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 .
Detailed Action
This action is in response to the claims filed 8/14/2024:
Claims 21 – 34 are pending.
Claims 21 and 32 are independent.
Specification
The disclosure is objected to because of the following informalities:
The title of the invention is not descriptive. A new title is required that is clearly indicative of the invention to which the claims are directed.
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.
Claims 24, 27, 28, and 34 are 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.
Regarding claim 24, “an interval of an input value on a positive side further than a predetermined input value is larger than 0, and an interval of an input value on a negative side further than the predetermined input value is 0 or asymptotic to 0.” is grammatically indefinite. The claim appears to attach the properties “larger than 0,” and “asymptotic to 0” to the interval rather than the value which does not make sense mathematically. In the interest of further examination the claim limitation is interpreted as “a value of the differential function of the activation function is greater than 0 over an interval of input values greater than a predetermined input value, and is 0 over an interval of input values less than the predetermined input value.”
Regarding claim 34, “the activation function” lacks antecedent basis. Claim 33 recites “activation functions” such that it would be unclear which of the plural activation functions was “the activation function.” “An activation function” is recommended.
Claims 27 and 28 are rejected with respect to their dependence on claim 24.
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 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 21 and 32 are rejected under U.S.C. §103 as being unpatentable over the combination of Koivisto (US11995551B2) and Wang (US20190080238A1).
Regarding claim 21, Koivisto teaches A learning method of optimizing a neural network implemented by an information processing device including one or more hardware processors, the method comprising:([Col. 3 l. 12-20] "a training application 140 resides in the memory 116 included in the compute instance 110(1) and executes on the processor(s) 112 included in the compute instance 110(1). The training application 140 generates a trained neural network 190 based on a source neural network 130.")
by the one or more hardware processors, acquiring configuration information of the neural network;([Col. 7 l. 42-53] "a training data store (not shown) stores training data and parameters related to training and/or pruning the neural networks […] The parameters may include, but are not limited to, the number of layers, the number of filters per layer, the number of kernels per filter, the number of training iterations, the number of hidden layers, the learning rate, etc.")
acquiring a plurality of pieces of training information including an input vector and a target vector serving as a target of an output vector;([Col. 7 l. 42-53] "a training data store (not shown) stores training data and parameters related to training and/or pruning the neural networks" [Col. 5 l. 29] "L(x, y, W) = LD(x, y, W) + λR(W)" x interpreted as training input vector, y interpreted as target of output vector)
executing a first processing of repeating executing a forward direction process, generating an error vector, executing a reverse direction process, ([Col. 3 l. 30-40] "Each of the convolutional layers (not shown in FIG. 1) converts a three-dimensional (3D) input volume into a 3D output volume using any number of three-dimensional (3D) filters" [Col. 5 l. 9-15] "the training engine 160 implements any number and type of machine learning techniques (e.g., gradient descent) with regularization to train the source neural network 130" [Col. 7 l. 42-53] "the number of training iterations" [Col. 21 l. 45-55] "the blocks may sometimes be executed in the reverse order" [Col. 5 l. 29] "L(x, y, W) = LD(x, y, W) + λR(W)" Koivisto describes ordinary iterative gradient based neural network training. Gradient descent is interpreted as a reverse direction process relying on the loss error vectors L(x, y, W) = LD(x, y, W) + AR(W). Koivisto iteratively performs gradient descent with the forward direction convolution process)
and updating each of a plurality of weight coefficients a predetermined number of times;([Col. 5 l. 9-15] "the training engine 160 implements any number and type of machine learning techniques (e.g., gradient descent) with regularization to train the source neural network 130" [Col. 7 l. 42-53] "the number of training iterations")
in the forward direction process, assigning the input vector to an input layer of the neural network, causing operation data to be propagated in a forward direction, and causing the output vector to be output from an output layer;([Col. 3 l. 30-40] "Each of the convolutional layers (not shown in FIG. 1) converts a three-dimensional (3D) input volume into a 3D output volume using any number of three-dimensional (3D) filters")
in the generating of the error vector, generating the error vector based on the output vector and the target vector;([Col. 5 l. 9-15] "the training engine 160 implements any number and type of machine learning techniques (e.g., gradient descent) with regularization to train the source neural network 130" [Col. 7 l. 42-53] "the number of training iterations")
in the reverse direction process, assigning the error vector to the output layer of the neural network and causing error data to be propagated in a reverse direction;([Col. 3 l. 30-40] "Each of the convolutional layers (not shown in FIG. 1) converts a three-dimensional (3D) input volume into a 3D output volume using any number of three-dimensional (3D) filters" [Col. 5 l. 9-15] "the training engine 160 implements any number and type of machine learning techniques (e.g., gradient descent) with regularization to train the source neural network 130" [Col. 7 l. 42-53] "the number of training iterations" [Col. 21 l. 45-55] "the blocks may sometimes be executed in the reverse order" [Col. 5 l. 29] "L(x, y, W) = LD(x, y, W) + λR(W)" Koivisto describes ordinary iterative gradient based neural network training. Gradient descent is interpreted as a reverse direction process relying on the loss error vectors L(x, y, W) = LD(x, y, W) + AR(W). Koivisto iteratively performs gradient descent with the forward direction convolution process)
in the updating of each of the plurality of weight coefficients, updating each of the plurality of weight coefficients included in the neural network such that an objective function obtained by adding a basic loss function and an L2 regularization term multiplied by a regularization strength is minimized;([Col. 5 l. 29-45] "an overall loss term L(x,y,W) [...] L(x, y, W) = LD(x, y, W) + λR(W) […] λ is the associated regularization parameter, and the regularization loss term R(W) is based on the p-norm of the weights (e.g., p is equal to 1 for L1 norm, p is equal to 2 for L2 norm, etc.)." Koivisto is explicit that the objective is a loss function which is minimized)
after executing the first processing: specifying an inactive node and an inactive channel among a plurality of nodes and a plurality of channels included in the neural network; deleting the inactive node and the inactive channel from the neural network;([Col. 3 l. 60-67] "the filter pruning engine 170 identifies one or more filters in the intermediate neural network 162 having average magnitudes lower than a pruning threshold 164, and generates the pruned neural network 162 that omits the identified filters" [Col. 7 l. 1-20] "the filter pruning engine 170 removes each filter included in the pruning list from the intermediate neural network 162 […] More precisely, the filter pruning engine 170 removes the kernels that operate on the channel corresponding to the pruned filter from the intermediate neural network 162" Koivisto explicitly deletes an inactive (low-magnitude filter) and is explicit that this also results in deleting the corresponding output channel)
wherein the method further comprises: in a first iteration of the first processing, setting the regularization strength such that the regularization strength increases as a target deletion ratio increases, and([Col. 5 l. 39-55] "the training engine 160 dynamically varies the regularization parameter(s) during training based on a complexity reduction target 154. The training engine 160 may acquire any type of complexity reduction target 154 and vary the regularization parameter(s) in any technically feasible fashion. For instance, in some embodiments, the training engine 160 sets the complexity reduction target 154 equal to a target number of filters to be pruned per layer, a target pruning ratio" [Col. 4 l. 62-67] "as persons skilled in the art will recognize, increasing the level of regularization typically increases the aggressiveness with which the filter pruning engine 170 removes filters from the intermediate neural network 162" Koivisto's target pruning ratio is interpreted as synonymous with a target deletion ratio. Koivisto describes increasing regularization strength when a greater target pruning ratio is sought)
in a second or subsequent iteration of the first processing, changing the regularization strength to be smaller than the regularization strength in the first iteration.([Col. 7 l. 20-35] "the regularization implemented by the re-training engine 180 may be weaker than the regularization implemented by the training engine 160").
However, Koivisto does not explicitly teach determining whether or not a size of the neural network from which the inactive node and the inactive channel are deleted is a target size or less after the inactive node and the inactive channel are deleted; completing the first processing and outputting the configuration information of the neural network when the size of the neural network is the target size or less; and re-executing the first processing when the size of the neural network is not the target size or less.
Wang, in the same field of endeavor, teaches determining whether or not a size of the neural network from which the inactive node and the inactive channel are deleted is a target size or less after the inactive node and the inactive channel are deleted; completing the first processing and outputting the configuration information of the neural network when the size of the neural network is the target size or less; and re-executing the first processing when the size of the neural network is not the target size or less, ([¶0024] "The method 100 may include evaluating whether the pruning ratio P is equal to a target pruning ratio PT. Example pruning target ratios may be in the range of 50 and 90 percent. However, what is a reasonable target pruning ratio is highly dependent on the application. If not, the pruning ratio P may be increased to be P+ΔP, where ΔP as an increment amount, or set to PT, whichever is smaller. The increment ΔP may be constant or may vary with each iteration of step 112 according to a non-linear function. The method 100 may then continue at step 104 by pruning the CNN model again with P equal to the value set at step 112. The method may then continue from step 104 as described above").
Koivisto as well as Wang are directed towards pruning convolutional neural networks. Therefore, Koivisto as well as Wang 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 Koivisto with the teachings of Wang by repeating the pruning process until a threshold is reached. Wang provides as additional motivation for combination ([¶0003] “Multi-iteration pruning and retraining usually can help improve accuracy”).
Regarding claim 32, claim 32 is directed towards a device for performing the method of claim 21. Therefore, the rejection applied to claim 21 also applies to claim 32.
Claims 22-27, 30, 33, and 34 are rejected under U.S.C. §103 as being unpatentable over the combination of Koivisto and Wang and in further view of Bansal (“Minnorm training: an algorithm for training over-parameterized deep neural networks”, 2018).
Regarding claim 22, the combination of Koivisto and Wang teaches The method according to claim 21.
However, the combination of Koivisto and Wang doesn't explicitly teach wherein, for each of the plurality of weight coefficients, in the updating, a gradient is calculated based on the objective function, a step width is calculated based on the gradient and a corresponding past gradient, and the plurality of weight coefficients are updated based on the calculated step width such that the objective function is decreased.
Bansal, in the same field of endeavor, teaches for each of the plurality of weight coefficients, in the updating, a gradient is calculated based on the objective function, a step width is calculated based on the gradient and a corresponding past gradient, and the plurality of weight coefficients are updated based on the calculated step width such that the objective function is decreased.([p. 4 §2] "We perform a pair of iterative updates similar to the dual-ascent algorithm to optimize for the network weights...where η is the step size for weight updates. The above equation corresponds to gradient descent on the Lagrangian and can be applied to arbitrary deep networks via standard automatic differentiation. This weight update step can be replaced by multiple gradient steps or alternative neural network training approaches such as momentum [28] or Adam [15]. The second step is the update of the Lagrange multipliers with step size s" Step width interpreted as synonymous with step size.).
The combination of Koivisto and Wang as well as Bansal are directed towards pruning neural networks. Therefore, the combination of Koivisto and Wang as well as Bansal 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 the combination of Koivisto and Wang with the teachings of Bansal by using Bansal’s training optimizations (such as ADAM) and any of Bansal’s activation functions. Bansal provides as additional motivation for combination ([p. 4 §2] “This weight update step can be replaced by multiple gradient steps or alternative neural network training approaches such as momentum [28] or Adam [15]”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 23, the combination of Koivisto, Wang, and Bansal teaches The method according to claim 22, wherein an activation function including an interval of an input value at which a differential function becomes 0 or an interval of an input value at which the differential function is asymptotic to 0 is set in the neural network.(Bansal [p. 8 §3.3.1] "We consider a shallow neural network with weight vector W = [w1w2] and sigmoidal output...where y˜µ = yµ/2 + 1/2 is the output label {−1, 1} recoded to {0, 1}. Differentiating with respect to the weights yields the gradient descent update" Bansal explicitly teaches that the derivative of the sigmoid activation function is mapped from 0 to 1.).
Regarding claim 24, the combination of Koivisto, Wang, and Bansal teaches The method according to claim 23, wherein, in the differential function of the activation function, an interval of an input value on a positive side further than a predetermined input value is larger than 0, and an interval of an input value on a negative side further than the predetermined input value is 0 or asymptotic to 0.(Bansal [p. 8 §3.3.1] "We consider a shallow neural network with weight vector W = [w1w2] and sigmoidal output...where y˜µ = yµ/2 + 1/2 is the output label {−1, 1} recoded to {0, 1}. Differentiating with respect to the weights yields the gradient descent update" Bansal explicitly teaches that the derivative of the sigmoid activation function is mapped from 0 to 1.).
Regarding claim 25, the combination of Koivisto, Wang, and Bansal teaches The method according to claim 23, wherein activation functions set in all nodes and channels included in all intermediate layers of the neural network are identical to one another.(Bansal [p. 11 §4] "We train a fully connected network with 2 hidden layers and 800 hidden units with ReLU activations on 50K examples for each algorithm").
Regarding claim 26, the combination of Koivisto, Wang, and Bansal teaches The method according to claim 22, wherein, in the specifying, a node and a channel for which norms of weight vectors are a predetermined threshold value or less are specified as the inactive node and the inactive channel.(Koivisto [Col. 3 l. 60-67] "the filter pruning engine 170 identifies one or more filters in the intermediate neural network 162 having average magnitudes lower than a pruning threshold 164, and generates the pruned neural network 162 that omits the identified filters" [Col. 7 l. 1-20] "the filter pruning engine 170 removes each filter included in the pruning list from the intermediate neural network 162 […] More precisely, the filter pruning engine 170 removes the kernels that operate on the channel corresponding to the pruned filter from the intermediate neural network 162" [Col. 6 l. 1-3] "the filter pruning engine 170 sets the average magnitude for a given filter equal to the L2 norm of the filter. Koivisto explicitly deletes an inactive (low-magnitude filter) and is explicit that this also results in deleting the corresponding output channel).
Regarding claim 27, the combination of Koivisto, Wang, and Bansal teaches The method according to claim 24, wherein the activation function is ReLU.(Bansal [p. 11 §4] "We train a fully connected network with 2 hidden layers and 800 hidden units with ReLU activations on 50K examples for each algorithm").
Regarding claim 30, the combination of Koivisto, Wang, and Bansal teaches The method according to claim 22, wherein, in the updating, the weight coefficient is updated by an algorithm of Adaptive Moment Estimation (Adam).(Bansal [p. 4 §2] "This weight update step can be replaced by multiple gradient steps or alternative neural network training approaches such as momentum [28] or Adam [15].").
Regarding claim 33, the combination of Koivisto and Wang teaches the one or more processors specifies a node and a channel for which norms of weight vectors are a predetermined threshold value or less as the inactive node and the inactive channel.(Koivisto [Col. 3 l. 60-67] "the filter pruning engine 170 identifies one or more filters in the intermediate neural network 162 having average magnitudes lower than a pruning threshold 164, and generates the pruned neural network 162 that omits the identified filters" [Col. 7 l. 1-20] "the filter pruning engine 170 removes each filter included in the pruning list from the intermediate neural network 162 […] More precisely, the filter pruning engine 170 removes the kernels that operate on the channel corresponding to the pruned filter from the intermediate neural network 162" [Col. 6 l. 1-3] "the filter pruning engine 170 sets the average magnitude for a given filter equal to the L2 norm of the filter. Koivisto explicitly deletes an inactive (low-magnitude filter) and is explicit that this also results in deleting the corresponding output channel).
However, the combination of Koivisto and Wang doesn't explicitly teach The device according to claim 32, wherein activation functions set in all nodes and channels included in all intermediate layers of the neural network include an interval of an input value at which a differential function becomes 0 or an interval of an input value at which the differential function is asymptotic to 0,
for each of the plurality of weight coefficients, the one or more processors calculates a gradient based on the objective function, calculates a step width based on the gradient and a corresponding past gradient, and updates the plurality of weight coefficients based on the calculated step width such that the objective function is decreased, and.
Bansal, in the same field of endeavor, teaches The device according to claim 32, wherein activation functions set in all nodes and channels included in all intermediate layers of the neural network include an interval of an input value at which a differential function becomes 0 or an interval of an input value at which the differential function is asymptotic to 0,([p. 8 §3.3.1] "We consider a shallow neural network with weight vector W = [w1w2] and sigmoidal output...where y˜µ = yµ/2 + 1/2 is the output label {−1, 1} recoded to {0, 1}. Differentiating with respect to the weights yields the gradient descent update" Bansal explicitly teaches that the derivative of the sigmoid activation function is mapped from 0 to 1.)
for each of the plurality of weight coefficients, the one or more processors calculates a gradient based on the objective function, calculates a step width based on the gradient and a corresponding past gradient, and updates the plurality of weight coefficients based on the calculated step width such that the objective function is decreased, and([p. 4 §2] "We perform a pair of iterative updates similar to the dual-ascent algorithm to optimize for the network weights...where η is the step size for weight updates. The above equation corresponds to gradient descent on the Lagrangian and can be applied to arbitrary deep networks via standard automatic differentiation. This weight update step can be replaced by multiple gradient steps or alternative neural network training approaches such as momentum [28] or Adam [15]. The second step is the update of the Lagrange multipliers with step size s" Step width interpreted as synonymous with step size.).
The combination of Koivisto and Wang as well as Bansal are directed towards pruning neural networks. Therefore, the combination of Koivisto and Wang as well as Bansal 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 the combination of Koivisto and Wang with the teachings of Bansal by using Bansal’s training optimizations (such as ADAM) and any of Bansal’s activation functions. Bansal provides as additional motivation for combination ([p. 4 §2] “This weight update step can be replaced by multiple gradient steps or alternative neural network training approaches such as momentum [28] or Adam [15]”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 34, the combination of Koivisto, Wang, and Bansal teaches The device according to claim 33, wherein the activation function is ReLU, and(Bansal [p. 11 §4] "We train a fully connected network with 2 hidden layers and 800 hidden units with ReLU activations on 50K examples for each algorithm")
the one or more processors update the plurality of weight coefficients in accordance with an optimization algorithm of Adaptive Moment Estimation (Adam).(Bansal [p. 4 §2] "This weight update step can be replaced by multiple gradient steps or alternative neural network training approaches such as momentum [28] or Adam [15].").
Claim 28 is rejected under U.S.C. §103 as being unpatentable over the combination of Koivisto, Wang, and Bansal and in further view of Lai (“PRUNE: Preserving Proximity and Global Ranking for Network Embedding”, 2017).
Regarding claim 28, the combination of Koivisto, Wang, and Bansal teaches The method according to claim 24.
However, the combination of Koivisto, Wang, and Bansal doesn't explicitly teach, wherein the activation function is ELU.
Lai, in the same field of endeavor, teaches The method according to claim 24, wherein the activation function is ELU ([p. 7 §4.1] "Model Setup. For all experiments, our model fixes node embedding and hidden layers to be 128- dimensional, proximity representation to be 64-dimensional. Exponential Linear Unit (ELU) [4] activation is adopted in hidden layers for faster learning").
The combination of Koivisto, Wang, and Bansal as well as Lai are directed towards pruning neural networks. Therefore, the combination of Koivisto, Wang, and Bansal as well as Lai 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 the combination of Koivisto, Wang, and Bansal with the teachings of Lai by using an exponential linear unit. Lai provides as a motivation for combination ([p. 7 §4.1] " Exponential Linear Unit (ELU) [4] activation is adopted in hidden layers for faster learning").
Claims 29 and 31 are rejected under U.S.C. §103 as being unpatentable over the combination of Koivisto, Wang, Bansal, and Theodorakopoulos (US20180137417A1).
Regarding claim 29, the combination of Koivisto, Wang, and Bansal teaches The method according to claim 23.
However, the combination of Koivisto, Wang, and Bansal doesn't explicitly teach, wherein the activation function is hyperbolic tangent.
Theodorakopoulos, in the same field of endeavor, teaches the activation function is hyperbolic tangent([¶0040] "Directly after the convolutions an additive bias and nonlinearity (sigmoidal, hyperbolic tangent etc.) or a rectified linear unit (RELU, leaky RELU etc.) is applied to each feature map (34, 39 in FIG. 1).").
The combination of Koivisto, Wang, and Bansal as well as Theodorakopoulos are directed towards pruning neural networks. Therefore, the combination of Koivisto, Wang, and Bansal as well as Theodorakopoulos 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 the combination of Koivisto, Wang, and Bansal with the teachings of Theodorakopoulos by using an htan activation function. One of ordinary skill in the art would recognize that this is a routine design choice as reinforced by Theodorakopoulos who explicitly discloses that it provides nonlinearity ([¶0040] "Directly after the convolutions an additive bias and nonlinearity (sigmoidal, hyperbolic tangent etc.)”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 31, the combination of Koivisto, Wang, and Bansal teaches The method according to claim 22.
However, the combination of Koivisto, Wang, and Bansal doesn't explicitly teach, wherein, in the updating, the weight coefficient is updated by an algorithm of RMSprop.
Theodorakopoulos, in the same field of endeavor, teaches in the updating, the weight coefficient is updated by an algorithm of RMSprop. ([¶0077] "using in one embodiment a back-propagation algorithm (e.g Stochastic Gradient Descend, AdaDelta, Adaptive Gradient, Adam, Nesterov's Accelerated Gradient, RMSprop etc.)").
The combination of Koivisto, Wang, and Bansal as well as Theodorakopoulos are directed towards pruning neural networks. Therefore, the combination of Koivisto, Wang, and Bansal as well as Theodorakopoulos 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 the combination of Koivisto, Wang, and Bansal with the teachings of Theodorakopoulos by using RMSProp. One of ordinary skill in the art would recognize that this is a routine design choice that would lead to obvious and expected results as reinforced by Theodorakopoulos. This motivation for combination also applies to the remaining claims which depend on this combination.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Wang (“Structured Probabilistic Pruning for Convolutional Neural Network Acceleration”, 2017) is directed towards channel-pruning convolutional neural networks trained and using a pruning ratio for retraining with ADAM and Relu activation.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SIDNEY VINCENT BOSTWICK whose telephone number is (571)272-4720. The examiner can normally be reached M-F 7:30am-5:00pm 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 on (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.
/SIDNEY VINCENT BOSTWICK/Examiner, Art Unit 2124
