Detailed Action
This Office Action is in response to the remarks entered on 06/03/2026. Amended claims 1, 6 and 11 have been entered. Claims 1-15 are presently pending.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 112
Amended claims were received and entered on 06/03/2026. 35 U.S.C. 112(b) rejections have been withdrawn.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 rejection have been withdrawn.
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 1-15 are rejected under 35 U.S.C. 103 as being unpatentable over Veshchikov et al. (US 11270227 B2, hereinafter ‘Veshchikov’) in view of Chen et al. (US 9939792 B2, hereinafter ‘Chen’) in view of Amini (US 20190261243 A1, hereinafter ‘Amini’).
Regarding claim 1, Veshchikov teaches:
A method for dynamically generating a plurality of machine learning models for processing a [Veshchikov, Abstract] The machine learning system generates a plurality of machine learning models using a plurality of subsets of data elements)
receiving a [Veshchikov, Fig. 4, block 72] The training data elements are stored in a database)
analyzing, by a machine learning engine, at least one characteristic of the [Veshchikov, Fig. 4, block 72] and [col 4, line 44 – col 5, line 11] discloses partitioning the dataset and assigning records having the same hash value to the same subset)
generating, by the machine learning engine, a first machine learning model for processing the The limitation and spec broadly recite ‘generating’ process and it allows broad interpretation of ‘generating’ process, as ‘generating a machine learning model’ encompasses create, initialize, train, combining, selecting, configuring a model. [Veshchikov, col 3, line 62 – col 4, line 20] The training environment for providing training data to a plurality of model training portions 22, 24 and 26 are coupled to partitioned data, wherein the training environment includes resources for training a ML model with training data and training algorithm that can be applied to the training data. This process corresponds to configuration of a machine learning model)
training, by the machine learning engine, the first machine learning model; ([Veshchikov, Fig. 4, block 74] and [col 7, lines 31-55] A plurality of machine learning models (including the first machine learning model) are trained with assigned subsets of data elements)
directing, by the machine learning engine, the first machine learning model to generate a first output by processing [Veshchikov, Fig. 4, block 76 and 78] and [col 7, lines 31-55] The plurality of machine learning model (including the first model) provide outputs to an aggregator, and then a subset to be changed is selected)
generating, by the machine learning engine, a second machine learning model based upon the at least one characteristic of the the at least one characteristic of the , wherein the second machine learning model is generated in a distinct form from the first machine learning model by building a model ensemble; and (The instant specification [0018] discloses that the model may receive new information that allows the predictive model to learn and change over time e.g., to improve its prediction accuracy by receiving back new results). Veshchikov performs the same process.
[Veshchikov, Fig. 4, blocks 78, 80 and 82] and [col 7, lines 31-55] After training the first machine learning model in Step 74, the system selects an assigned subset to be changed or deleted in Step 78. [col 5, line 39 – col 6, line 16] when any associated machine learning model that was trained using the assigned subset of data that should be changed or deleted, a new machine learning model (i.e., the second machine learning model) is generated using the changed subset. The machine learning system (includes the first model) continues to perform inference operations while the new machine learning model (i.e., the second machine learning model) is being trained. [col 6, lines 10-16] Other machine learning models (i.e., the first machine learning model) in the machine learning system may be used (execution of the first model) while the new model (i.e., the second machine learning model) is being trained with the modified data subset)
directing, by the machine learning engine, the second machine learning model to generate at least a second output by processing ([Veshchikov, Fig. 4, block 82]. [col 6, lines 14-16] and [col 7, lines 31-55] The retrained model is placed back into inference operation in machine learning system 10 to generate further inference results)
Veshchikov does not specifically disclose:
processing user-specified data set and a user-specified task; receiving a user-specified data set and a user-specified task;
analyzing, by a machine learning engine, at least one characteristic of the user-specified data set and at least one characteristic of the user- specified task;
selecting, by the machine learning engine, a plurality of encoders based upon the at least one characteristic of the user-specified data set and at least one characteristic of the user-specified task;
directing, by the machine learning engine, each of the selected plurality of encoders to encode the received user-specified data set;
wherein generating the first machine learning model is based upon the at least one characteristic of the user-specified data set and the at least one characteristic of the user-specified task;
directing, by the machine learning engine, to generate output by processing the at least one encoding of the user- specified data set;
Chen teaches:
receiving a [Chen, col 4, lines 24-41] The controller forwards both the task and the data to the neural network and trains the neural network based on the task feature and the data feature)
analyzing, by a machine learning engine, at least one characteristic of the [Chen, col 6, lines 40-47] discloses input feature and task feature information is forwarded to learning module 620, which contains a neural network. [Chen, col 6, lines 51-67] The artificial neural network analyzes the features)
wherein generating the first machine learning model is based upon the at least one characteristic of the task; ([Chen, col 4, lines 24-41] The controller forwards both the task and the data to the neural network and trains the neural network based on the task feature and the data feature. [Chen, col 2, lines 18-23] teaches there are a plurality of neural network in the learning model which includes the first and the second neural network)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Veshchikov and Chen to use the method of using both task data and data set of Chen to implement the machine learning system of Veshchikov. The suggestion and/or motivation to do so is to improve the performance of the machine learning system by introducing task-specific mode selection of [Chen, col 3, lines 42-54] thereby improving task-specific (subset) performance of each machine learning algorithm in a plurality of machine learning models [Veshchikov, line 26-45].
However, Chen does not specifically disclose:
processing user-specified data set; receiving a user-specified data set and a user-specified task;
selecting, by the machine learning engine, a plurality of encoders based upon the at least one characteristic of the user-specified data set and at least one characteristic of the user-specified task;
directing, by the machine learning engine, each of the selected plurality of encoders to encode the received user-specified data set;
directing, by the machine learning engine, to generate output by processing the at least one encoding of the user- specified data set;
Amini teaches:
processing user-specified data set; receiving a user-specified data set and a user-specified task; ([Amini, 0068] A user may specify the type of scene a camera is intended to capture (i.e., the user-specified task) and input information to the system indicating where the camera is installed and input indicative of deployment characteristics of the camera system (i.e., the user specified data set), and the user may also update entered information by providing new inputs via a similar interface during operation of the camera system. [Amini, 0070] The received encoded data are provided to the machine learning techniques such as deep learning and neural networks to detect physical objects in the captured video)
selecting, by the machine learning engine, a plurality of encoders based upon the at least one characteristic of the user-specified data set and at least one characteristic of the user-specified task; ([Amini, 0075] discloses machine learning techniques (i.e., the machine learning engine) can be implemented to select a channel from a set of available channels based on how certain combinations of network conditions and encoding schemes impact a quality level of a resulting video stream. [Amini, 0061 and 0067] collectively discloses that an encoder is selected to encode video captured by camera 610a and encoder parameters are important to selecting an appropriate channel [Amini, 0068] A user may specify the type of scene a camera is intended to capture (i.e., the user-specified task) and input information to the system indicating where the camera is installed and input indicative of deployment characteristics of the camera system (i.e., the user specified data set), and the user may also update entered information by providing new inputs via a similar interface during operation of the camera system. [Amini, 0070] The received encoded data are provided to the machine learning techniques such as deep learning and neural networks to detect physical objects in the captured video)
directing, by the machine learning engine, each of the selected plurality of encoders to encode the received user-specified data set; ([Amini, 0075] discloses machine learning techniques (i.e., the machine learning engine) can be implemented to select a channel from a set of available channels based on how certain combinations of network conditions and encoding schemes impact a quality level of a resulting video stream. [Amini, 0061 and 0067] collectively discloses that an encoder is selected to encode video captured by camera 610a and encoder parameters are important to selecting an appropriate channel. [Amini, 0070] The received encoded data are provided to the machine learning techniques such as deep learning and neural networks to detect physical objects in the captured video)
directing, by the machine learning engine, to generate output by processing the at least one encoding of the user- specified data set; ([Amini, 0075] discloses machine learning techniques (i.e., the machine learning engine) can be implemented to select a channel from a set of available channels based on how certain combinations of network conditions and encoding schemes impact a quality level of a resulting video stream. [Amini, 0061 and 0067] collectively discloses that an encoder is selected to encode video captured by camera 610a and encoder parameters are important to selecting an appropriate channel. [Amini, 0070] The received encoded data are provided to the machine learning techniques such as deep learning and neural networks to detect physical objects in the captured video)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Veshchikov, Chen and Amini to use the method of selecting an encoder to encode the data of Amini to implement the machine learning system of Veshchikov. The suggestion and/or motivation to do so is to improve the performance of the system, as compressing the data using a selected encoder (an encoder specializing in encoding specific data) reduces the amount of data processed by each of the machine learning model.
Regarding claim 6, Veshchikov teaches:
A non-transitory, computer-readable medium comprising instructions tangibly stored on the non-transitory computer- readable medium, wherein the instructions are executable by at least one processor to perform a method for dynamically generating a plurality of machine learning models for processing a user-specified data set, the method comprising ([Veshchikov, col 6, line 38 – col 7, line 7] discloses that the data processing system is implemented using processor, memory and user interface).
Claim 6 is a computer readable medium claim which implements the same features as the method claim 1 and is rejected for at least the same reasons.
Regarding claim 2, Veshchikov teaches:
wherein generating the first machine learning model further comprises generating a neural network. (The limitation and spec broadly recite ‘generating’ process and it allows broad interpretation of ‘generating’ process, as ‘generating a machine learning model’ encompasses create, initialize, train, combining, selecting, configuring a model. [Veshchikov, col 3, line 62 – col 4, line 43] The training environment for providing training data to a plurality of model training portions 22, 24 and 26 are coupled to partitioned data, wherein the training environment includes resources for training a ML model with training data and training algorithm that can be applied to the training data. This process corresponds to configuration of a machine learning model. The machine learning model may include neural networks)
Claim 7 is a computer readable medium claim which implements the same features as the method claim 2 and is rejected for at least the same reasons.
Regarding claim 3, Veshchikov teaches:
wherein generating the second machine learning model further comprises generating a neural network. (The limitation and spec broadly recite ‘generating’ process and it allows broad interpretation of ‘generating’ process, as ‘generating a machine learning model’ encompasses create, initialize, train, combining, selecting, configuring a model. [Veshchikov, col 3, line 62 – col 4, line 43] The training environment for providing training data to a plurality of model training portions 22, 24 and 26 are coupled to partitioned data, wherein the training environment includes resources for training a ML model with training data and training algorithm that can be applied to the training data. This process corresponds to configuration of a machine learning model. The machine learning model may include neural networks)
Claim 8 is a computer readable medium claim which implements the same features as the method claim 3 and is rejected for at least the same reasons.
Regarding claim 4, Veshchikov teaches:
further comprising providing, by the machine learning engine, access to at least one output selected from the group consisting of the first output and the second output. ([Veshchikov, col 5, line 12-40] After the training phase of machine learning system 10 is complete, the models analyze the input and outputs are computed. The selected outputs of the plurality of models are provided to aggregator 34)
Claim 9 is a computer readable medium claim which implements the same features as the method claim 4 and is rejected for at least the same reasons.
Regarding claim 5, Veshchikov in view of Chen teaches:
The method of claim 1, further comprising directing, by the machine learning engine, the second machine learning model to determine a residual of the first output. ([Chen, col 5, line 44-54] The output of the controller, which contains the machine learning models, is being accessed and the difference between a target output and a current output is calculated)
Claim 10 is a computer readable medium claim which implements the same features as the method claim 5 and is rejected for at least the same reasons.
Regarding claim 11, Veshchikov in view of Chen in view of Amini teaches:
A machine learning system, comprising: a user interface, configured to select: a user-specified data set; and a user-specified task; ([Amini, 0068] A user may specify the type of scene a camera is intended to capture (i.e., the user-specified task) and input information to the system indicating where the camera is installed and input indicative of deployment characteristics of the camera system (i.e., the user specified data set), and the user may also update entered information by providing new inputs via a similar interface (i.e., the user interface) during operation of the camera system. [Amini, 0070] The received encoded data are provided to the machine learning techniques such as deep learning and neural networks to detect physical objects in the captured video)
Claim 11 is a machine learning system claim which implements the same features as the method claim 1 and is rejected for at least the same reasons.
Claim 12 is a machine learning system claim which implements the same features as the method claim 2 and is rejected for at least the same reasons.
Claim 13 is a machine learning system claim which implements the same features as the method claim 3 and is rejected for at least the same reasons.
Claim 14 is a machine learning system claim which implements the same features as the method claim 4 and is rejected for at least the same reasons.
Claim 15 is a machine learning system claim which implements the same features as the method claim 5 and is rejected for at least the same reasons.
Response to Arguments
Response to Arguments under 35 U.S.C. 112(b)
Amended claims were received and entered on 06/03/2026. 35 U.S.C. 112(b) rejections have been withdrawn.
Response to Arguments under 35 U.S.C. 101
Applicant’s arguments, see [Remarks, pages 2-5], filed 06/03/2026, with respect to 35 U.S.C. 101 rejections have been fully considered and are persuasive. The 35 U.S.C. 101 rejections of 1-15 have been withdrawn.
Response to Arguments under 35 U.S.C. 103
Arguments: Applicant asserts that (a) switching from a sequential execution model to a parallel execution mode is not “generating a second machine learning model”, (b) the execution mode of Chen is fundamentally different from a machine learning model, a way of running a program task, not a trained model that processes data to produce outputs, and (c) Chen describes switching execution modes for the same task, not generating a new ML model while a first ML model is executing [Remarks, page 6].
Examiner’s Response: Applicant’s arguments with respect to claims 1-15 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US-20180046942-A1 (This prior art is pertinent because it discloses generating a machine learning model while using another machine learning model to process input data)
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 JUN KWON whose telephone number is (571)272-2072. The examiner can normally be reached Monday – Friday 8:00AM – 5:00PM ET.
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, Abdullah Kawsar can be reached at (571)270-3169. 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.
/JUN KWON/Examiner, Art Unit 2127
/ABDULLAH AL KAWSAR/Supervisory Patent Examiner, Art Unit 2127
Prosecution Insights