiating the predetermined network action by at least one of proactively re-routing connections in the network, provisioning additional network resources, or instructing deployment of virtualized network functions (VNFs).].”).
Regarding claim 10, the claim is similar to claim 1 and rejected under the same rationales. Alhussein further teaches the additional limitations A non-transitory computer-readable medium configured to store a program executable by a processing system, the program including instructions configured to cause the processing system to perform steps of: (Alhussein, ⁋18, “In accordance with embodiments of the present disclosure, there is provided a non-transitory computer-readable medium storing machine executable instructions. The machine executable instructions, when executed by a processor of a device for [A non-transitory computer-readable medium configured to store a program executable by a processing system, the program including instructions configured to cause the processing system to perform steps of:]”).
Regarding claims 11-13, 15, and 17-18, the claims are similar to claims 2-4, 6, and 8-9 and are rejected under the same rationales.
Regarding claim 19, the claim is similar to claim 1 and rejected under the same rationales. Alhussein further discloses the additional limitations A computing system comprising: a processing device and memory comprising instructions that, when executed, cause the processing device to (Alhussein, claim 13, “An apparatus for managing network traffic in a communication network, the apparatus comprising: at least one processor; and at least one machine-readable medium storing machine executable instructions which when executed by the at least one processor configure the apparatus to: [A computing system comprising: a processing device and memory comprising instructions that, when executed, cause the processing device to]”).
Regarding claim 20, the claim is similar to claim 2 and rejected under the same rationales.
Claims 5 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Alhussein, et al., US Pre-Grant Publication 2023/0216811A1 (“Alhussein”) in view of Amiri, et al., US Pre-Grant Publication US20210150305A1 (“Amiri”) and further in view of Kawashima, et al., Non-Patent Literature “The Aleatoric Uncertainty Estimation Using a Separate Formulation with Virtual Residuals” (“Kawashima”) and Arnold, et al., US Pre-Grant Publication 2021/0312323A1(“Arnold”).
Regarding claim 5, Alhussein in view of Amiri and Kawashima teaches the method of claim 1. While the combination teaches a system that predicts the uncertainty related to the forecasting future demand of network traffic using a hybrid forecasting and uncertainty DNN, the combination does not explicitly teach wherein the forecasting DNN and the uncertainty DNN include different components.
Arnold teaches wherein the forecasting DNN and the uncertainty DNN include different components. (Arnold, ⁋17, “a stacked meta-modeling approach is employed to generate the performance prediction and the uncertainty prediction. For example, embodiments generate two levels of meta-models: (1) a first level meta-model to predict the performance of a base model [wherein the forecasting DNN] and (2) a second level meta-meta-model to predict the uncertainty of the first level meta-model [and the uncertainty DNN]…Using a stacked meta-modeling approach enables embodiments to work with models of any architecture [include different components.]”).
Alhussein, in view of Amiri and Kawashima, Arnold are both in the same field of endeavor (i.e. uncertainty quantification). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Alhussein, in view of Amiri and Kawashima, Arnold to teach the above limitation(s). The motivation for doing so is that using a cascaded architecture allows for more robust selections of models to be used uncertainty analyses (cf. Arnold, ⁋17, “Using a stacked meta-modeling approach enables embodiments to work with models of any architecture (e.g., the stacked meta-modeling approach is agnostic to the model architecture and is not reliant on having access the inner parameters of the model) and capture multiple types of uncertainty (e.g., aleatoric uncertainty, epistemic uncertainty, etc.).”).
Regarding claim 14, the claim is similar to claim 5 and rejected under the same rationales.
Allowable Subject Matter
Claims 7 and 16 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims. The following is a statement of reasons for indication of allowable subject matter:
Regarding claim 7, Below are the closest cited references, each of which disclose various aspects of the claimed invention:
Kawashima, et al., “The Aleatoric Uncertainty Estimation Using a Separate Formulation with Virtual Residuals” discloses a system that uses virtual residuals to separate the model prediction and uncertainty estimation functions for regression problems. Kawashima teaches a loss function for determining aleatoric uncertainty in regression problems by separating the model prediction and uncertainty estimation using virtual residuals. While Kawashima teaches using residuals, Kawashima does not explicitly teach a joint loss function that jointly includes (i) a distance between predicted future points and corresponding actual future points and (ii) a distance between the residual vector and an estimated residual vector output.
Brando, et al., “Uncertainty Modelling in Deep Networks: Forecasting Short and Noisy Series” teaches a system that considers the effects of homoscedastic and heteroscedastic aleatoric uncertainty in forecasting models. Brando introduces aleatoric uncertainty into a base forecasting neural network model to force the model to adapt to the added uncertainty. While Brando teaches a system that considers the aleatoric uncertainty in forecasting models, Brando does not explicitly teach the joint loss function that jointly includes (i) a distance between predicted future points and corresponding actual future points and (ii) a distance between the residual vector and an estimated residual vector output.
Griffiths, et al., “Achieving robustness to aleatoric uncertainty with heteroscedastic Bayesian optimization” teaches a heteroscedastic Bayesian optimization system capable of representing and minimizing aleatoric noise across the input space. Griffiths’s system employs a heteroscedastic Gaussian process surrogate model in conjunction with two heuristics: an augmented expected improvement heuristic and an aleatoric noise-penalized expected improvement heuristic. While Griffiths teaches a system that considers aleatoric uncertainty using Bayesian optimization, Griffiths does not explicitly teach using a joint loss function that jointly includes (i) a distance between predicted future points and corresponding actual future points and (ii) a distance between the residual vector and an estimated residual vector output.
While the above prior arts disclose the aforementioned concepts, however, none of the prior arts, individually or in reasonable combination, discloses all the limitations in the manner recited in claim 7. Specifically, the claim requires updating weights of the forecasting DNN and the uncertainty DNN by optimizing a loss function that jointly includes (i) a distance between predicted future points and corresponding actual future points and (ii) a distance between the residual vector and an estimated residual vector output by the uncertainty DNN. While the references cited above mention aspects of forecasting and uncertainty quantification using DNNs, they do not recite claim 7’s specific joint loss function of updating weights of the forecasting DNN and the uncertainty DNN by optimizing a loss function that jointly includes (i) a distance between predicted future points and corresponding actual future points and (ii) a distance between the residual vector and an estimated residual vector output by the uncertainty DNN. Therefore, claim 7 is allowable over the prior art.
Regarding claim 16, the claim is similar to claim 7 and allowable over the prior art under the same rationales.
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 NICHOLAS S WU whose telephone number is (571)270-0939. The examiner can normally be reached Monday - Friday 8:00 am - 4:00 pm EST.
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, Michelle Bechtold can be reached at 571-431-0762. 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.
/N.S.W./Examiner, Art Unit 2148
/MICHELLE T BECHTOLD/Supervisory Patent Examiner, Art Unit 2148
Prosecution Insights