DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claims 1, 3-4, 6, 8, and 10 -15 is pending and claim 1 is an independent claim.
Response to Arguments
Applicant’s arguments, see Arguments page 6, filed on 07/21/2026, with respect to 35 USC § 101 claim rejections have been fully considered and are persuasive. The 35 USC § 101 claim rejections of claim 15 have been withdrawn.
Applicant’s arguments, see Arguments pages 6-12, filed on 07/21/2026, with respect to 35 USC § 101 claim rejections have been fully considered and are persuasive. The 35 USC § 101 claim rejections of claims 1, 3-4, 6, 8, and 10-15 have been withdrawn.
Applicant’s arguments with respect to claim(s) 1, 3-4, 6, 8, and 10-15, see Arguments pages 12-16, filed on 07/21/2026, with respect to 35 USC § 103 claim rejections have been fully considered but they are not persuasive.
The Applicant argues that Johnsson is directed to "selecting a machine learning model for execution in a resource constrained environment." The Applicant further argues that the cited paragraph 0006 of Steelberg merely discloses a method for classifying a media segment using classification neural networks and reference Penha was cited by the Examiner only for the "voice call" element and does not remedy this deficiency (Arguments, pages 13-14).
The Examiner respectfully disagrees. Johnsson teaches on how to select a highest ranked ML model. Steelberg, on para 0006, provides systems and methods for classifying a media file (e.g., audio, video, multimedia file) and one of the methods for classifying a first media segment of a first data type having a corresponding media segment of a second data type includes: extracting a first set of media features of the first media segment of the first data type; generating, using an engine prediction neural network, a best candidate neural network based at least on the first set of media features; determining whether a predicted value of accuracy of the best candidate neural network is above a predetermined accuracy threshold; when the predicted value of accuracy of the best candidate neural network is below the predetermined accuracy threshold, classifying the corresponding media segment of a second data type using a second classification neural network; and selecting, based at least on results of the classification of the corresponding media segment of a second data type, a third classification neural network to classify the first media segment of the first data type … Thus, the Examiner disagrees with the Applicant’s argument and characterization stating that the cited paragraph of Steelberg merely discloses a method for classifying a media segment. The cited Steelberg paragraph not only discloses a media file, but it discloses it in detail and in depth, with various features and predicted value of accuracy. Moreover, Steelberg discloses computational task that comprises a speech-to-text conversion and the at least one minimum performance threshold that comprises a maximum word error rate, since it teaches that, in transcription, the number of outputs is effectively reduced to a single number (the word error rate) per engine per audio segment. It teaches that the engine prediction neural network can be trained to predict the best-candidate engine by associating dominant features (e.g., weights of a layer) of an audio segment to an accuracy rating (e.g., word error rate) of an engine). Additionally, the Steelberg reference improves the Johnsson reference because it would have enabled to be on a more competitive edge in the current business climate because of the processing and analyzing of all sorts of big data (i.e., structured, unstructured, and semi-structured data) using neural networks to discover the market trends, customer behaviors, and other useful indicators relating to their markets, product, and/or services. Penha was used for its provision of the various embodiments for a real-time call monitoring machine learning framework 100 that uses several different sources of data for the voice call in predicting whether the party is likely to drop the voice call before solving a query. In Penha the different sources of data may include the audio 110 of the voice call, but “voice call” was not the reference provides (Johnsson, 8th page, 5th para; Steelberg, para 00114, 0054, 0004, 0006; Penha, para 0034 and Figure 1).
The Applicant argues that none of the cited references teach or suggest selecting a machine learning model that meets a minimum performance threshold for the speech-to-text conversion (i.e., "a maximum word error rate" as recited by claim 1) where that threshold is itself determined based on a complexity indicator of the computational task, thereby enabling adaptive behavior depending on task complexity (e.g., simple vs. complex utterances in the voice call) (Arguments, pages 14-16).
The Examiner respectfully disagrees. Johnsson, in Fig. 1 d, displays the apparatus 104, that may comprise an arrangement as depicted to select a machine learning model M to be deployed in an execution environment 102 having resource constraints 302… The apparatus 104 may further comprise a memory 142 comprising one or more memory units to store data on. The memory 142 comprises instructions executable by the processor. The memory 412 is arranged to be used to store e.g. measurements, photos, location information, ML models, metadata, instructions, configurations and applications to perform the methods herein when being executed by the processing unit 147. Johnsson also discloses some tasks that have certain latency requirements in the range of 50 ps to 200 ms. Thus, it is also essential to consider the latency requirements of the task while selecting the associated ML model. Moreover, Steelberg discloses a second classification engine to classify the alternate data set where the second engine (machine learning model) has the same/closest performance as the first classification engine implemented on the alternate data set (Johnsson, 5th page, 5th para; Johnsson, 2nd page, 3rd para; Steelberg, para 0015)
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.
Claims 1-2, 6, 8, 10 and 12-15 are rejected under 35 U.S.C. 103 as being unpatentable over Johnsson et al. Pat App No. WO 2022161644 A1 (Johnsson) in view of Steelberg Pat App No. US 20200066278 A1 (Steelberg), and further in view of Penha Pat App No. US 20220239775 A1 (Penha).
Regarding Claim 1, Johnsson discloses a computer-implemented (Johnsson, 4th page, 6th para, a server, a computer, or any computing device) method for performing a computational task using a machine learning model (Johnsson, 2nd page, 3rd para, each task could be having a plurality of trained ML models), the method comprising:
obtaining an indication about a computational task to be performed (Johnsson, 3rd page, 2nd para, receive a request for a machine learning model solving a task);
obtaining at least one minimum performance threshold for the computational task (Johnsson, 7th page, 1st para, the ML models 108 fulfill the latency requirements defined by the task T).
obtaining a plurality of machine learning models (Johnsson, 2nd page, 8th para, retrieving, from a model store, a first set of machine learning models; OR, 2nd page, 3rd para, having a plurality of trained ML models), wherein each machine learning model in the plurality of machine learning models is associated with at least one performance indicator and at least one resource consumption indicator, for the computational task to be performed (Johnsson, 2nd page, 8th para – 3rd page, 1st para, retrieving, from a model store, a first set of machine learning models that solves the task T using at least a subset of features F… at least one suitable machine learning model to be deployed, wherein the determining is based on the calculated complexity and the resource constraints of the execution environment; OR, Johnsson, 6th page, 5th para, Each of the above-mentioned tasks would have a set of ML models with different accuracies, hyperparameters, complexities, feature sets, data sampling requirements, hardware requirements and software requirements);
choosing a machine learning model out of the plurality of machine learning models based at least on the at least one minimum performance threshold for the computational task (Johnsson, 7th page, 1st para, check whether the first set of the ML models 108 fulfill the latency requirements defined by the task T), the at least one performance indicator of each machine learning model in the plurality of machine learning models (Johnsson, 7th page, 9th para, assigns a rank to each ML model in the second set of ML models (or suitable ML models) based on their historical predictive performance) and the at least one resource consumption indicator of each machine learning model in the plurality of machine learning models (Johnsson, 7th page, 8th para, the first set of machine learning models 108. The resource shortage function is trained based on the calculated complexity and resource constraints), wherein the choosing the machine learning model out of the plurality of machine learning models comprises choosing, from among machine learning models that meet the at least one minimum performance threshold, a machine learning model that meets the at least one minimum performance threshold and minimizes resource consumption (Johnsson, 6th page, 6th para – 7th page, 8th para, Action 202: In this action, the apparatus 104 retrieves a first set of machine learning (ML) models 108 associated with the task T using at least a subset of features F. In order to retrieve the first set of ML model, the apparatus 104 transmits a request to the model store 106 to determine if ML models associated with the task T and using the feature set F or subset of features (from properties PTI, PT2,... Pin) exists therein. In response to the request, the model store searches for ML models solving task T having the feature set F or the subset of the features… the apparatus 104 may also check whether the first set of the ML models 108 fulfill the latency requirements defined by the task T. Action 203: In this action, the apparatus 104 determines a complexity (Ci) for each ML model (Mi where i =1 to n)) in the first set of ML models 108… Action 204: In this action, the apparatus 104 requests resource constraints from the execution environment 102. The resource constraints 302 comprise at least one of hardware constraints, software constraints, sampling requirements and resource usage of the execution environment. Action 205: In this action, the apparatus 104 determines from the first set of machine learning models 108 a second set of machine learning model 306 with at least one suitable machine learning model that can be deployed. The determining is performed based on the calculated complexity and the resource constraints 302 received from the execution environment 102; [i.e., apparatus 104 determines and retrieves /i.e., chooses/ if ML model(s) associated for task T exists, and also the same apparatus 104 checks if it fulfils the latency requirements, i.e., meeting the at least one minimum performance threshold, and the same apparatus also determines whether resource usage, constraints and execution environment, i.e., consumption resource, is minimized]); and
performing the computational task using the chosen machine learning model (Johnsson, 8th page, 5th para, a highest ranked ML model is selected and transmitted to the execution environment 102. Subsequently, the highest ranked ML model is deployed in the execution environment 102).
Johnsson does not specifically disclose wherein the computational task comprises processing of a voice call and the at least one minimum performance threshold comprises at least one minimum performance threshold relating to the processing of the voice call, and the obtaining the at least one minimum performance threshold for the computational task comprises: obtaining a complexity indicator of the computational task and choosing the at least one minimum performance threshold according to at least the complexity indicator of the computational task, and wherein the computational task comprises a speech-to-text conversion and the at least one minimum performance threshold comprises a maximum word error rate.
However, Steelberg, in the same field of endeavor, discloses:
wherein the computational task comprises processing of a voice and the at least one minimum performance threshold comprises at least one minimum performance threshold relating to the processing of the voice, and the obtaining the at least one minimum performance threshold for the computational task comprises: obtaining a complexity indicator of the computational task and choosing the at least one minimum performance threshold according to at least the complexity indicator of the computational task (Steelberg, para 0006, provided herein are embodiments of systems and methods for classifying a media file (e.g., audio, video, multimedia file). One of the methods for classifying a first media segment of a first data type having a corresponding media segment of a second data type includes: extracting a first set of media features of the first media segment of the first data type; generating, using an engine prediction neural network, a best candidate neural network based at least on the first set of media features; determining whether a predicted value of accuracy of the best candidate neural network is above a predetermined accuracy threshold; when the predicted value of accuracy of the best candidate neural network is below the predetermined accuracy threshold, classifying the corresponding media segment of a second data type using a second classification neural network; and selecting, based at least on results of the classification of the corresponding media segment of a second data type, a third classification neural network to classify the first media segment of the first data type … );
wherein the computational task comprises a speech-to-text conversion and the at least one minimum performance threshold comprises a maximum word error rate (Steelberg, para 00114, In transcription, the number of outputs is effectively reduced to a single number (the word error rate) per engine per audio segment; Steelberg, para 0054, The engine prediction neural network can be trained to predict the best-candidate engine by associating dominant features (e.g., weights of a layer) of an audio segment to an accuracy rating (e.g., word error rate) of an engine).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Steelberg in the method of Johnsson because this would enable to be on a more competitive edge in the current business climate because of the processing and analyzing of all sorts of big data (i.e., structured, unstructured, and semi-structured data) using neural networks to discover the market trends, customer behaviors, and other useful indicators relating to their markets, product, and/or services (Steelberg, para 0004).
Johnsson in view of Steelberg does not specifically disclose voice call.
However, Penha, in the same field of endeavor, discloses voice call (Penha, para 0034, Therefore, turning now to FIG. 1, various embodiments of the disclosure provide a real-time call monitoring machine learning framework 100 that uses several different sources of data for the voice call in predicting whether the party is likely to drop the voice call before solving a query. Accordingly, the different sources of data may include the audio 110 of the voice call; [i.e., the “audio” may include “voice call”]).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Penha in the method of Johnsson in view of Steelberg because this would enable to avoid unresolved voice calls by predicting that a party is likely to drop the voice call which would then allow for one or more actions to be taken to avoid the party from dropping the voice call (Penha, para 0002).
Regarding Claim 6, Johnsson in view of Steelberg and Penha disclose the computer-implemented method according to claim 1, wherein the at least one resource consumption indicator of each machine learning model in the plurality of machine learning models (Johnsson, 2nd page, 3rd – 4th para, for a data-driven network each task could be having a plurality of trained ML models and having varying feature sets, accuracies, complexities, data sampling requirements, and hardware requirements. Also, the LTE system or NR base stations are typically a resource-constrained system without excess memory… An existing solution to solve the aforementioned problem include performing field trials of ML model executions for a customer network to determine negative impacts related to the ML model. The solution includes executing a trial and collecting relevant data about resource usage and Key performance Indicators (KPIs) for the ML model) comprises at least one of: an estimated energy consumption of each machine learning model for the computational task, a number of parameters in each machine learning model, a number of processing operations needed to perform the computational task using each machine learning model, an estimated processing time needed to perform the computational task using each machine learning model, and/or an estimated hardware resource consumption of each machine learning model for the computational task (Johnsson, 7th page, 3rd para, The model parameters [i.e., the model parameters of each ML model] correspond to the number of variables that need to be estimated during a training process; [citation is mapped to “a number of parameters in each machine learning model”] ).
Regarding Claim 8, Johnsson in view of Steelberg and Penha disclose the computer-implemented method according to claim 1, wherein the obtaining the at least one minimum performance threshold for the computational task comprises:
obtaining a complexity indicator of the computational task (Johnsson, 2nd page, 8th para – 3rd page, 1st para, The complexity of each machine learning model in the first set of machine learning models is calculated. The method comprises determining, from the first set of machine learning models, at least one suitable machine learning model to be deployed, wherein the determining is based on the calculated complexity and the resource constraints of the execution environment); and
choosing the at least one minimum performance threshold according to at least the complexity indicator of the computational task (Johnsson, 8th page, 5th para, The apparatus 104 further executes the resource shortage function using the resource constraints 302 and complexity of the model to check if a ML model (Mi(F, T)) is compatible for deployment. In step 214, after the execution of the resource shortage function execution, the apparatus 104 creates a second set of ML models that may be deployed on the execution environment 102 without causing resource shortages. Further, in step 215, the apparatus 104 assigns a rank to each ML model in the second set of ML models based on their historical predictive performance).
Regarding Claim 10, Johnsson in view of Steelberg and Penha disclose the computer-implemented method according to claim 1, wherein the obtaining the at least one minimum performance threshold for the computational task comprises:
obtaining metadata related to the computational task (Johnsson, 5th page, 5th para, The apparatus 104, may comprise an arrangement as depicted in Fig. 1 d to select a machine learning model M to be deployed in an execution environment 102 having resource constraints 302… The apparatus 104 may further comprise a memory 142 comprising one or more memory units to store data on. The memory 142 comprises instructions executable by the processor. The memory 412 is arranged to be used to store e.g. measurements, photos, location information, ML models, metadata, instructions, configurations and applications to perform the methods herein when being executed by the processing unit 147); and
choosing the at least one minimum performance threshold according to at least the metadata related to the computational task (Johnsson, 5th page, 3rd para - 5th para, The second set of ML model contains at least one ML model that meet deployment suitability of the execution environment 102, where the deployment suitability is defined by the hardware and software configuration of the execution environment, latency requirements, sampling time of features, and performance requirements of the task…The memory 142 comprises instructions executable by the processor. The memory 412 is arranged to be used to store e.g. measurements, photos, location information, ML models, metadata, instructions, configurations and applications to perform the methods herein when being executed by the processing unit 147; [“performance requirements of the task” as “minimum performance threshold …related to the computational task”; “according to at least the metadata” as “The memory 142 comprises instructions executable by the processor. The memory 412 is arranged to be used to store e.g. … metadata, … to perform the methods herein when being executed by the processing unit 147”]).
Regarding Claim 12, Johnsson in view of Steelberg and Penha disclose the computer-implemented method according to claim 1, wherein the obtaining the at least one minimum performance threshold for the computational task comprises:
monitoring a first performance indicator during another computational task (Johnsson, 7th page, 8th para, The resource shortage function is trained based on the calculated complexity and resource constraints as inputs to determine the suitability of each machine learning model for deployment. In an embodiment, the resource shortage function 304 checks whether the resource constraints 302 of the execution environment (for example, a base station) is compatible with each ML model (Mi(F, T)). The resource shortage function 304 will be further elaborated in Fig.3. After the execution of the resource shortage function 304, the second set of ML models (suitable to deploy) is created by the apparatus 104; [“monitoring” as “checking”]); and
adjusting the at least one minimum performance threshold based on the monitoring of the first performance indicator, wherein the at least one minimum performance threshold comprises a lower level performance indicator than the first performance indicator (Johnsson, 8th page, 7th para, the resource shortage function is performed by executing a rule-based policy on each machine learning model from the first set of machine learning models, where the rule-based policy defines a preferred machine learning model for varying measures of the complexity value and the resource constraint. In an example, the rule-based policy could be programmed to analyze each ML model (Mi (F, T)) based on pre-defined policies provided by a user. In yet another embodiment herein, the resource shortage function could be a dynamic function, where a neural network is updated continuously based on deployment data and historic performance of the ML models).
Regarding Claim 13, Johnsson in view of Steelberg and Penha disclose the computer-implemented method according to claim 1.
Steelberg further teaches:
in response to none of the plurality of machine learning models fulfilling the at least one minimum performance threshold, choosing a machine learning model that has the at least one performance indicator closest to the at least one minimum performance threshold out of the plurality of machine learning models or directing the choosing the machine learning model to a human (Steelberg, para 0015, if none of the best candidate engines has a predicted accuracy value above a predetermined accuracy threshold, identifying an alternate data set associated with the first image portion, the alternate data set comprises non-image data; requesting a second classification engine to classify the alternate data set; receiving, from the second classification engine, a second classification result of the alternate data set; and selecting, based at least on the second classification result of the alternate data set, a third image classification engine to re-classify the portion of the image. The second classification engine is trained to classify data in a same class as the alternate data set. The first and third image classification engines can be different; [i.e., “The second classification engine is trained to classify data in a same class as the alternate data set” as “the second engine (machine learning model) has the same/closest performance as the first classification engine implemented on the alternate data set”]).
Regarding Claim 14, Johnsson discloses a computing device, comprising at least one processor and at least one memory including computer program code, the at least one memory and the computer program code configured to cause the computing device to, with the at least one processor, perform the method according to claim 1 (Johnsson, 5th page, 5th – 6th para, The apparatus comprises a processing unit 147 with one or more processors. The apparatus 104 may further comprise a memory 142 comprising one or more memory units to store data on. The memory 142 comprises instructions executable by the processor. The memory 412 is arranged to be used to store e.g. measurements, photos, location information, ML models, metadata, instructions, configurations and applications to perform the methods herein when being executed by the processing unit 147. Thus, it is herein provided the apparatus 104 e.g. comprising the processing unit 147 and a memory 142, said memory 142 comprising instructions executable by said processing unit 147 whereby said apparatus 104 is operative to: receive a request for a machine learning model solving a task T using a feature set F).
Regarding Claim 15, Johnsson discloses a non-transitory storage medium storing (Johnsson, 3rd page, 3rd para, computer program product ) program code configured to perform the method according to claim 1 when the program code is executed on a computer (Johnsson, 3rd page, 2nd para, there is provided a computer program comprising computer-executable instructions for causing an apparatus to perform the method according to the first aspect of the present disclosure, when the computer-executable instructions are executed on a processing unit included in the apparatus).
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Johnsson in view of Steelberg, further in view of Penha, and further in view of Drake et al. Pat App No. US 20210174958 A1 (Drake).
Regarding Claim 3, Johnsson in view of Steelberg and Penha disclose the computer-implemented method according to claim 1.
Johnsson in view of Steelberg and Penha do not specifically disclose wherein the at least one minimum performance threshold comprises at least one of: a maximum execution time, a minimum accuracy, a minimum precision, a minimum recall, a minimum specificity, a maximum miss-rate, a maximum fall-out, a minimum F1 score, a minimum area under curve, and/or a minimum kappa statistic.
However, Drake, in the same field of endeavor, discloses wherein the at least one minimum performance threshold comprises at least one of: a maximum execution time, a minimum accuracy, a minimum precision, a minimum recall, a minimum specificity, a maximum miss-rate, a maximum fall-out, a minimum F1 score, a minimum area under curve, and/or a minimum kappa statistic (Drake, 0297, Machine learning techniques can be used to assess the commercial testing modalities most optimal for cost/performance/commercial reach as defined in the initial question. A threshold check can be performed: … A subset of assays may be selected from a set of assays to be performed on a given sample based on the total cost of performing the subset of assays, subject to the threshold for assay performance, such as desired minimum accuracy; [i.e., “minimum accuracy” is mapped with the citation] ).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Drake in the method of Johnsson in view of Steelberg and Penha because this would enable a machine learning model to be implemented and used in a production according to the specified performance threshold(s) to classify a new sample once the model has been trained and satisfied one or more specified criteria, (e.g., a desired minimum accuracy, positive predictive value (PPV), negative predictive value (NPV), …, area under the curve (AUC), or a combination thereof) (Drake, para 0316).
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Johnsson in view of Steelberg, further in view of Penha, and further in view of Zhu et al. Pat App No. US 20220305950 A1 (Zhu).
Regarding Claim 4, Johnsson in view of Steelberg and Penha disclose the computer-implemented method according to claim 1.
Johnsson in view of Steelberg and Penha do not specifically disclose wherein the computational task comprises a regression task and the at least one minimum performance threshold comprises at least one of: a maximum mean square error, a maximum root mean square error, and/or a maximum sum of squares error.
However, Zhu, in the same field of endeavor, discloses wherein the computational task comprises a regression task and the at least one minimum performance threshold comprises at least one of: a maximum mean square error, a maximum root mean square error, and/or a maximum sum of squares error (Zhu, para 0020, regression artificial neural network or autoencoder is used to encrypt, compress, decode, and decompress battery cell voltage data. Empirical testing has demonstrated downsizing of a dataset to 1/10 of its original size. The 10-to-1 compression ratio can be further increased if the dataset has more signals. The maximum root-mean-square error (RMSE) among ten RMSEs was 0.01V; [The citation is used to “maximum root mean square error” [i.e., maximum RMSE]] ).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Zhu in the method of Johnsson in view of Steelberg and Penha because this would enable the RMSE to be used to train the neural network along with the training data and to evaluate the performance of the neural network with a given testing data (Zhu, para 0038).
Claim 11 is rejected under 35 U.S.C. 103 as being unpatentable over Johnsson in view of Steelberg, further in view of Penha, further in view of Dunne et al. Pat App No. US 20200272899 A1 (Dunne), and further in view of Shoaib et al. Pat App No. CN 108351984 A (Shoaib).
Regarding Claim 11, Johnsson in view of Steelberg and Penha disclose the computer-implemented method according to claim 1, wherein the obtaining the at least one minimum performance threshold for the computational task comprises:
obtaining test input data (Johnsson, 2nd page, 5th para, the test ML models are executed with different parameters (inputs and features of the model) to collect performance data; [“different parameters (inputs and features of the model)” as “test input data”; OR, Johnsson, 2nd page, 3rd para, execute a plurality of tasks…Some examples of the tasks include beamforming, scheduling, Coordinated multi-point (CoMP) transmission/reception, handover decisions, etc.; [“examples of the tasks” as “input data”]);
obtaining an expected output data for the test input data (Johnsson, 2nd page, 3rd para, some tasks have certain latency requirements in the range of 50 ps to 200 ms. Thus, it is also essential to consider the latency requirements of the task while selecting the associated ML model; OR, Johnsson, 7th page, 2nd para, The complexity of each machine learning model is computed based on parameters comprising at least one of model parameters, model type, model size, training method, number of input features, and feature-sampling cost; [“computed complexity” as “expected output data”]);
feeding the complexity into at least one machine learning model in the plurality of machine learning models, thus obtaining at least one output data (Johnsson, 7th page, 8th para, The resource shortage function is trained based on the calculated complexity and resource constraints as inputs to determine the suitability of each machine learning model for deployment. [i.e., “calculated/computed complexity” as “output data” to be fed/input to train the resource shortage function of the ML model] );
Johnsson in view of Steelberg and Penha do not specifically disclose comparing the at least one output data and the expected output data, and obtaining the at least one minimum performance threshold based at least on the comparison of the at least one output data and the expected output data.
However, Dunne, in the same field of endeavor discloses:
comparing the at least one output data and the expected output data (Dunne, para 0036, learning is accomplished during a “training” process in which the values of the weights in each layer are determined. As an example, the training process may include causing the neural network to process a task for which an expected/desired output is known, comparing the activations generated by the neural network to the expected/desired output, and determining the values of the weights in each layer based on the comparison results; OR, Dunne, para 0091, In operation block 414, the centralized site/device 406 may train the neural network using the acquired (or adjusted) training data. For example, the centralized site/device 406 may cause the neural network to process a task for which an expected/desired output is known, compare the outputs/activations generated by each layer the neural network to the expected/desired outputs, determine the values of the weights in each layer based on the results of the comparison, and generate a trained neural network based on the determined weight values).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Dunne in the method of Johnsson in view of Steelberg and Penha because this would enable generating the neural network difference model by comparing the updated neural network to the trained neural network may include determining one or more neural network layers or one or more neural network weights of the one or more neural network layers to freeze based on a mean of activations of layers in the neural network which have the potential to solve a variety of long-standing technical challenges (Dunne, para 0002-0005)
Johnsson in view of Steelberg, Penha and Dunne do not specifically disclose obtaining the at least one minimum performance threshold based at least on the comparison of the at least one output data and the expected output data.
However, Shoaib, in the same field of endeavor, discloses obtaining the at least one minimum performance threshold based at least on the comparison of the at least one output data and the expected output data (Shoaib, 4th page, 2nd para – 7th para, the neural network may be represented as a function f (FORMULA VIII, XI, tj) wherein each input x associated with weights wi, and each neuron has a threshold tj. in each neuron, calculating wixi, applies a nonlinear function, and the result is compared with the threshold. using the hyperbolic tangent (tanh) as a nonlinear function will get the following example:
tanh (Bi + B0 E wixi) >tj (1)
… non-linear function to the pixel value of each neuron, and each neuron by the result of the nonlinear function is compared with one or more threshold values to output. the output value of neuron output vector. The input vector to create the output vector of the process called feature extraction. method of model and input data based on different needs for different types of feature extraction method based on the feature extraction of different, neural network can be applied to all kinds of data of known or unknown characteristic, comprising a voice amplitude data, seismic data, or other sensor data. the output vector may be provided to the classifier (e.g., a model-based machine learning classifier). For example, classifier can realize a support vector machine, a decision tree, a Fisher linear discriminant, a linear discriminant analysis (LDA) or other classification method. analyzing the classifier output vector and the input image is classified as a category in a set of categories. in the binary classifier, for example, an image can be classified as containing an object of interest (e.g., a face) (output "1") or not containing an object of interest (output "0"). neural network is typically trained to determining neuronal threshold value and a classifier model parameters. input data and available classifier output label is provided to the training algorithm as the training algorithm tries to make all classifier outputs output error tag is minimized. solving to obtain the parameter value and the minimum error threshold value can be realized).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Shoaib in the method of Johnsson in view of Steelberg, Penha and Dunne because this would enable efficiently realizing a convolutional neural network (Shoaib, Abstract).
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MULUGETA T. DUGDA whose telephone number is (703)756-1106. The examiner can normally be reached Mon - Fri, 4:30am - 7:00pm.
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, Paras D. Shah can be reached at 571-270-1650. 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.
/MULUGETA TUJI DUGDA/Examiner, Art Unit 2653
/DOUGLAS GODBOLD/Primary Examiner, Art Unit 2655
Prosecution Insights