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 .
This action is in response to amendments filed March 18th, 2026. The status of the claims is as follows. Claims 1, 9 and 25 are amended. Claims 1-30 are currently pending.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-2; 9-10; 17-18; 25-26 are rejected under 35 U.S.C. 103 as being unpatentable over Bao et al. (“Using Distillation to Improve Network Performance after Pruning and Quantization” [2019], hereinafter “Bao”) in view of Miles et al. (“Cascaded channel pruning using hierarchical self-distillation” [2020], hereinafter “Miles”) further in view of Courville et al. (US20220036189A1, hereinafter “Courville”)
Regarding Claim 1,
Bao discloses receiving an initial neural network model; (Bao [Figure 1];
PNG
media_image1.png
206
352
media_image1.png
Greyscale
Where the teacher model read on as an initial neural network model being present in the framework indicates at some point the receiving of an initial neural network model)
pruning the initial neural network model based on a first threshold to generate a pruned network and a pruned set of weights …; (Bao [Section 3 Paragraph 2]; “Firstly, the filter-level pruning algorithm [4] is used to prune the teacher model. The steps of pruning operation: first setting the pruning rate, then we evaluate the importance of all the convolution kernels in the whole network and find the least important one and cut it out. Then judging whether the pruning rate at this time is higher than the set pruning rate. If not, we repeat the pruning operation or end the pruning”
Bao [Figure 2];
PNG
media_image2.png
280
242
media_image2.png
Greyscale
Wherein a pruned network and pruned weights are disclosed)
applying a quantization process to the pruned network to produce a pruned and quantized network; (Bao [Section 3 Paragraph 3]; “Secondly, We need to quantize[19] the pruned student model. The core operation of quantization is to select the appropriate scaling function”)
generating a teacher model … ; (Bao [Figure 1];
PNG
media_image3.png
201
354
media_image3.png
Greyscale
Wherein a model generated by incorporating the pruned set of weights with the pruned network is interpreted as simply the initial neural network model, thus the teacher model passed in as the initial neural network model is read on as a generated teacher model)
generating an initial student model from the pruned and quantized network;
(Bao [Section 3 Paragraph 3]; “After quantization, we get a new student model” wherein the new student model generated after the quantization of the pruned student model reads on a generated initial student model from the pruned and quantized network)
training the initial student model using the teacher model to output a trained student model. (Bao [Section 3 Paragraph 4]; “Finally, we use the knowledge distillation[13] to train student model, and use the information provided by teacher to guide student training. The key of joint training is to continuously optimize distillation loss function, which is defined as
PNG
media_image4.png
21
203
media_image4.png
Greyscale
which represent the cross-entropy of hard target, represent the cross-entropy of soft target, represent probability of studnet model, and represent probability with tempruature of student and teacher model, is a tunable parameter to balance both cross-entropies. After training, we got the final student model”
Bao [Figure 1];
PNG
media_image3.png
201
354
media_image3.png
Greyscale
wherein the better student model output from knowledge distillation reads on an output trained student model)
Bao fails to disclose but Miles discloses generating a teacher model by incorporating nodes and connections associated with the … pruned set of weights into the pruned network (Miles [Section I Paragraph 2]; “We propose a formulation for the updates of this importance score by extending the idea of "teaching-assistants" (TA) for knowledge distillation [23]. To enable the contribution of previously pruned filters to be re-considered into the student network, we use the gradients from a lesser-pruned TA, thus providing gradients from a model with a higher capacity. We describe the use of passing down surrogate gradients from the TAs to update the pruning masks as cascaded pruning, since the pruning is performed in a sequential fashion starting from the largest model in the hierarchy”
Miles [Figure 1];
PNG
media_image5.png
391
524
media_image5.png
Greyscale
wherein cascaded channel pruning involving “teaching-assistant” models for knowledge distillation of the final student model reads on intermediary teacher models; wherein the intermediary teacher models generated in a cascading sequence are increasingly pruned, thus reading on generating of a teacher model by incorporating nodes and connections associated with the pruned sets of weights with the pruned network)
It would have been obvious to modify Bao’s method of pruning and quantizing an inputted neural network to generate a teacher model to train a generated student model to utilize Miles’ method of generating a teacher model by incorporating nodes and connections of the pruned set of weights. One would have been motivated to do so because “By ensuring each teacher is sufficiently large to provide useful knowledge distillation, while being sufficiently small such that this gradient update is stable, very effective training can emerge,” (Miles [Page 6 Paragraph 3]) thus pruning the teacher model improves training stability.
The combination of Bao/Miles fails to explicitly disclose but Courville discloses pruning the initial neural network model based on a first threshold to generate a pruned network and a pruned set of weights, the pruned set of weights comprising a subset of the plurality of initial weights of the initial neural network model that are less than the first threshold, the pruned network excluding the subset of the plurality of initial weights (Courville [0019]; “In accordance with at least some of the preceding aspects, the method further includes for each of one or more additional convolutional layers of the convolutional neural network: generating a row pruning mask identifying a plurality of kernel rows to be pruned from a first filter of the additional convolutional layer and generating a pruned filter in accordance with the generated row pruning mask and the first filter, the pruned filter comprising at least the kernel rows of the first filter not identified as kernel rows to be pruned. The row pruning mask is generated by: generating, using the pseudo-random number generator, a sequence of pseudo-random numbers sequence based on the seed value; and for each kernel row of a first filter of the convolutional layer, determining whether to identify the kernel row of the first filter as one of the plurality of kernel rows to be pruned based on the pseudo-random number to the pruning threshold.” wherein a pruning threshold disclosed for generation of a pruned convolutional network; wherein the pruned sets of weights comprise new weights that are masked from the plurality of initial weights, thus the masked weights being read as the pruned network excluding the subset of the plurality of initial weights
Courville [0101]; “At 410, the random number generated by the PRNG is compared to the pruning threshold. If the random number is at or below the pruning threshold, this indicates that the row should be pruned, and the method proceeds to step 412. If the random number is above the pruning threshold, this indicates that the kernel row should not be pruned, and the method proceeds to step 414. It will be appreciated that the pruning threshold value and the comparison operation are arbitrary: the pruning threshold may in various embodiments be used to prune random numbers falling either above or below the pruning threshold, and the pruning threshold may be compared to either discrete or continuous values.” wherein the pruned set of weights is determined through a comparison of their value being less than the pruning threshold)
It would have been obvious to modify Bao/Miles’s method of pruning and quantizing an inputted neural network to generate a teacher model through incorporated nodes and connections to train a generated student model by implementing Courville’s pruned network to explicitly exclude the subset of the initial weights deemed to be insufficient to a threshold. One would have been motivated to do so because it “removes individual weights that are deemed to be low-value. These approaches can potentially achieve high compression while maintaining high accuracy, and the compression enables faster inference times” (Courville [0009]).
Regarding Claim 2,
Bao/Miles/Courville teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Bao/Miles/Courville further discloses providing an input to the teacher model and the initial student model; (Bao [Figure 3];
PNG
media_image6.png
141
319
media_image6.png
Greyscale
Bao [Section 3 Paragraph 4]; “Finally, we use the knowledge distillation[13] to train student model, and use the information provided by teacher to guide student training. The key of joint training is to continuously optimize distillation loss function, which is defined as
PNG
media_image7.png
25
206
media_image7.png
Greyscale
which represent the cross-entropy of hard target, represent the cross-entropy of soft target, represent probability of studnet model, and represent probability with tempruature of student and teacher model, is a tunable parameter to balance both cross-entropies. After training, we got the final student model. The framework as shown in the Fig 3”)
applying a model loss function to adjust a set of second parameters of the initial student model; (Bao [Section 3 Paragraph 4]; “The key of joint training is to continuously optimize distillation loss function, which is defined as
PNG
media_image7.png
25
206
media_image7.png
Greyscale
which represent the cross-entropy of hard target, represent the cross-entropy of soft target, represent probability of studnet model, and represent probability with tempruature of student and teacher model, is a tunable parameter to balance both cross-entropies”)
outputting the trained student model based on the adjusted set of second parameters of the initial student model; (Bao [Figure 1];
PNG
media_image3.png
201
354
media_image3.png
Greyscale
wherein the better student model output from the knowledge distillation framework reads on an output trained student model based on an adjusted set of parameters)
Claims 9-10 recites an apparatus comprising a memory and at least one processor to execute the same method of Claims 1-2. Claims 9-10 are rejected for reasons set forth in the rejection of Claims 1-2.
Claims 17-18 recites an apparatus comprising means to execute the same method of Claims 1-2. Claims 17-18 are rejected for reasons set forth in the rejection of Claims 1-2.
Claims 25-26 recites a non-transitory computer readable medium having encoded thereon program code to execute the same method of Claims 1-2. Claims 25-26 are rejected for reasons set forth in the rejection of Claims 1-2.
Claims 3, 11, and 19 are rejected under 35 U.S.C. 103 as being anticipated by Bao et al. (Using Distillation to Improve Network Performance after Pruning and Quantization” [2019], hereinafter “Bao”) in view of Miles et al. (“Cascaded channel pruning using hierarchical self-distillation” [2020], hereinafter “Miles”) further in view of Courville et al. (US20220036189A1, hereinafter “Courville”) further in view of Yanzhi Wang et al. (US20210192352A1, hereinafter “Yanzhi Wang”).
Regarding Claim 3,
Bao/Miles/Courville teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Bao/Miles/Courville fails to explicitly disclose but Yanzhi discloses in which the pruning and quantization are iteratively applied. (Yanzhi Wang [0004]; “(a) performing weight pruning on weights in the DNN model by solving a first subproblem and a second subproblem iteratively until convergence to reduce the number of weights in the DNN model to a set of non-zero weights; and (b) performing weight quantization on the set of non-zero weights by solving the first subproblem and the second subproblem iteratively until convergence to generate a weight pruned and quantized DNN model”)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify Bao/Miles/Courville’s method of pruning and quantizing an inputted neural network to generate a teacher model to train a generated student model to incorporate Yanzhi’s method of performing the pruning and quantizing iteratively. The motivation to do so lies in how dynamically updating parameters through iterations “is the key reason why ADMM-based framework outperforms conventional regularization method in DNN weight pruning and quantization” (Yanzhi Wang [0049]).
Claim 11 recites an apparatus comprising a memory and at least one processor to execute the same method of Claim 3. Claim 11 is thus rejected for reasons set forth in the rejection of Claim 3.
Claim 19 recites an apparatus comprising means to execute the same method of Claim 3. Claim 19 is thus rejected for reasons set forth in the rejection of Claim 3.
Claims 4, 12 , 20 and 27 are rejected under 35 U.S.C. 103 as being anticipated by Bao et al. (Using Distillation to Improve Network Performance after Pruning and Quantization” [2019], hereinafter “Bao”) in view of Miles et al. (“Cascaded channel pruning using hierarchical self-distillation” [2020], hereinafter “Miles”) in view of Courville et al. (US20220036189A1, hereinafter “Courville”) in view of Yanzhi Wang et al. (US20210192352A1, hereinafter “Yanzhi Wang”) and further in view of Xin Wang et al. (US20190080238, hereinafter “Xin Wang”)
Regarding Claim 4,
The combination of Bao/Miles/Courville/Yanzhi Wang teaches the method of Claim 3 (and thus the rejection of Claim 3 is incorporated). The combination fails to explicitly disclose but Xin Wang discloses in which a pruning ratio is increased with each iteration. (Xin Wang [Abstract]; “A convolution neural network (CNN) model is trained and pruned at a pruning ratio. The model is then trained and pruned one or more times without constraining the model according to any previous pruning step. The pruning ratio may be increased at each iteration until a pruning target is reached.”)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify Bao/Miles/Courville/Yanzhi Wang’s method of iteratively pruning and quantizing a neural network for training a pruned and quantized student model with a teacher model to incorporate Xin Wang’s increasing pruning ratio over iterations. The motivation to do so lies in how increasing the pruning ratio to allow more coefficients to be pruned allows networks “to reduce computational load” ([Xin Wang 0002])
Claim 12 recites an apparatus comprising a memory and at least one processor to execute the same method of Claim 4. Thus, Claim 12 is rejected for reasons set forth in the rejection of Claim 4.
Claim 20 recites an apparatus comprising means to execute the same method of Claim 4. Thus, Claim 12 is rejected for reasons set forth in the rejection of Claim 4.
Regarding Claim 27,
Bao/Miles/Courville teaches the non-transitory computer readable medium and code-executable method of Claim 25 (and thus the rejection of Claim 25 is incorporated). Bao/Miles/Courville fails to explicitly disclose but Yanzhi discloses to iteratively apply both of a pruning process to prune the subset of the plurality of initial weights of the initial neural network model and the quantization process to the pruned network. (Yanzhi Wang [0004]; “(a) performing weight pruning on weights in the DNN model by solving a first subproblem and a second subproblem iteratively until convergence to reduce the number of weights in the DNN model to a set of non-zero weights; and (b) performing weight quantization on the set of non-zero weights by solving the first subproblem and the second subproblem iteratively until convergence to generate a weight pruned and quantized DNN model”)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify Bao/Miles/Courville’s method of pruning and quantizing an inputted neural network to generate a teacher model to train a generated student model to incorporate Yanzhi Wang’s method of performing the pruning and quantizing iteratively. The motivation to do so lies in how dynamically updating parameters through iterations “is the key reason why ADMM-based framework outperforms conventional regularization method in DNN weight pruning and quantization” (Yanzhi Wang [0049]).
The combination of Bao/Miles/Courville/Yanzhi Wang fails to disclose but Xin Wang discloses wherein a pruning ratio of the pruning process is increased with each iteration. (Xin Wang [Abstract]; “A convolution neural network (CNN) model is trained and pruned at a pruning ratio. The model is then trained and pruned one or more times without constraining the model according to any previous pruning step. The pruning ratio may be increased at each iteration until a pruning target is reached.”)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify Bao/Miles/Courville/Yanzhi Wang’s method of iteratively pruning and quantizing a neural network for training a pruned and quantized student model with a teacher model to incorporate Xin Wang’s increasing pruning ratio over iterations. The motivation to do so lies in how increasing the pruning ratio to allow more coefficients to be pruned allows networks “to reduce computational load” ([Xin Wang 0002])
Claims 5-6, 13-14, 21-22, and 28-29 are rejected under 35 U.S.C. 103 as being anticipated by Bao et al. (Using Distillation to Improve Network Performance after Pruning and Quantization” [2019], hereinafter “Bao”) in view of Miles et al. (“Cascaded channel pruning using hierarchical self-distillation” [2020], hereinafter “Miles”) in view of Courville et al. (US20220036189A1, hereinafter “Courville”) in view of Xiong et al. (“Towards Efficient Compact Network Training on Edge-Devices” [2019], hereinafter “Xiong”)
Regarding Claim 5,
Bao/Miles/Courville teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Bao/Miles/Courville fails to explicitly disclose but Xiong discloses in which the quantization process comprises a quantization-aware training process. (Xiong [Section 3A Paragraph 1]; “Training-aware Symmetric Quantization For Network Training[:] Generally speaking, quantization can be seen as a mapping method, which maps floating-point values to integers. When the data is represented as low quantization precision, the amount of memory accesses can be reduced, which has significant impact on the energy consumption. Besides, the computation can be faster due to the reduced quantization precision of the data … As for inference, post quantization with symmetric quantization method usually performs worsen than with asymmetric quantization method. However, we argue that the training-aware quantization can close the gap between symmetric and asymmetric quantization.” Wherein training-aware symmetric quantization is read as a quantization-aware training process)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify Bao/Miles/Courville’s method of pruning and quantizing an inputted neural network to generate a teacher model to train a generated student model to incorporate Xiong’s training-aware quantization. The motivation to do so lies in how “training-aware symmetric quantization is introduced to quantize all of the data types in the training process, which is hardware-friendly while maintaining the training accuracy.” (Xiong [Introduction, Page 62 Paragraph 4]).
Regarding Claim 6,
The combination of Bao/Miles/Courville/Xiong teaches the method of Claim 5 (and thus the rejection of Claim 5 is incorporated). The combination already discloses in which the quantization-aware training process includes uniform symmetric quantization based on a learnable step size. (Xiong [Section 3A Paragraph 2]; “For a clearer description of the quantization, the real floating point number is denoted as r and the corresponding quantized number is denoted as q, then the quantization process can be formulated as:
PNG
media_image8.png
50
193
media_image8.png
Greyscale
where S represents the quantization scale which specifies the step size of the quantizer and Z represents the Zero-point which is used to ensure that zero is quantized without quantization error. In addition, Round is a function to approximate a floating value to the nearest integer value.”
Xiong [Section 3A Paragraph 5]; “Specifically, an additive noise which obeys the uniform distribution is added to the scaled real floating-point number, which can be formulated as:
PNG
media_image9.png
62
459
media_image9.png
Greyscale
Where ϵ∼U(−0.5,0.5) represents uniform distribution.”)
Claims 13-14 recites an apparatus comprising a memory and at least one processor to execute the same method of Claims 5-6. Thus, Claims 13-14 are rejected for reasons set forth in the rejection of Claims 5-6.
Claims 21-22 recites an apparatus comprising means to execute the same method of Claims 5-6. Thus, Claims 21-22 are rejected for reasons set forth in the rejection of Claims 5-6.
Claims 28-29 recites a non-transitory computer readable medium having encoded thereon program code to execute the same method of Claims 5-6. Thus, Claims 28-29 are rejected for reasons set forth in the rejection of Claims 5-6.
Claims 7-8; 15-16; 23-24; 30 are rejected under 35 U.S.C. 103 as being anticipated by Bao et al. (Using Distillation to Improve Network Performance after Pruning and Quantization” [2019], hereinafter “Bao”) in view of Miles et al. (“Cascaded channel pruning using hierarchical self-distillation” [2020], hereinafter “Miles”) in view of Courville et al. (US20220036189A1, hereinafter “Courville”) in view of Li et al. (US11429860B2, hereinafter “Li”).
Regarding Claim 7,
Bao/Miles/Courville teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Bao/Miles/Courville fails to explicitly disclose but Li discloses in which the teacher model is untrained. (Li [Col. 7 Line 52]; “initialization component 124 may create a teacher DNN model (which may be pre-trained), and provide the initialized but untrained teacher DNN model to training component 126 for training … Training component 126 is generally responsible for training the student DNN based on the teacher”)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify Bao/Miles/Courville’s method of pruning and quantizing a neural network for training a pruned and quantized student model with a teacher model to incorporate Li’s usage of an untrained teacher model. The motivation to do so lies in how “An untrained DNN model therefore maybe considered to have a different internal structure than the same DNN model that has been trained” (Li [Col. 9 Line 35]).
Regarding Claim 8,
Bao/Miles/Courville teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Bao/Miles/Courville already discloses in which the trained student model is trained based on a model loss function which includes a cross-entropy loss (Bao [Section 3 Paragraph 4]; “The key of joint training is to continuously optimize distillation loss function, which is defined as
PNG
media_image4.png
21
203
media_image4.png
Greyscale
which represent the cross-entropy of hard target, represent the cross-entropy of soft target, represent probability of studnet model, and represent probability with tempruature of student and teacher model, is a tunable parameter to balance both cross-entropies”).
Bao/Miles/Courville fails to disclose but Li discloses in which the trained student model is trained based on a model loss function which includes … a Kullback-Leibler divergence. (Li [Col. 8 Line 33]; “Evaluating component 128 is generally responsible for evaluating the student DNN model to determine if it is sufficiently trained to approximate the teacher. In particular, in an embodiment, evaluating component 128 evaluates the output distributions of the student and teacher DNNs, determines the difference (which may be determined as an error signal) between the outputs and also determines whether the student is continuing to improve or whether the student is no longer improving (i.e. the student output distribution shows no further trend towards convergence with the teacher output). In one embodiment, evaluating component 128 computes the Kullback-Leibler (KL) divergence between the output distributions)”
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify Bao/Miles/Courville’s method of pruning and quantizing a neural network for training a pruned and quantized student model with a teacher model to incorporate Li’s usage of Kullback-Leibler divergence in training. The motivation to do so lies in how “Using KL divergence to determine an error signal between the teacher and student output distributions provides an advantage over other alternatives such as regression because minimizing the KL divergence is equivalent to minimizing the cross entropy of the distributions” (Li [Col. 15 Line 51]).
Claims 15-16 recite an apparatus comprising a memory and at least one processor to execute the same method of Claims 7-8. Thus, Claims 15-16 are rejected for reasons set forth in the rejection of Claims 7-8.
Claims 23-24 recite an apparatus comprising means to execute the same method of Claims 7-8. Thus, Claims 23-24 are rejected for reasons set forth in the rejection of Claims 7-8.
Claim 30 recites a non-transitory computer readable medium having encoded thereon program code to execute the same method of Claim 8. Thus, Claim 30 is rejected for reasons set forth in the rejection of Claim 8.
Response to Arguments
The Examiner acknowledges the Applicant’s amendments in which Claims 1, 9 and 25 are amended.
Applicant’s arguments filed March 18th, 2026, traversing the rejection of claims 1-30 under 35 U.S.C. § 112(f) have been fully considered, and are fully persuasive.
Applicant’s arguments regarding the 35 U.S.C. § 103 rejections of Claims 1-30 of the previous office action have been considered, but are not fully persuasive.
Examiner believes that the current claim language limitations reciting a preserved pruned set of weights, under broadest reasonable interpretation, is disclosed by the prior art. It is not obvious whether the “pruned set of weights” refers to the subset of weights that were removed or “pruned” from the initial model weights or if it instead refers to the subset of weights remaining after different weights were “pruned” from it. Notably, the recitation of a “pruned set of weights” could fall under both definitions. Similar to how a pruned branch could both refer to a branch that had twigs and leaves pruned from it or a branch that itself was pruned from a tree, the recitation of “preserved pruned set of weights” does not make clear whether the pruned weights are the important weights leftover after the removal of the unimportant weights or the unimportant weights themselves. The conventional understanding one familiar in the art may have of preserved pruned weights in the context of knowledge distillation typically reads on such knowledge distillation utilizing a network that had unimportant weights pruned from it. As such, examiner’s broadest reasonable interpretation of the “preserved pruned weights” which are used in applicant’s system is to read the “preserved pruned set of weights” being simply the neural network after unimportant weights were pruned from it.
Importantly, Examiner notes that clarifying the language surrounding the nature of the “preserved pruned set of weights” through applicant’s specification would significantly change the scope of the claim language, particularly [0024].
The rejection of Claim 1 under 35 U.S.C. § 103 has been maintained. Similarly, the rejection of Claims 9, 17, and 25 under 35 U.S.C. § 103 have been maintained.
The rejection of Claims 2-8, 10-16, 18-24 and 26-30 under 35 U.S.C. § 103, which depend directly or indirectly from Claims 1, 9, 17 and 25 respectively, have been maintained.
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 JONATHAN J KIM whose telephone number is (571)272-0523. The examiner can normally be reached 8-6.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Matt Ell can be reached on (571) 270-3264. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of 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.
/JONATHAN J KIM/Examiner, Art Unit 2141
/MATTHEW ELL/Supervisory Patent Examiner, Art Unit 2141
Prosecution Insights