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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 03/20/2026 has been entered.
Response to Remarks
Claim Rejections – 35 U.S.C. 103
Applicant’s arguments have been fully considered and they are not persuasive.
Applicant argues (pg. 9-10) that the cited references do not teach “in response to selectively identifying the first machine learning model based on the measured performance, creating a shared encoder based on the first encoder” and “communicating a shared encoder based on the first encoder with other participating entities of the plurality of participating entities, the shared encoder configured to be applied by the other participating entities to produce encoded features using their respective training data sets to train their respective machine learning models”. More specifically, Applicant argues that Perepu (and the other cited references) does not teach selectively identifying a model based on measured performance and using the first filter in the global model.
Examiner respectfully disagrees. Kumar teaches the element of identifying a model based on measured performance and using the filter/encoder/update of weight in the global model. See Kumar [¶0031]: “the quality of the central model may be measured as the accuracy against a test set compiled from a selection (e.g., a random selection) of samples from local nodes” Kumar teaches that the measured performances of the models are assessed and compared. Kumar [¶0043]: “the K arms of the CMAB paradigm correspond to the indices of the local nodes. A chosen arm indicates the local node whose update will be used to update the central model” Kumar teaches that the algorithm used has arms that correspond to the local nodes. One of the local models is chosen based on the aforementioned performance and will provide a weight update. Kumar [¶0051]: “Once moderator 106 selects an update to apply, moderator 106 may send … the recommended update to central node or server 104. In response, central node or server 104 may apply the update (at 220), and send the update (at 222 and 224) to local nodes” Kumar teaches that the update from the selected model is sent to the central model, which is then updated at the local nodes, creating a shared encoder.
The foregoing applies to all independent claims and their dependent claims.
Claim Rejections – 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 3-7, 9-13, 15-20 are rejected under 35 U.S.C. 103 as being unpatentable over Kumar et al. (WO 2021070189 A1) hereinafter known as Kumar in view of Perepu et al. (WO 2021064737 A1) hereinafter known as Perepu in view of Daykin et al. (US 20210097381 A1) hereinafter known as Daykin.
Regarding independent claim 1, Kumar teaches:
A system comprising: a processing unit operatively coupled to memory; (Kumar ¶ [0060]: “FIG. 4 is a block diagram of an apparatus 400 (e.g., a user 102 and/or central node or server 104 and/or moderator 106), according to some embodiments. As shown in FIG. 4, the apparatus may comprise: processing circuitry (PC) 402, which may include one or more processors. … memory devices (e.g., random access memory, flash memory) …” Kumar teaches a system that has a processor and memory.)
and an artificial intelligence (AI) platform in communication with the processing unit, the Al platform to apply input encodings in federated learning, the Al platform comprising: (Kumar ¶ [0022]: “Users 104 may register with the central node or server 102 to indicate their willingness to participate in the federated learning of the central model, and may do so continuously or on a rolling basis. Upon registration (and potentially at any time thereafter), the central node or server 102 may select a model type and/or model architecture for the local user 104 to train. The central node or server 102 may transmit an initial model to the users 104.” Kumar teaches a platform that apply input encodings for federated learning. This is because the model type and/or model architecture is an encoding for the input.)
a registration manager configured to arrange a plurality of participating entities in a collaborative relationship to train a single machine learning model, the plurality of participating entities comprising a first participating entity and a second participating entity, each of the plurality of participating entities training a respective machine learning model using a respective encoder and respective training data set, the first participating entity training a first machine learning model with a first encoder on a first training data set to produce first encoded features and the second participating entity training a second machine learning model with a second encoder on a second training data set to produce second encoded features; (Kumar ¶ [0022]: “Upon registration (and potentially at any time thereafter), the central node or server 102 may select a model type and/or model architecture for the local user 104 to train. … The users 104 may train their individual models locally with their own data. The results of such local training may then be reported back to central node or server 102, which may pool the results and update the global model.” Kumar ¶ [0025]: “Local nodes (e.g. users 104) maintain local models… Updates to local models happen according to the frequency of the data that are input to the respective model. Depending upon a user specified strategy, updates are sent to the central node or server 102, mediated by moderator 106. The updates may include the weights of the neural network.” Kumar teaches that there are multiple local nodes that each maintain a model, which is an encoder, and each of these local nodes train the model to get updates. The server arranges this and also gives the model architecture for each node to train. Furthermore, the updates made by each of the nodes are sent to the central node to train a single model.)
an evaluator, operatively coupled to the registration manager, configured to evaluate respective models from the plurality of participating entities, including: measure performance of the first and second machine learning models; (Kumar ¶ [0030]: “The moderator 106 may accordingly schedules the local updates such that the quality of the central model satisfies one or more requirements.” Kumar ¶ [0034]: “In embodiments, requirements may be specified in a language that allows combining predicates on different model properties such as accuracy, norm (e.g., L2- norms) based on the model weights, and other model properties.” Kumar teaches that the performance of the first and second models are measured and then evaluated to determine if they meet a performance requirement and select if this standard is met. Also, the evaluator is operatively coupled to the registration manager because it is the same apparatus.)
and selectively identifying the first machine learning model based on the measured performance; (Kumar ¶ [0030]: “The moderator 106 may accordingly schedules the local updates such that the quality of the central model satisfies one or more requirements.” Kumar ¶ [0034]: “In embodiments, requirements may be specified in a language that allows combining predicates on different model properties such as accuracy, norm (e.g., L2- norms) based on the model weights, and other model properties.” Kumar teaches that the performance of the first and second models are measured and then evaluated to determine if they meet a performance requirement and select if this standard is met.)
and in response to selectively identifying the first machine learning model based on the measured performance, creating a shared encoder based on the first encoder; (Kumar [¶0031]: “the quality of the central model may be measured as the accuracy against a test set compiled from a selection (e.g., a random selection) of samples from local nodes” Kumar teaches that the measured performances of the models are assessed and compared. Kumar [¶0043]: “the K arms of the CMAB paradigm correspond to the indices of the local nodes. A chosen arm indicates the local node whose update will be used to update the central model” Kumar teaches that the algorithm used has arms that correspond to the local nodes. One of the local models is chosen based on the aforementioned performance and will provide a weight update. Kumar [¶0051]: “Once moderator 106 selects an update to apply, moderator 106 may send … the recommended update to central node or server 104. In response, central node or server 104 may apply the update (at 220), and send the update (at 222 and 224) to local nodes” Kumar teaches that the update from the selected model is sent to the central model, which is then updated at the local nodes, creating a shared encoder.)
Kumar does not explicitly teach:
and a director, operatively coupled to the evaluator, configured to communicate the shared encoder to other participating entities of the plurality of participating entities, the shared encoder configured to be applied by the plurality of participating entities to train their respective machine learning models, …
However, Perepu teaches:
and a director, operatively coupled to the evaluator, configured to communicate the shared encoder to other participating entities of the plurality of participating entities, the shared encoder configured to be applied by the other participating entities to produce encoded features using their respective training data sets to train their respective machine learning models, … (Perepu ¶ [0053]: “In some embodiments, one or more individual users 104 having their own individual models of potentially different model type and model architecture may convert (e.g., by distilling) their local model into a distilled model of a specified model type and model architecture. For example, the central node or server 102 may instruct each user about what model type and model architecture the user 104 should distill a model into. The model type will be common to each user 104, but the model architecture may be different in some embodiments.” Perepu teaches a central server that acts as the director and gives each of the entities an encoder to distill the model into. The model architecture is the encoder that is shared among the entities and is common so that, as described above, the single machine learning model can be formed. Perepu ¶ [0057]: “Typically, out of these filters, only a small number (e.g., two or three) of the filters will work well … a global model can then be constructed which takes the filters of each of the different users’ models for each layer and concatenates them … the global model final layer weights may be determined by an averaging” Perepu teaches that in regards to the global model encoder that is given to the entities to distill into, it is formulated by identifying and weighting the various models based on their performances. Perepu teaches that only a small portion of the filters work well, indicating that these are selected based on their performance. Furthermore, Perepu combines these filters of the models of the entities to create an aggregated global model. In determining how the models are given emphasis, a weighting scheme is used to greater weigh the more important and well-performing filters of the models.)
Kumar and Perepu are in the same field of endeavor as the present invention, as the
references are directed to training machine learning models locally and then sharing the models with other entities via federated learning. It would have been obvious, before the effective filing date of the claimed invention, to a person of ordinary skill in the art, to combine the idea of each local entity applying an encoder taught in Kumar with receiving a common encoder from a server taught in Perepu. Perepu provides this additional functionality. As such, it would have been obvious to one of ordinary skill in the art to modify the teachings of Kumar to include teachings of Perepu because the combination would allow for a common encoder to be used among different local entities in the federated learning network to make the models more homogeneous. This has the benefit of potentially regularizing the model to account for differences between testing data from different local entities.
Perepu does not explicitly teach:
… the director further configured to merge models from the plurality of participating entities trained using the shared encoder to form the single machine learning model.
However, Daykin teaches:
… the director further configured to merge models from the plurality of participating entities trained using the shared encoder to form the single machine learning model. (Daykin ¶ [0195]: “Certain embodiments provide a method of training a model, comprising: receiving a first training model generated based on first training data set acquired in first cohort and a second training model generated based on second training data set acquired in second cohort, calculate a first evaluation value and a second evaluation value by inputting data set acquired in the first cohort to the first training model and the second training model.” Daykin ¶ [0196]: “The first training model and the second training model may be synthesized based on the first evaluation value and the second evaluation value.” Daykin teaches that the first and second models can be merged by their evaluation scores.)
Daykin is in the same field of endeavor as the present invention, since it is directed to locally training machine learning models to share and combine with one another. It would have been obvious, before the effective filing date of the claimed invention, to a person of ordinary skill in the art, to combine sharing the machine learning models between nodes in Kumar as modified by Perepu with merging these models once they are shared as taught in Daykin. Daykin provides this additional functionality. As such, it would have been obvious to one of ordinary skill in the art to modify the teachings of Kumar as modified by Perepu to include teachings of Daykin because the combination would allow for a merge of machine learning models from different local nodes. This has the benefit of potentially creating a model that better represents the data in different local nodes.
Regarding dependent claim 3 Kumar, Perepu, and Daykin teach:
The system of claim 1,
Daykin teaches:
wherein the evaluator is configured to test combinations of the first and second machine learning models, (Daykin ¶ [0195]: “Certain embodiments provide a method of training a model, comprising: receiving a first training model generated based on first training data set acquired in first cohort and a second training model generated based on second training data set acquired in second cohort, calculate a first evaluation value and a second evaluation value by inputting data set acquired in the first cohort to the first training model and the second training model.” Daykin ¶ [0196]: “The first training model and the second training model may be synthesized based on the first evaluation value and the second evaluation value.” Daykin teaches that the first and second models can be merged by their evaluation scores. Daykin teaches that after the training is done, the models are tested, so therefore, this resulting model is tested.)
Perepu teaches:
and wherein the selectively identifying at least the first machine learning model includes selectively identifying a combination of the first and second machine learning models. (Perepu ¶ [0053]: “In some embodiments, one or more individual users 104 having their own individual models of potentially different model type and model architecture may convert (e.g., by distilling) their local model into a distilled model of a specified model type and model architecture. For example, the central node or server 102 may instruct each user about what model type and model architecture the user 104 should distill a model into. The model type will be common to each user 104, but the model architecture may be different in some embodiments.” Perepu teaches a central server that acts as the director and gives each of the entities an encoder to distill the model into. In the case when the first and second models use the same encoder, the selective identification, which is a combination of the first and second models, must use a combination of the first and second encoders.)
Daykin is in the same field of endeavor as the present invention, since it is directed to locally training machine learning models to share and combine with one another. It would have been obvious, before the effective filing date of the claimed invention, to a person of ordinary skill in the art, to combine sharing the machine learning models between nodes in Kumar as modified by Perepu with merging and testing these models once they are shared as taught in Daykin. Daykin provides this additional functionality. As such, it would have been obvious to one of ordinary skill in the art to modify the teachings of Kumar as modified by Perepu to include teachings of Daykin because the combination would allow for a merge of machine learning models from different local nodes and subsequent testing of the merged models. This has the benefit of potentially testing and creating a model that better represents the data in different local nodes.
Regarding dependent claim 4 Kumar, Perepu, and Daykin teach:
The system of claim 3,
Perepu teaches:
wherein the evaluator is further configured to create a union of the first and second encoders, wherein the shared encoder is the union of the first and second encoder. (Perepu ¶ [0053]: “In some embodiments, one or more individual users 104 having their own individual models of potentially different model type and model architecture may convert (e.g., by distilling) their local model into a distilled model of a specified model type and model architecture. For example, the central node or server 102 may instruct each user about what model type and model architecture the user 104 should distill a model into. The model type will be common to each user 104, but the model architecture may be different in some embodiments.” Perepu teaches a central server that acts as the director and gives each of the entities an encoder to distill the model into. In the case when the first and second models use the same encoder, the selective identification, which is a union of the first and second models, must use a union of the first and second encoders.)
The motivation to combine are for substantially similar reasons as those of claim 1.
Regarding dependent claim 5 Kumar, Perepu, and Daykin teach:
The system of claim 1,
Daykin teaches:
wherein: the evaluator is configured to share the first and second machine learning models with each of the first and second entities; (Daykin ¶ [0045]: “Certain embodiments provide an apparatus comprising processing circuitry configured to: obtain first trained parameters for a model, wherein the first trained parameters have been generated by training the model using data from a first data cohort; obtain second trained parameters for the model, wherein the second trained parameters have been generated by training the model using data from a second, different data cohort; determine a first evaluation value by inputting data from the first data cohort into a model having the first trained parameters; and determine a second evaluation value by inputting data from the first data cohort into a model having the second trained parameters.” Daykin teaches that the second evaluation values are created by inputting data into a model that was trained at a different local node. This means that the models of the first and second entities were shared with each other.)
the first entity is configured to evaluate the first and second machine learning models with first test data and measure first performance of the first and second models based on the first test data; (Daykin ¶ [0045]: “Certain embodiments provide an apparatus comprising processing circuitry configured to: obtain first trained parameters for a model, wherein the first trained parameters have been generated by training the model using data from a first data cohort; obtain second trained parameters for the model, wherein the second trained parameters have been generated by training the model using data from a second, different data cohort; determine a first evaluation value by inputting data from the first data cohort into a model having the first trained parameters; and determine a second evaluation value by inputting data from the first data cohort into a model having the second trained parameters.” Daykin teaches that the second evaluation values are created by inputting data into a model that was trained at a different local node. This means that the models of the first and second entities were shared with each other. Daykin ¶ [0112]: “At stage 74, the performance scores are used to generate a set of influence values, which may also be referred to as evaluation values.” Daykin teaches that the evaluation value are the performance scores.)
the second entity is configured to evaluate the first and second machine learning models with second test data, different from the first test data, and measure second performance of the first and second models based on the second test data; (Daykin ¶ [0045]: “Certain embodiments provide an apparatus comprising processing circuitry configured to: obtain first trained parameters for a model, wherein the first trained parameters have been generated by training the model using data from a first data cohort; obtain second trained parameters for the model, wherein the second trained parameters have been generated by training the model using data from a second, different data cohort; determine a first evaluation value by inputting data from the first data cohort into a model having the first trained parameters; and determine a second evaluation value by inputting data from the first data cohort into a model having the second trained parameters.” Daykin teaches that the second evaluation values are created by inputting data into a model that was trained at a different local node. This means that the models of the first and second entities were shared with each other. Daykin ¶ [0112]: “At stage 74, the performance scores are used to generate a set of influence values, which may also be referred to as evaluation values.” Daykin teaches that the evaluation value are the performance scores. Each entity calculates a second evaluation score that uses a model different to its own so the terms first and second are interchangeable.)
and the first and second entities are configured to share the first and second measured performance data with the evaluator; (Daykin ¶ [0045]: “Certain embodiments provide an apparatus comprising processing circuitry configured to: obtain first trained parameters for a model, wherein the first trained parameters have been generated by training the model using data from a first data cohort; obtain second trained parameters for the model, wherein the second trained parameters have been generated by training the model using data from a second, different data cohort; determine a first evaluation value by inputting data from the first data cohort into a model having the first trained parameters; and determine a second evaluation value by inputting data from the first data cohort into a model having the second trained parameters.” Daykin teaches that the circuitry gets a first evaluation value. Since the data from each local node is not shared with anyone, it was shared from the local node to the circuit.)
The motivation to combine are for substantially similar reasons as those of claim 1.
Regarding dependent claim 6 Kumar, Perepu, and Daykin teach:
The system of claim 1,
Kumar teaches:
wherein selectively identifying at least the first machine learning model is based on assessing output data for the first machine learning model with respect to a threshold. (Kumar ¶ [0058]: “In some embodiments, the selection criteria includes a rule that during any N updates to the central model, an accuracy of the central model is not changed by more than a threshold amount.” Kumar teaches that the selection of the machine learning model to be added to the updates is based on a threshold on the amount that it would change the accuracy of the central model. Only if the update for the model is less than this threshold is the model selected.)
The motivation to combine are for substantially similar reasons as those of claim 1.
Claim 7 is substantially similar to claim 1 but has the following additional elements:
Regarding independent claim 7, Kumar teaches:
A computer program product to apply input encodings in federated learning, the computer program product comprising a computer readable storage medium having program code embodied therewith, the program code executable by a processor to cause the processor to: (Kumar ¶ [0060]: “FIG. 4 is a block diagram of an apparatus 400 (e.g., a user 102 and/or central node or server 104 and/or moderator 106), according to some embodiments. As shown in FIG. 4, the apparatus may comprise: processing circuitry (PC) 402, which may include one or more processors.” Kumar teaches a system with that has a processor. Kumar ¶ [0022]: “Users 104 may register with the central node or server 102 to indicate their willingness to participate in the federated learning of the central model, and may do so continuously or on a rolling basis. Upon registration (and potentially at any time thereafter), the central node or server 102 may select a model type and/or model architecture for the local user 104 to train. The central node or server 102 may transmit an initial model to the users 104.” Kumar teaches a computer program that applies input encodings for federated learning.)
The motivation to combine are for substantially similar reasons as those of claim 1.
The product dependent claim 9 is rejected for substantially the same reasons as system claim 3.
The product dependent claim 10 is rejected for substantially the same reasons as system claim 4.
The product dependent claim 11 is rejected for substantially the same reasons as system claim 5.
The product dependent claim 12 is rejected for substantially the same reasons as system claim 6.
The product independent claim 13 is rejected for substantially the same reasons as system claim 1.
The method dependent claim 15 is rejected for substantially the same reasons as system claim 3.
The method dependent claim 16 is rejected for substantially the same reasons as system claim 4.
The method dependent claim 17 is rejected for substantially the same reasons as system claim 5.
The method dependent claim 18 is rejected for substantially the same reasons as system claim 6.
Regarding dependent claim 19, Kumar, Perepu, and Daykin teach:
The method of claim 18,
Kumar teaches:
wherein the selectively identifying at least the first machine learning model comprises selectively identifying the second machine learning model based on assessing output data for the second machine learning model with respect to the threshold, … (Kumar ¶ [0058]: “In some embodiments, the selection criteria includes a rule that during any N updates to the central model, an accuracy of the central model is not changed by more than a threshold amount.” Kumar teaches that the selection of the machine learning model to be added to the updates is based on a threshold on the amount that it would change the accuracy of the central model. Only if the update for the model is less than this threshold is the model selected. This criterion is for the both the first and second models.)
Perepu teaches:
… and wherein the shared encoder is a union of the first and second encoders. (Perepu ¶ [0053]: “In some embodiments, one or more individual users 104 having their own individual models of potentially different model type and model architecture may convert (e.g., by distilling) their local model into a distilled model of a specified model type and model architecture. For example, the central node or server 102 may instruct each user about what model type and model architecture the user 104 should distill a model into. The model type will be common to each user 104, but the model architecture may be different in some embodiments.” Perepu teaches a central server that acts as the director and gives each of the entities an encoder to distill the model into. In the case when the first and second models use the same encoder, the selective identification, which is a union of the first and second models, must use a union of the first and second encoders.)
The motivation to combine are for substantially similar reasons as those of claim 1.
The product dependent claim 20 is rejected for substantially the same reasons as system claim 19.
Claims 2, 8, 14 are rejected under 35 U.S.C. 103 as being unpatentable over Kumar in view of Perepu in view of Daykin in view of Hall et al. (WO 2021056043 A1) hereinafter known as Hall.
Regarding dependent claim 2 Kumar, Perepu, and Daykin teach:
The system of claim 1,
Daykin teaches:
wherein at least the first participating entity trains at least two machine learning models, each of the two machine learning models having … and a separate training data set, (Daykin ¶ [0103]: “Three-fold evaluation divides each data cohort into three equal subdivisions (more generally: n-fold would divide into n subdivisions). The models are then trained on two (n-1) of these subdivisions and tested on one (the one not used for training in the case of evaluating at the same institution). Each fold uses a different two training subdivisions and evaluation subdivision.” Daykin teaches a three-fold evaluation that splits the training data into three parts. Two of these parts are used for training, and therefore two models are created using two separate training data sets.)
Kumar teaches:
… having a separate encoder… (Kumar ¶ [0022]: “The users 104 may train their individual models locally with their own data. The results of such local training may then be reported back to central node or server 102, which may pool the results and update the global model.” Kumar ¶ [0025]: “Local nodes (e.g. users 104) maintain local models, which may in some embodiments be neural networks with identical architecture. Updates to local models happen according to the frequency of the data that are input to the respective model. Depending upon a user specified strategy, updates are sent to the central node or server 102, mediated by moderator 106. The updates may include the weights of the neural network.” Kumar teaches that there are multiple local nodes that each maintain a model, which is an encoder, and each of these local nodes train the model to get updates.)
Kumar, Perepu, and Daykin do not teach:
and the first participating entity is configured to measure performance of the at least two machine learning models on first test data, and selectively identify and share with the evaluator one of the at least two machine learning models based on their measured performance.
However, Hall teaches:
and the first participating entity is configured to measure performance of the at least two machine learning models on first test data, and selectively identify and share with the evaluator one of the at least two machine learning models based on their measured performance. (Hall ¶ [0042]: “training a plurality of Student models, wherein each Student model is a Teacher model at a first node … and training the Student model using the Teacher model at the node using the node dataset” Hall teaches that the first node, or entity, trains the at least two models on its own data set. Hall ¶ [00121]: “These may include an Average Voting method, a weighted averaging method, a Mixture of Experts Layers (or learned weighting) method, or even using a further distillation method where a final model is distilled from the plurality of student models using simple distillation or modified decentralised distillation.” Hall teaches that the performance of the models is evaluated, with examples of methods such as the Average Voting method. After the models are evaluated, they are selected via weighting.)
Hall is in the same field of endeavor as the present invention, since it is directed to distributing the evaluation of the performance of a machine learning model to another entity. It would have been obvious, before the effective filing date of the claimed invention, to a person of ordinary skill in the art, to combine evaluating machine learning models in different nodes in Kumar as modified by Perepu as modified by Daykin with a single node measuring the performance of the models from other nodes as taught in Hall. Hall provides this additional functionality. As such, it would have been obvious to one of ordinary skill in the art to modify the teachings of Kumar as modified by Perepu as modified by Daykin to include teachings of Hall because the combination would allow one node to measure performance of the machine learning model for subsequent evaluation by another entity. This has the benefit of being able to have a common dataset that each model is measured against, which may improve the objectivity in the performance measurements of these models.
The product dependent claim 8 is rejected for substantially the same reasons as system claim 2.
The method dependent claim 14 is rejected for substantially the same reasons as system claim 2.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KYU HYUNG HAN whose telephone number is (703) 756-5529. The examiner can normally be reached on MF 9-5.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Alexey Shmatov can be reached on (571) 270-3428. 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.
/Kyu Hyung Han/
Examiner
Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123
Prosecution Insights