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 regarding the 103 rejection have been fully considered but are moot in light of a new rejection.
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.
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 1-5, 7, 10-14, 17,19-20, 21, 24-25 and 27-28 are rejected under 35 U.S.C. 103 as being unpatentable over Liu et al. “ASKs: Convolution with any-shape kernels for efficient neural networks” (as cited by the IDS on 01/06/2023) in view of Kundu, Souvik, et al. "Pre-defined sparsity for low-complexity convolutional neural networks.".
As per claim 1, Liu discloses a method for processing data using a convolutional neural network, comprising:
receiving an input data patch (Introduction, "we propose to use efficient convolution kernels with irregular shapes to build compact CNNs" Section 3.2, "Given the input data, X ∈ ℝc x h x w where c is the number of input channels, h and w are the height and width of the input data,". Please note the input data corresponds to Applicant’s an input data patch); and
generating an output of the convolutional neural network (Y = X * w +b, - where Y € R’*"*“is the output feature map with n channels – e.g. Section 3.2 and output Y under Algorithm 1) based on processing the input data patch with a shaped kernel of the convolutional neural network (Section 3.2, (see Fig. 5(a)) "Fig. 5. Convolution with cross-shaped kernel and decomposed calculation of cross-shaped convolution." Please note a cross-shaped kernel corresponds to Applicant’s a shaped kernel.)
Although Liu et al. further teaches wherein the shaped kernel comprises a partially fixed kernel including: a set of learned weights in other elements of the shaped kernel, the set of learned weights having values learned during the training of the shaped kernel (e.g. Table 1, using candidate ASKs, weights, ACC% represents the accuracy on the test set), Liu et al. does not explicitly disclose a predefined fixed weight in a center element of the shaped kernel
However, Kundu teaches a predefined fixed weight in a center element of the shaped kernel, the predefined fixed weight remaining fixed during training of the shaped kernel (pg. 4
PNG
media_image1.png
524
740
media_image1.png
Greyscale
” We say a 3D filter of size k × k × Ci has pre-defined sparsity if some of the k2 × Ci parameters are fixed to be zero before training and held fixed throughout training and inference”)
Liu as well as Kundu are directed towards training convolutional neural network. Therefore, it would have been obvious before the effective filing date of the claimed invention to combine the teachings of Liu with the teachings of Kudu by having partially fixed weights in order to reduce workload imbalance.. This motivation for combination also applies to the remaining claims which depend on this combination.
As per claim 2, Liu et al. teaches claim limitation as applied above in claim 1. Liu further teaches wherein: the shaped kernel is associated with a layer of a convolutional neural network model, (Liu Section 4, "In this section, we conduct experiments on benchmarks to evaluate the performance of [any-shape kernels] ASKs. We first introduce the datasets and experimental setup. Then, we replace regular convolution kernels with the candidate ASKs" (Section 4.1) " we evaluate ASKs for both single-branch networks (VGGNets [41]) and multiple-branch networks (ResNets [42] and DenseNet [43])" In other words, the convolutional layers of the evaluated CNNs (VGGNets, ResNets and DenseNet) have kernel replaced with ASKs. Therefore, it is interpreted that a convolutional layer with ASKs corresponds to a shaped kernel associated with a layer of a CNN.) and the input data patch comprises input data element values generated, at least in part, by a square convolution kernel of a preceding layer of the convolutional neural network model.(Liu Introduction, "We evaluate our method by replacing all the regular 3 x 3 kernels in VGGNets and ResNets with high-efficiency ASKs." Please note the shaped convolution kernels consist of 3 x 3 squares as depicted in figure 1 of Liu. However, the shapes are sparse, meaning that only a portion of the square contains weights. Therefore, the evaluated convolutional neural networks which use 3 x 3 ASK kernels, would comprise generation of data patches using square convolutional kernels.).
As per claim 3, Liu et al. teaches claim limitation as applied above in claim 1. Liu further teaches wherein the shaped kernel comprises a cruciform kernel.(Liu (See Fig. 5(a)) Section 3.2, "Fig. 5. Convolution with cross-shaped kernel and decomposed calculation of cross-shaped convolution." Please note a cross-shaped kernel corresponds to Applicant’s cruciform kernel.).
As per claim 4, Liu et al. discloses the method as applied above in claim 1.
Kundu further teaches wherein the partially fixed kernel comprises a partially fixed [cruciform] kernel (§3.1 ¶1 “We say a 3D filter of size k × k × Ci has pre-defined sparsity if some of the k2 × Ci parameters are fixed to be zero before training and held fixed throughout training and inference”)
Liu teaches cruciform kernel and wherein the other elements of the shaped kernel are branch elements of the partially fixed cruciform kernel (pg. 3
PNG
media_image2.png
278
1136
media_image2.png
Greyscale
C and D show cruciform kernels with branch elements)
As per claim 13, Liu teaches the method of claim 10. Kundu teaches wherein the partially fixed kernel comprises a partially fixed cruciform kernel and wherein the other elements of the shaped kernel are branch elements of the partially fixed cruciform kernel. (see mapping of claim 4)
As per claim 5, Liu et al. teaches claim limitation as applied above in claim 1. Liu further discloses wherein: the input data patch comprises a set of m input data elements, the shaped kernel comprises a set of n weight elements, n < m, (Liu Section 3.2, (see Fig. 5(a)) "Fig. 5. Convolution with cross-shaped kernel and decomposed calculation of cross-shaped convolution." In other words, the cross-shaped kernel corresponds to a set of weight element less than the number of input data elements of input data patch depicted in figure 5.) and processing the input data patch with the shaped kernel comprises processing n input data elements of the input data patch with n corresponding elements of the shaped kernel. (Liu Section 3.2, "Specifically, assume the number of weights in an ASK is s, we take the c input channels as a whole, shift s copies of it in s required directions, and concat them to form input channels. These input channels are then convolved by pointwise kernels to get an output." Please note the processing of the input data with shaped kernel to generate convolution output corresponds to process of using ASKs within a convolutional neural network to generate an output.).
As per claim 7, Liu et al. teaches claim limitation as applied above in claim 5. Liu further teaches wherein n is an even multiple of four. (Liu, See Fig. 1 (e) "ASK_4a". Please note Liu’s the use of a shaped kernel where number of weights is 4 in ASK_4a.).
As per claim 10, Liu discloses a method for training a convolutional neural network, comprising:
receiving an input data patch associated with a target label (Introduction, "we propose to use efficient convolution kernels with irregular shapes to build compact CNNs" Section 3.2, "Given the input data, X ∈ ℝc x h x w where c is the number of input channels, h and w are the height and width of the input data,". Please note the input data corresponds to Applicant’s an input data patch); and
generating an output of the convolutional neural network (Y = X * w +b, - where Y € R’*"*“is the output feature map with n channels – e.g. Section 3.2 and output Y under Algorithm 1) based in part on processing the input data patch using a shaped kernel of the convolutional neural network, wherein the shaped kernel comprises a partially fixed kernel (Section 3.2, (see Fig. 5(a)) "Fig. 5. Convolution with cross-shaped kernel and decomposed calculation of cross-shaped convolution." Please note a cross-shaped kernel corresponds to Applicant’s a shaped kernel.)
computing a loss based on the generated output and the target label;(Section 4.1, "we train for 160 epochs and use Stochastic Gradient Descent (SGD) optimization algorithm.." and (Section 3.2) "Since we do not introduce any non-differentiable operations or additional parameters, the backpropagation of loss can be solved automatically according to the original chain rule." In other words, Stochastic Gradient Descent involves adjusting parameters to minimize loss, which is calculated based on target label of the supervised dataset. Therefore, Liu teaches computing a loss using target labels to refine parameters (weights).)
a set of learnable weights in other elements of the shaped kernel (e.g. Table 1, using candidate ASKs, weights, ACC% represents the accuracy on the test set),
and refining one or more of the other elements of the shaped kernel based on the loss such that the set of learnable weights have values learned during training of the shaped kernel (Section 4.1, "we train for 160 epochs and use Stochastic Gradient Descent (SGD) optimization algorithm.." In other words, Liu teaches the use of Stochastic Gradient Descent to optimize weights of the CNN, which is based on the calculated loss.) such that the set of learnable weights have values learned during training of the shaped kernel (e.g. Table 1, using candidate ASKs, weights, ACC% represents the accuracy on the test set),
Liu does not explicitly disclose a predefined fixed weight in a center element of the shaped kernel
However, Kundu teaches a predefined fixed weight in a center element of the shaped kernel, the predefined fixed weight remaining fixed during training of the shaped kernel (pg. 4
PNG
media_image1.png
524
740
media_image1.png
Greyscale
” We say a 3D filter of size k × k × Ci has pre-defined sparsity if some of the k2 × Ci parameters are fixed to be zero before training and held fixed throughout training and inference”)
Liu as well as Kundu are directed towards training convolutional neural network. Therefore, it would have been obvious before the effective filing date of the claimed invention to combine the teachings of Liu with the teachings of Kudu by having partially fixed weights in order to reduce workload imbalance.. This motivation for combination also applies to the remaining claims which depend on this combination.
As per claim 11, Liu et al. teaches claim limitation as applied above in claim 10.
Liu further teaches wherein: the shaped kernel is associated with an internal layer of a convolutional neural network model, (Liu Section 4.3, "We replace all the regular kernels in...ResNets…with ASKs" and "we evaluate ASKs for...multiple-branch networks...ResNets [42]...on several benchmark visual datasets" The examiner notes that the reference cited by Liu regarding ResNets, He et al. "Deep Residual Learning for Image Recognition" (as cited by the IDS on states that "Residual Networks. Next we evaluate 18-layer and 34-layer residual nets (ResNets). The baseline architectures are the same as the above plain nets, expect that a shortcut connection is added to each pair of 3×3 filters as in Fig. 3 (right)." Therefore, Liu teaches shaped kernels associated with an internal layer of a convolution neural network, as all kernels are replaced including internal layers.) and the input data patch comprises input data element values generated, at least in part, by a square convolution kernel of a preceding layer of the convolutional neural network model.(Liu Introduction, "We evaluate our method by replacing all the regular 3 x 3 kernels in VGGNets and ResNets with high-efficiency ASKs." In other words, the shaped convolution kernels consist of 3 x 3 squares as depicted in figure 1 of Liu. However, the shapes are sparse, meaning that only a portion of the square contains weights. Therefore, the evaluated convolutional neural networks which use 3 x 3 ASK kernels, would comprise generation of data patches using square convolutional kernels.).
As per claim 12, Liu et al. teaches claim limitation as applied above in claim 10.
Liu further teaches wherein the shaped kernel comprises a cruciform kernel. (Liu (See Fig. 5(a)) Section 3.2, "Fig. 5. Convolution with cross-shaped kernel and decomposed calculation of cross-shaped convolution." Please note a cross-shaped kernel corresponds to Applicant’s a cruciform kernel).
As per claim 14, Liu et al. teaches claim limitation as applied above in claim 10. Liu further teaches wherein a number of other elements in the shaped kernel is an even multiple of four (Liu See Fig. 1 (e) "ASK_4a". Please note Liu disclosed the use of a shaped kernel where number of weights is 4 in ASK_4a.).
As per claims 17, 18, 20, and 21, Liu – Kundu et al. recite a system that implements the methods of claims 1, 3, 5, 7 with substantially the same limitations, respectively. Therefore, the rejection applied to claims 1, 3, 5, 7 also applies to claims 17, 18, 20, 21.
In addition, Claims 17, 18, 20, 21 recite a processing system, comprising: a memory comprising computer-executable instructions; one or more processors configured to execute the computer-executable instructions and cause the processing system to perform an operation comprising (The examiner notes that generic computing hardware--including a processor and a memory--is inherent throughout the Liu disclosure.) substantially the same limitations, respectively, as the method of claims 1, 3, 5 and 7.
As per claims 24, 25, 26, 28, Liu – Kundu et al. recite a system that implements the methods of claims recite a system that implements the methods of claims 10, 11, 12, 14 with substantially the same limitations, respectively. Therefore, the rejection applied to claims 10, 11, 12, 14 also applies to claims 24, 25, 26, 28.
In addition, Claims 24, 25, 26, 28 recite A processing system, comprising: a memory comprising computer-executable instructions; one or more processors configured to execute the computer-executable instructions and cause the processing system to perform an operation comprising (The examiner notes that generic computing hardware--including a processor and a memory--is inherent throughout the Liu disclosure.) substantially the same limitations, respectively, as the method of claims 10, 11, 12, 14.
As per claims 19 and 27, they recite systems that implement the methods of claims 4, 13 with substantially the same limitations, respectively. Therefore, the rejection applied to claims 4, 13 also applies to claims 19, 27.
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Liu – Kundu et al. and further in view of Weng et al. “A Quality-Oriented Reconfigurable Convolution Engine Using Cross-Shaped Sparse Kernels for Highly-Parallel CNN Acceleration”.
As per claim 6, Liu further teaches wherein processing n elements of the set of m input data elements of the input data patch with n corresponding elements of the shaped kernel comprises: performing an elementwise multiplication between n - 1 input data elements and n - 1 weight elements; (Liu Section 3.2, "we take the c input channels as a whole, shift...concat them to form...These input channels are then convolved by pointwise kernels to get an output." The examiner notes the term comprises is an inclusive term that does not exclude additional elementwise multiplications in addition to the n - 1 input data and weight elements. Therefore, the multiplication occurring between kernel and input data, as depicted in figure 5 of Liu, would is interpreted to correspond to elementwise multiplication between n-1 input data elements and n-1 elements.) and refraining from processing a center weight element using multiplication ( Liu Section 3.2 . – “We calculate the mean relative value of parameters in each group in different layers to denote their contributions…mean relative value of parameters in center. Please note examiner interprets this claim limitation as instead of multiplying the center element, one can skip it or apply a different operation such as calculate the mean of the center weight as taught in Liu Section 3.2)
However, Liu – Kundu et al. doesn't explicitly teach and processing the center weight element with a skip connection.
Weng, in the same field of endeavor, teaches and processing a center weight element with a skip connection.(Section IV, "For the proposed method, we replace dense kernels with center-sharing cross-shaped sparse kernels in the models...We remove [batch normalization] BN layers and add a skip connection to improve training stability" In other words, Weng teaches replacing a batch normalization layer with a skip connection. In the combination of Liu and Weng, the input data patch and weight elements are processed along with a skip connection in the activation layer of the convolutional neural network. Therefore, a center weight (among the plurality of weights) is processed alongside a skip connection.).
Liu – Kundu et al. as well as Weng are directed towards training convolutional neural network. Therefore, Liu – Kundu et al. as well as Weng are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Liu – Kundu et al. with the teachings of Weng by adding a skip connection. Weng provides as additional motivation for combination “add a skip connection to improve training stability.” In other words, it would have been obvious to add a skip connection to improve training performance. This motivation for combination also applies to the remaining claims which depend on this combination.
Claims 8, 15, 22 and 29 are rejected under 35 U.S.C. 103 as being unpatentable over Liu – Kundu et al. and further in view of Lee et al. “Efficient SIMD Implementation for Accelerating Convolutional Neural Network”.
As per claim 8, Liu-Kundu et al. teaches the method as applied above in claim 5. However, Liu – Kundu et al. doesn't explicitly teach wherein processing the n input data elements of the input data patch with the n corresponding elements of the shaped kernel comprises using single instruction, multiple data (SIMD) operations to apply multiple weight elements in parallel.
Lee, in the same field of endeavor, teaches wherein processing the n input data elements of the input data patch with the n corresponding elements of the shaped kernel comprises using single instruction, multiple data (SIMD) operations to apply multiple weight elements in parallel. (Section 5, "it is important to accelerate CNN by only the CPU efficiently. In this paper, we propose a method to accelerate CNN by using the Single Instruction Multiple Data (SIMD) unit" and (Introduction) "SIMD takes advantage of data parallelism and it has been widely used in signal and image processing." In other words, in the combination of Liu – Kundu et al. and Lee, SIMD would be used to accelerate the multiplication operations completed while processing input data.).
Liu – Kundu et al. as well as Lee are directed towards Convolutional neural networks. Therefore, Liu as well as Lee are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Liu-Kundu et al. with the teachings of Lee by performing the processing of input data using single instruction, multiple data operations (SIMD). In other words, it would have been obvious to use SIMD operations which are “widely used in signal and image processing” in order to “accelerate CNN…using…(SIMD).” Thereby, decreasing the computation resources need to implement the CNN. This motivation for combination also applies to the remaining claims which depend on this combination.
As per claim 15, Liu – Kundu et al. teaches the method of claim 10. However, Liu – Kundu et al. doesn't explicitly teach wherein refining the one or more weight elements comprises using single instruction, multiple data (SIMD) operations to refine multiple weight elements in parallel.
Lee, in the same field of endeavor, teaches wherein refining the one or more weight elements comprises using single instruction, multiple data (SIMD) operations to refine multiple weight elements in parallel.(Section 5, "it is important to accelerate CNN by only the CPU efficiently. In this paper, we propose a method to accelerate CNN by using the Single Instruction Multiple Data (SIMD) unit" and (Introduction) "SIMD takes advantage of data parallelism and it has been widely used in signal and image processing. ").
Liu – Kundu et al. as well as Lee are directed towards Convolutional neural networks. Therefore, Liu- Kundu et al. as well as Lee are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Liu- Kundu et al. with the teachings of Lee by performing the processing of input data using single instruction, multiple data operations (SIMD). In other words, it would have been obvious to use SIMD operations which are “widely used in signal and image processing” in order to “accelerate CNN…using…(SIMD).” Thereby, decreasing the computation resources need to implement the CNN. This motivation for combination also applies to the remaining claims which depend on this combination.
Claims 22 and 29 recite systems that implement the methods of claims 8, 15 with substantially the same limitations, respectively. Therefore, the rejection applied to claims 8, 15 also applies to claims 22, 29.
Claims 9, 16, 23 and 30 are rejected under 35 U.S.C. 103 as being unpatentable over Liu – Kundu et al. and further in view of Purdum “Introduction to pointers”.
As per claim 9, Liu – Kundu et al. teaches the method of claim 1. However, Liu – Kundu et al. doesn't explicitly teach retrieving a first set of weight elements using one or more pointers; incrementing the one or more pointers using one or more fixed offsets; and retrieving a second set of weight elements using the one or more pointers..
Purdum, in the same field of endeavor, teaches retrieving a first set of weight elements using one or more pointers; ((Purdum, Exercises, Answer 2) "Pointers allow functions to have direct access to data that would otherwise be out of scope...Pointers also allows arrays to be passed to functions in a more memory-efficient manner than pass-by-value would permit." Listing 8-5 depicts incrementally displaying, using pointers and offsets, values stored in the array displayed in Figure 8-8 on page 184. In other words, Purdum teaches retrieving a set of values from an array (weight elements) using pointers.) incrementing the one or more pointers using one or more fixed offsets; ((Purdum, page 185) "it [the method for displaying array values via pointers using offsets] also works just fine because all pointer math is also scaled to fit the underlying data type. In this case, any increment increases the offset by 2 because the scalar is 2 (each int requires two bytes of memory)." In other words, Purdum teaches access values of an array (set of elements) via pointers by incrementing by fixed offsets based on the data type size.) and retrieving a second set of weight elements using the one or more pointers.(Listing 8-5 depicts incrementally displaying, using pointers and offsets, values stored in the array displayed in Figure 8-8 on page 184. In other words, in the combination of Liu and Purdum, both sets of weight elements are retrieved from using pointers and offsets.).
Liu - Kundu et al. as well as Purdum are directed towards retrieving elements from memory. Therefore, Liu – Kundu et al. as well as Purdum are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Liu with the teachings of Purdum by using pointers and offsets to retrieve stored value. Purdum provides as additional motivation for combination "Pointers also allows arrays to be passed to functions in a more memory-efficient manner than pass-by-value would permit.” In other words, it would have been obvious to use pointers and fixed offsets to retrieve weight elements as pointers enable more efficient use of computation resources.
As per claim 16, Liu – Kundu et al. teaches the method of as applied above in claim 10.
However, Liu – Kundu et al. doesn't explicitly teach wherein refining the one or more weight elements comprises: retrieving a first set of weight elements using one or more pointers;
incrementing the one or more pointers using one or more fixed offsets; and retrieving a second set of weight elements using the one or more pointers.
Purdum, in the same field of endeavor, teaches wherein refining the one or more weight elements comprises: retrieving a first set of weight elements using one or more pointers; ((Purdum, Exercises, Answer 2) "Pointers allow functions to have direct access to data that would otherwise be out of scope...Pointers also allows arrays to be passed to functions in a more memory-efficient manner than pass-by-value would permit." Listing 8-5 depicts incrementally displaying, using pointers and offsets, values stored in the array displayed in Figure 8-8 on page 184. In other words, Purdum teaches retrieving a set of values from an array (weight elements) using pointers.) incrementing the one or more pointers using one or more fixed offsets; ((Purdum, page 185) "it [the method for displaying array values via pointers using offsets] also works just fine because all pointer math is also scaled to fit the underlying data type . In this case, any increment increases the offset by 2 because the scalar is 2 (each int requires two bytes of memory)." In other words, purdum teaches access values of an array (set of elements) via pointers by incrementing by fixed offsets based on the data type size.) and retrieving a second set of weight elements using the one or more pointers.(Listing 8-5 depicts incrementally displaying, using pointers and offsets, values stored in the array displayed in Figure 8-8 on page 184. In other words, in the combination of Liu and Purdum, both sets of weight elements are retrieved from using pointers and offsets.).
Liu – Kundu et al. as well as Purdum are directed towards retrieving elements from memory. Therefore, Liu – Kundu et al. as well as Purdum are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Liu with the teachings of Purdum by using pointers and offsets to retrieve stored value. Purdum provides as additional motivation for combination "Pointers also allows arrays to be passed to functions in a more memory-efficient manner than pass-by-value would permit.” In other words, it would have been obvious to use pointers and fixed offsets to retrieve weight elements as pointers enable more efficient use of computation resources.
Claims 23 and 30 recite systems that implement the methods of claims 9, 16 with substantially the same limitations, respectively. Therefore, the rejection applied to claims 9, 16 also applies to claims 23, 30.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Kundu, Souvik, et al. "pSConv: A pre-defined sparse kernel based convolution for deep CNNs." 2019 57th Annual Allerton Conference on Communication, Control, and Computing (Allerton). IEEE, 2019.
THIS ACTION IS MADE FINAL. 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 KEVIN W FIGUEROA whose telephone number is (571)272-4623. The examiner can normally be reached Monday-Friday, 10AM-6PM 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, MIRANDA HUANG can be reached at (571)270-7092. 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.
KEVIN W FIGUEROA
Primary Examiner
Art Unit 2124
/Kevin W Figueroa/Primary Examiner, Art Unit 2124
Prosecution Insights