DETAILED ACTION
The office action is a response to an application filed on November 24, 2024, wherein claims 1-13 are pending and ready for examination.
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 Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 10 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
As per specification in para [0017] “the first P value and the last Q values from the first temporal sequence of output values...”
Claim 10 is unclear that “the first P values and the last Q values” as being indefinite that are pointing to claim 1. “P” and “Q” are not defined.
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.
Claims 1, 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over Vasseur et al. (Vasseur hereafter) (US 20150193694 A1) in view of Morrison et al. (Morrison hereafter) (US 20230197093 A1) and in further view of Iyer et al. (Iyer hereafter) (US 20240311962 A1).
Regarding claim 1 Vasseur teaches, A method comprising:
upsampling a network traffic trace representing variations over time of a traffic-related counter based on (Vasseur; [0019] The collection of the first set of status data is then scheduled. Another request may also be received by the network policy engine to begin collecting a second set of status data from the plurality of network devices corresponding to at least one of the network devices simulating the type of network attack):
obtaining a first temporal sequence of input values of the traffic-related counter at an input rate (Vasseur; [0079] a first data set is received by a network device that is indicative of the statuses of a plurality of network devices when a type of network attack is not present. [0087] The first phase of data collection may entail gathering attack-free data with a large spectrum of conditions (e.g., the density of the nodes in the network, traffic load, etc.). As shown in FIG. 4B, data collection may be initiated by the agent on the field area router (FAR) (e.g., FAR-1), which will need to query the policy engine for authorization) wherein the input values are in a first range (Vasseur; [0093] The policy engine, in turn, may respond with a data collection response/reply message, as shown in FIG. 4C... the policy engine may require the FAR to reduce the duration or the traffic overhead of the data gathering or to postpone the data collection or even not perform data collection according to the policy.);
Vasseur fails to explicitly teach, preprocessing the first temporal sequence of input values to generate a first sequence of scaled values at an output rate equal to the input rate multiplied by an upsampling factor K, wherein the scaled values are in a second range derived from the first range by a non-linear function
Morrison teaches, preprocessing the first temporal sequence of input values (frame-rate network 276) to generate a first sequence of scaled values (the sample-rate network 278) at an output rate equal to the input rate multiplied by an upsampling factor K (are related by an upsampling factor k.), wherein the scaled values are in a second range derived from the first range by a non-linear function ([0046]The neural vocoder 175 outputs a synthesized output sample s.sub.t, which is raw audio data and which is a combination of the excitation value and the prediction value.) (Morrison; [0046] The neural vocoder 175 outputs a synthesized output sample s.sub.t, which is raw audio data and which is a combination of the excitation value and the prediction value. [0048] Time resolutions of the frame-rate network 276 and the sample-rate network 278 are related by an upsampling factor k. For every frame processed by the frame-rate network 276, the sample-rate network 278 produces k samples without overlap between frames...);
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur to include the above recited limitations as taught by Morrison in order to reduce regions of silence. (Morrison; [0049])
Vasseur-Morrison fails to explicitly teach, applying a trained iterative denoising process to the first sequence of scaled values stacked with a first noise signal at the output rate to generate a first sequence of denoised upsampled values at the output rate in the second range, wherein the first noise signal has values in the second range
postprocessing of the first sequence of denoised upsampled values to generate a first temporal sequence of output values in the first range at the output rate, wherein each input value in the first temporal sequence of input values is equal to a sum of K corresponding successive output values
However, in the same field of endeavor Iyer teaches, applying a trained iterative ((e.g., for k=3)) denoising process to the first sequence of scaled values stacked with a first noise signal at the output rate (flow upsampling and denoising network 3700 predicts the filter mask consisting of local 3×3 filters from the input (e.g., for k=3)) to generate a first sequence of denoised upsampled values at the output rate in the second range (the predicted filters to denoise the optical flow after the optical flow is upsampled to the source resolution.), wherein the first noise signal has values in the second range (The final optical flow estimate is then denoised into the final optical flow estimate that is used for frame generation) (Iyer; [0422]The flow upsampling and denoising network 3700 predicts the filter mask consisting of local 3×3 filters from the input (e.g., for k=3). The flow upsampling and denoising network 3700 uses the predicted filters to denoise the optical flow after the optical flow is upsampled to the source resolution. Thus, the upsampling and denoising network 3700 enables upsampling, as well as denoising, for input optical flow that is less than the source resolution of the frame data. The upsampled and denoised optical flow data is then used as input to the optical flow prediction network to perform optical flow estimation for the purposes of frame generation. the denoising portion 3300 of the optical flow prediction network 3200 predicts local filters from the output of the stage prior to the stage that calculates final optical flow estimate for use in frame generation. The final optical flow estimate is then denoised into the final optical flow estimate that is used for frame generation);
postprocessing of the first sequence of denoised upsampled values to generate a first temporal sequence of output values in the first range at the output rate (to perform end-to-end training of a neural frame prediction network with an extrapolation dataset to generate extrapolation weights), wherein each input value in the first temporal sequence of input values is equal to a sum of K corresponding successive output values (The extrapolation weights and the interpolation weights can then be deployed along with the neural frame prediction network) (Iyer; [0424] the method 3800 includes operations to perform end-to-end training of a neural frame prediction network with an extrapolation dataset to generate extrapolation weights for frame generation and super resolution (3802) and also perform end-to-end training of a neural frame prediction network with an interpolation dataset to generate interpolation weights for frame generation and super resolution (3804). The extrapolation weights and the interpolation weights can then be deployed along with the neural frame prediction network... the n×n optical flow is upsampled to 1×1 optical flow and then compared with the 1×1 dense optical flow training data that serves as a pseudo ground truth. This comparison is used to generate a ‘flow’ loss. The predicted frame is also compared with ground truth frame to generate ‘image’ loss. The flow loss and image loss are added to generate a total loss. The total loss is then backpropagated to update the parameters of the network).
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur-Morrison to include the above recited limitations as taught by Iyer in order to improve the quality of the predicted frame (Iyer; [0424]).
Regarding claim 13 Vasseur teaches, An apparatus comprising:
at least one processor (Vasseur; [Fig. 2] 220); and
at least one memory including a computer program code, the at least one memory (Vasseur; [Fig. 2] 240) and computer program code configured to, with the at least one processor, cause the apparatus to:
upsampling a network traffic trace representing variations over time of a traffic-related counter based on (Vasseur; [0019] The collection of the first set of status data is then scheduled. Another request may also be received by the network policy engine to begin collecting a second set of status data from the plurality of network devices corresponding to at least one of the network devices simulating the type of network attack):
obtaining a first temporal sequence of input values of the traffic-related counter at an input rate (Vasseur; [0079] a first data set is received by a network device that is indicative of the statuses of a plurality of network devices when a type of network attack is not present. [0087] The first phase of data collection may entail gathering attack-free data with a large spectrum of conditions (e.g., the density of the nodes in the network, traffic load, etc.). As shown in FIG. 4B, data collection may be initiated by the agent on the FAR (e.g., FAR-1), which will need to query the policy engine for authorization) wherein the input values are in a first range (Vasseur; [0093] The policy engine, in turn, may respond with a data collection response/reply message, as shown in FIG. 4C... the policy engine may require the FAR to reduce the duration or the traffic overhead of the data gathering or to postpone the data collection or even not perform data collection according to the policy.);
Vasseur fails to explicitly teach, preprocessing the first temporal sequence of input values to generate a first sequence of scaled values at an output rate equal to the input rate multiplied by an upsampling factor K, wherein the scaled values are in a second range derived from the first range by a non-linear function
Morrison teaches, preprocessing the first temporal sequence of input values (frame-rate network 276) to generate a first sequence of scaled values (the sample-rate network 278) at an output rate equal to the input rate multiplied by an upsampling factor K (are related by an upsampling factor k.), wherein the scaled values are in a second range derived from the first range by a non-linear function ([0046]The neural vocoder 175 outputs a synthesized output sample s.sub.t, which is raw audio data and which is a combination of the excitation value and the prediction value.) (Morrison; [0046] The neural vocoder 175 outputs a synthesized output sample s.sub.t, which is raw audio data and which is a combination of the excitation value and the prediction value. [0048] Time resolutions of the frame-rate network 276 and the sample-rate network 278 are related by an upsampling factor k. For every frame processed by the frame-rate network 276, the sample-rate network 278 produces k samples without overlap between frames...);
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur to include the above recited limitations as taught by Morrison in order to reduce regions of silence. (Morrison; [0049])
Vasseur-Morrison fails to explicitly teach, applying a trained iterative denoising process to the first sequence of scaled values stacked with a first noise signal at the output rate to generate a first sequence of denoised upsampled values at the output rate in the second range, wherein the first noise signal has values in the second range
postprocessing of the first sequence of denoised upsampled values to generate a first temporal sequence of output values in the first range at the output rate, wherein each input value in the first temporal sequence of input values is equal to a sum of K corresponding successive output values
However, in the same field of endeavor Iyer teaches, applying a trained iterative ((e.g., for k=3)) denoising process to the first sequence of scaled values stacked with a first noise signal at the output rate (flow upsampling and denoising network 3700 predicts the filter mask consisting of local 3×3 filters from the input (e.g., for k=3)) to generate a first sequence of denoised upsampled values at the output rate in the second range (the predicted filters to denoise the optical flow after the optical flow is upsampled to the source resolution.), wherein the first noise signal has values in the second range (The final optical flow estimate is then denoised into the final optical flow estimate that is used for frame generation) (Iyer; [0422]The flow upsampling and denoising network 3700 predicts the filter mask consisting of local 3×3 filters from the input (e.g., for k=3). The flow upsampling and denoising network 3700 uses the predicted filters to denoise the optical flow after the optical flow is upsampled to the source resolution. Thus, the upsampling and denoising network 3700 enables upsampling, as well as denoising, for input optical flow that is less than the source resolution of the frame data. The upsampled and denoised optical flow data is then used as input to the optical flow prediction network to perform optical flow estimation for the purposes of frame generation. the denoising portion 3300 of the optical flow prediction network 3200 predicts local filters from the output of the stage prior to the stage that calculates final optical flow estimate for use in frame generation. The final optical flow estimate is then denoised into the final optical flow estimate that is used for frame generation);
postprocessing of the first sequence of denoised upsampled values to generate a first temporal sequence of output values in the first range at the output rate (to perform end-to-end training of a neural frame prediction network with an extrapolation dataset to generate extrapolation weights), wherein each input value in the first temporal sequence of input values is equal to a sum of K corresponding successive output values (The extrapolation weights and the interpolation weights can then be deployed along with the neural frame prediction network) (Iyer; [0424] the method 3800 includes operations to perform end-to-end training of a neural frame prediction network with an extrapolation dataset to generate extrapolation weights for frame generation and super resolution (3802) and also perform end-to-end training of a neural frame prediction network with an interpolation dataset to generate interpolation weights for frame generation and super resolution (3804). The extrapolation weights and the interpolation weights can then be deployed along with the neural frame prediction network... the n×n optical flow is upsampled to 1×1 optical flow and then compared with the 1×1 dense optical flow training data that serves as a pseudo ground truth. This comparison is used to generate a ‘flow’ loss. The predicted frame is also compared with ground truth frame to generate ‘image’ loss. The flow loss and image loss are added to generate a total loss. The total loss is then backpropagated to update the parameters of the network).
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur-Morrison to include the above recited limitations as taught by Iyer in order to improve the quality of the predicted frame (Iyer; [0424]).
Regarding claim 2 Vasseur-Morrison-Iyer teaches, The method of claim 1,
Vasseur-Morrison fails to explicitly teach, wherein preprocessing the first temporal sequence of input values comprises:
upsampling the first temporal sequence of input values by generating, from each of the input values, K upsampled values to generate a first sequence of upsampled values such that the sum of the K upsampled values is equal to the considered input value
applying a scaling function to each of the upsampled values to generate the first sequence of scaled values in the second range
Iyer further teaches, upsampling the first temporal sequence of input values by generating, from each of the input values, K upsampled values to generate a first sequence of upsampled values such that the sum of the K upsampled values is equal to the considered input value (Iyer; [0420] FIG. 37 illustrates an optical flow upsampling and denoising network 3700, according to an embodiment. The flow upsampling and denoising network 3700, takes as input 3702 a filter kernel size k, and an upsampling factor f as parameters, along with sub source resolution optical flow that is at a resolution that is less than the source resolution.);
applying a scaling function to each of the upsampled values to generate the first sequence of scaled values in the second range (Iyer; [0420]...kernel size that may be used is k=3, such that the optical flow upsampling and denoising network 3700 will predict 3×3 filter. The sub source resolution optical flow includes a motion vector for each n×n pixels of the rendered frames, such as a motion vector for each 8×8 pixels when n=8).
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur-Morrison to include the above recited limitations as taught by Iyer in order to improve the quality of the predicted frame (Iyer; [0424]).
Regarding claim 3 Vasseur-Morrison-Iyer teaches, The method of claim 2,
Vasseur-Morrison fails to explicitly teach, wherein generating the K upsampled values comprises replacing each of the input values by K equal upsampled values, each computed by dividing the considered input value by K
Iyer further teaches, wherein generating the K upsampled values comprises replacing each of the input values by K equal upsampled values, each computed by dividing the considered input value by K (Iyer; [0421]The filter predictor stage 3710 can predict the k×k filter kernel that corresponds to each pixel of upsampled optical flow in the form of h*f×w*f×2 for upsampling factor f. The filter kernels are normalized using softmax so that the weights sum to 1 (3714). In the filter application stage 3720, the network first upscales the input lower resolution optical flow by the input upsampling factor to generate an intermediate full resolution optical flow according to upsample factor f ).
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur-Morrison to include the above recited limitations as taught by Iyer in order to improve the quality of the predicted frame (Iyer; [0424]).
Regarding claim 5 Vasseur-Morrison-Iyer teaches, The method of claim 1,
Vasseur-Morrison fails to explicitly teach, wherein the first sequence of scaled values is used as conditioning signal for the trained iterative denoising process
Iyer further teaches, wherein the first sequence of scaled values is used as conditioning signal for the trained iterative denoising process (Iyer; [0405] the denoising portion 3300 is a portion of the optical flow prediction network 3200 of FIG. 32 and operates on concatenated intermediate output...[0408] the performance of operations to train optical flow prediction for a unified extrapolation/interpolation neural network with an extrapolation dataset to generate extrapolation weights for frame generation (3402) and also to train optical flow prediction for the unified extrapolation/interpolation neural network with an interpolation dataset to generate interpolation weights for frame generation).
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Li-Morrison to include the above recited limitations as taught by Iyer in order to improve the quality of the predicted frame (Iyer; [0424]).
Regarding claim 6 Vasseur-Morrison-Iyer teaches, The method of claim 1,
Vasseur-Morrison fails to explicitly teach, wherein the trained iterative denoising process uses a U-net model iteratively executed.
Iyer further teaches, wherein the trained iterative denoising process uses a U-net model iteratively executed. (Iyer; [0391]The unified network makes use of a single neural network having a modified U-Net architecture that accepts two rendered frames and optical flow between those frames as inputs, [0402]).
Regarding claim 11 Vasseur-Morrison-Iyer teaches, The method of claim 1,
Vasseur further teaches obtaining a second temporal sequence of input values of the traffic-related counter at the input rate in the second range (Vasseur; [0079] A second data set is also received that is indicative of the statuses of the plurality of network devices when the type of network attack is present.), wherein each of the first and second sequences concerns traffic via a respective communication channel over a same physical or logical transmission link ([0082] An agent on the FAR that is configured to both receive the statistics produced by the nodes and to query the nodes for additional information. In some embodiments, the FAR may also have full visibility of the IP topology of the PAN and may train and apply the learning machine in charge of the final decision about the detection of an attack.);
preprocessing the second temporal sequence of input values to generate a second sequence of scaled values at the output rate, wherein the scaled values of the second sequence of scaled values are in the second range (Vasseur; [0096] After a sufficient attack-free dataset has been collected, a dataset including attacks may be generated in order to appropriately train the ANN or other learning machine. Thus, the techniques herein provide mechanisms whereby attacks are generated and simulated (i.e., initiated by the network itself in a controlled manner). In a simple form, attacks may be generated using a traffic generator.);
applying a trained iterative denoising process to the second sequence of scaled values and a second noise signal at the output rate to generate a second sequence of denoised upsampled values at the output rate in the second range, wherein the second noise signal has values in the second range, wherein the trained iterative denoising process is applied jointly to the first sequence of scaled values, the first noise signal, the second sequence of scaled values and the second noise signal (Vasseur; [0096]
the learning machine hosted on the router (e.g., a FAR) may select nodes of its WPAN to act as attackers, thus being able to correctly label the gathered statistics. In other words, this component of the techniques and architecture manages the generation and collection of this type of labeled data.);
postprocessing of the second sequence of denoised upsampled values to generate a second temporal sequence of output values in the first range at the output rate, wherein each input value in the second temporal sequence of input values is equal to a sum of K corresponding successive output values in the second temporal sequence of output values (Vasseur; [0112] Whenever training has been carried out and the new ANN classifier or other model has been installed, the FAR agent may send to the policy engine a new CoAP message called a Classifier_Enabled message, to inform the NMS/policy engine that a classifier has been enabled.).
Regarding claim 12 Vasseur-Morrison-Iyer teaches, The method of claim 1,
Vasseur further teaches comprising: performing an operation on one or more network devices or network function based on one or more output values of the first temporal sequence of output values (Vasseur; [0112] Such a message may also include the computed classifier, in one embodiment, which can be used by the policy engine to double check data from another FAR in case of attack detection).
Claims 4 is rejected under 35 U.S.C. 103 as being unpatentable over Vasseur-Morrison-Iyer as applied to claim 1 above, and further in view of Saharia et al. (Saharia hereafter) (US 20230067841 A1).
Regarding claim 4 Vasseur-Morrison-Iyer teaches, The method of claim 1,
Vasseur-Morrison-Iyer fails to explicitly teach, wherein the trained iterative denoising process is based on a denoising diffusion probabilistic model
However, in the same field of endeavor Saharia teaches, wherein the trained iterative denoising process is based on a denoising diffusion probabilistic model (Saharia; [0045] an approach based on adapting the denoising diffusion probabilistic (DDPM) model to conditional image generation may be utilized.).
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur-Morrison-Iyer to include the above recited limitations as taught by Saharia in order to generate a target image in T refinement steps (Saharia; [0046]).
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Vasseur-Morrison-Iyer as applied to claim 1 above, and further in view of Saharia and Chen et al. (Chen hereafter) (US 20260134543 A1).
Regarding claim 7 Vasseur-Morrison-Iyer teaches, The method of claim 1,
Vasseur-Morrison fails to explicitly teach, comprising training the iterative denoising process by:
obtaining a temporal sequence of input training values of the traffic-related counter at the second rate, wherein the input training values are in the first range
obtaining a temporal sequence of input training values of the traffic-related counter at the second rate, wherein the input training values are in the first range
obtaining a sequence of aggregated training values at the first rate in the first range, the aggregated training values corresponding to an aggregated version of the input training values
scaling the input training values to generate a first sequence of scaled training values in the second range at the second rate
preprocessing the sequence of aggregated training values to generate a second sequence of scaled training values in the second range at the second rate
adding a noise signal in the second range to the first sequence of scaled training values to generate a sequence of noisy training values
applying one iteration of the iterative denoising process to the sequence of noisy training values using the first sequence of scaled training values as target signal and the second sequence of scaled training values as conditioning signal to generate a sequence of output denoised values
adapting one or more parameters of the iterative denoising process based on a loss function that evaluate a remaining noise in the sequence of output denoised values
Iyer further teaches, obtaining a temporal sequence of input training values of the traffic-related counter at the second rate, wherein the input training values are in the first range (Iyer; [0418] Fig. 36 Frame generation logic and associated circuitry that are configured to perform method 3600 can load extrapolation weights for a unified flow prediction network);
obtaining a sequence of aggregated training values at the first rate in the first range, the aggregated training values corresponding to an aggregated version of the input training values (Iyer; [0418] The frame generation logic can then receive RGB data associated with a pair of rendered frames, along with sub source resolution optical flow between those frames, such as, for example, 8×8 optical flow (3602).);
scaling the input training values to generate a first sequence of scaled training values in the second range at the second rate (Iyer; [0418] The frame generation logic can upscale the lower resolution optical flow to a denoised full resolution optical flow using a flow upsampling and denoising network (3625).);
preprocessing the sequence of aggregated training values to generate a second sequence of scaled training values in the second range at the second rate (Iyer; [0419] The frame generation logic and associated circuitry can pre-process the input frames by calculating an overall mean for the two input frames and subtracting the overall mean from each frame (3604));
adding a noise signal in the second range to the first sequence of scaled training values to generate a sequence of noisy training values (Iyer; [0410] the application can load pre-trained extrapolation weights into the optical flow prediction network for the unified interpolation/extrapolation network (3414). The unified interpolation/extrapolation network can then predict the optical flow between a rendered frame and a future frame (3416).);
Vasseur-Morrison fails to explicitly teach, applying one iteration of the iterative denoising process to the sequence of noisy training values using the first sequence of scaled training values as target signal and the second sequence of scaled training values as conditioning signal to generate a sequence of output denoised values
Saharia further teaches, applying one iteration of the iterative denoising process to the sequence of noisy training values using the first sequence of scaled training values as target signal and the second sequence of scaled training values as conditioning signal to generate a sequence of output denoised values (Saharia; [0077] FIG. 3A is a diagram illustrating an example architecture 300A for a neural network, in accordance with example embodiments. FIG. 3A is a description of an example U-Net architecture 300A with skip connections. The low resolution input image 305, x, may be interpolated to a target high resolution, and concatenated with the noisy high resolution image 310, y.sub.t.);
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur-Morrison-Iyer to include the above recited limitations as taught by Saharia in order to generate a target image in T refinement steps.( (Saharia; [0046]).
Vasseur-Morrison-Iyer-Saharia fails to explicitly teach, adapting one or more parameters of the iterative denoising process based on a loss function that evaluate a remaining noise in the sequence of output denoised values
However, in the same field of endeavor Chen teaches, adapting one or more parameters of the iterative denoising process based on a loss function that evaluate a remaining noise in the sequence of output denoised values (Chen; [0097] The parameters of the denoising diffusion model are adjusted to reduce the loss function, improving the accuracy of the denoising diffusion model's predictions. In some implementations, this adjustment could be performed using various optimization algorithms, such as stochastic gradient descent (SGD), Adam, RMSProp, or other suitable optimization algorithms. This iterative process continues until the denoising diffusion model is sufficiently trained to perform accurate panoptic segmentation,).
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur-Morrison-Iyer-Saharia to include the above recited limitations as taught by Chen in order to minimize this loss function, leading to more accurate predictions (Chen; [0046]).
Claim 8 is rejected under 35 U.S.C. 103 as being unpatentable over Vasseur-Morrison-Iyer as applied to claim 1 above, and further in view of Agarwal et al. (Agarwal hereafter) (US 20240428468 A1).
Regarding claim 8 Vasseur-Morrison-Iyer teaches, The method of claim 1,
Vasseur-Morrison-Iyer fails to explicitly teach, wherein postprocessing of the first sequence of denoised upsampled values comprises:
scaling the first sequence of denoised upsampled values to generate a first sequence of scaled upsampled values in the first range.
However, in the same field of endeavor Agarwal teaches, scaling the first sequence of denoised upsampled values to generate a first sequence of scaled upsampled values in the first range (Agarwal; [0087] An adaptive step size includes a scalar value (e.g., a value that indicates a single component such as magnitude) which modifies a noise representation of the denoising neural network at a specific denoising step. The adaptive step size varies at different steps of the denoising process);
generating the first temporal sequence of output values by adjusting values in the first sequence of scaled upsampled values such that each input value in the first temporal sequence of input values is equal to a sum of K corresponding successive output values (Agarwal; [0074] FIG. 5 shows the attention image synthesis system 102 generating a noise representation 514 from denoising neural network 500 and subsequently generating a modified noise representation 516 from the noise representation 514. Furthermore, based on the denoising neural network 518 processing the modified noise representation 516, the attention image synthesis system 102 generates an additional noise representation 520.).
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur-Morrison-Iyer to include the above recited limitations as taught by Agarwal in order to modify the noise representation (Agarwal; [0087]).
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Vasseur-Morrison-Iyer-Agarwal as applied to claim 1 above, and further in view of Bredehoft (Bredehoft hereafter) (US 20070061390 A1).
Regarding claim 9 Vasseur-Morrison-Iyer-Agarwal teaches, The method of claim 9,
Vasseur-Morrison-Iyer-Agarwal fails to explicitly teach, wherein adjusting values in the first sequence of scaled upsampled values comprises applying a linear scaling factor to values in the first sequence of scaled upsampled values, wherein the linear scaling factor is computed based on the input value and the sum of K corresponding successive output values
However, in the same field of endeavor Bredehoft teaches, wherein adjusting values in the first sequence of scaled upsampled values comprises applying a linear scaling factor to values in the first sequence of scaled upsampled values, wherein the linear scaling factor is computed based on the input value and the sum of K corresponding successive output values (Bredehoft; [0028] This matrix equation demonstrates that each component P.sub.k.sup.(0) of the interpolator ouput is a linear function of the initial conditions. [0029] The initial value for linear component P.sub.1 of a Nyquist filter is given by P.sub.1.sup.(0)[0]=y[Mj], which represents a sample value at the start of the interpolation interval. [0032] For an L-lobe linear-phase Nyquist filter whose starting and ending derivatives are zero, the cost of the above computation is only (L-1)/2+4 multiplies per input sample, and is independent of the upsampling factor M.).
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur-Morrison-Iyer-Agarwal to include the above recited limitations as taught by Bredehoft in order to provide good performance (Bredehoft; [0006]).
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Vasseur-Morrison-Iyer as applied to claim 1 above, and further in view of Suh et al. (Suh hereafter) (US20090175353 A1).
Regarding claim 10 Vasseur-Morrison-Iyer teaches, The method of claim 1,
Vasseur-Morrison-Iyer fails to explicitly teach, comprising discarding the first P values and the last Q values from the first temporal sequence of output values or signaling that the first P values and the last Q values from the first temporal sequence of output values are less reliable
However, in the same field of endeavor Suh teaches, comprising discarding the first P values and the last Q values from the first temporal sequence of output values or signaling that the first P values and the last Q values from the first temporal sequence of output values are less reliable (Suh; [0009] The MANE in FIG. 3 parses P, D, T, and Q values (S11) from each NALU (which is generated by the encoder or transmitted over wired or wireless networks) and, then, parsed P, D, T, and Q
values are compared to pre-defined values (p, d, t, q) in order for the MANE to decide whether it forwards or discards each NALU (S12), [0104].).
It would have been obvious to one of ordinary skilled in the art before the effective filing date to create the invention of Vasseur-Morrison-Iyer to include the above recited limitations as taught bySuh in order to control quality of video service (Suh; [0114]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to WILFRED THOMAS whose telephone number is (571)270-0353. The examiner can normally be reached Mon -Thurs 9:00 am-4:00 pm.
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, Noel R Beharry can be reached at 571-270-5630. 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.
/W. T/ Examiner, Art Unit 2416
/KENNY S LIN/ Primary Examiner, Art Unit 2416
