DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application is being examined under the pre-AIA first to invent provisions.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 01/01/2025, 10/15/2024, 04/26/2023, 10/17/2022, and 03/31/2022 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Response to Arguments
103 Arguments
Applicant asserts:
Applicants argues, starting on page 10-15, that the prior art does not teach all the recitations of the amended claim 69.
Examiner response:
Applicant’s arguments with respect to claim(s) 69 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.
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.
Claim(s) 69 is rejected under 35 U.S.C. 103 as being unpatentable over Dayong He et al; US 20190086988 A1 filed on Sep 18, 2017 (hereinafter “He”) in view of Hugo Tullberg et al; US 20220322195 A1 filed on Jun 19, 2019 (hereinafter “Tullbeg”) in further view of Michael McCourt et al; US 20200019888 A1 filed on Jul 15, 2019 (hereinafter “McCourt”) in further view of Erik Richard Stauffer et al; US 20210029643 A1 filed on Apr 17, 2019 (hereinafter “Stauffer”) in further view of Lan Vu et al; US 20200372360 A1 filed on May 20, 2019 (hereinafter “Vu”) in further view of Miguel Angel Hernandez Orozco et al; US 11182305 B1 filed on Sep 30, 2019 (hereinafter “Orozco”) in further view of Sebastian Faxer et al; US 20220303108 A1 filed on Fun 19, 2019 (hereinafter “Faxer”) in further view of Belghoul; Farouk et al; US 20190132708 A1 filed on Sep 28, 2018 (hereinafter “Belghoul”) in further view of Wang; Leyuan et al; US 11797876 B1 filed on Jun 26, 2019 (hereinafter “Wang”).
Regarding Claim 69, He teaches A method performed by a network node in a communication network, the network node comprising a processor and a transceiver, the method comprising: (He Paragraph 0044; “UE device 110 and machine learning system 140 may each include one or more devices 200 or components of device 200. As shown in FIG. 2, device 200 may include a bus 210, a processor 220, a memory 230, an input device 240, an output device 250, and a communication interface 260.” He Paragraph 0049; “Communication interface 260 may include a transceiver that enables device 200 to communicate with other devices and/or systems via wireless communications” Examiner notes that a network node (machine learning system 140) in a communication network (provider network) comprises a processor (processor 220) and a transceiver (communication interface 260))
receiving, via the transceiver, from a User Equipment (UE), UE capabilities comprising a machine learning based assistance capability, wherein the machine learning based assistance capability comprises a plurality of machine learning based functionalities, each machine learning based functionality of the plurality of machine learning based functionalities comprising a machine learning entity and at least one machine learning mode associated with the machine learning entity, (He Paragraph 0025; “Different machine learning modules may use different algorithms and/or different parameters for the algorithms to make decisions, predictions, and/or inferences. The smart engine may make adjustments to data sources and/or machine learning models/classifiers/algorithms used by a particular machine learning module based on the determined device status.” He Paragraph 0056; “Smart engine 330 may control the operation of machine learning processes on UE device 110. Smart engine 330 may obtain the device status of UE device 110… applications currently running on UE device 110, machine learning processes scheduled to run… whether to communicate with machine learning system 140 to perform a machine learning process” He Paragraph 0054; “Applications 310 may include applications running on UE device 110 that may make use of machine learning. For example, applications 310 may include … another type of application that may utilize machine learning.” He Paragraph 0061; “Machine learning modules 350 may include machine learning modules for particular applications. For example, a first machine learning module may be trained to perform behavior classification, a second machine learning module may be trained to perform object recognition in images” Examiner notes that machine learning system receives from a UE (smart engine in UE communicates with machine learning system) UE capabilities (applications) comprising a machine learning based assistance capability (applications make use of machine learning) comprises a plurality of machine learning based functionalities (plurality of applications include various functionalities as seen in Paragraph 0054); each machine learning based functionality comprises a machine learning entity (machine learning module for particular application) and at least one machine learning mode associated with machine learning entity (different machine learning models used in machine learning module))
the at least one machine learning mode of each machine learning based functionality comprises a fallback operating mode or a primary machine learning mode, (He Paragraph 0024; “Different machine learning models may be associated with different expected resource use. For example, a first machine learning model may use less battery power, processor time, memory, and/or network bandwidth, etc., and a second machine learning model may use more battery power, processor time, memory, and/or network bandwidth, etc.” Examiner notes that the at least one machine learning mode (machine learning model) of each machine learning based functionality comprises a fallback operating mode (a first machine learning model may use less battery power, processor time, memory, and/or network bandwidth, etc.) or a primary machine learning mode (a second machine learning model may use more battery power, processor time, memory, and/or network bandwidth, etc))
and the fallback operating mode applies when the primary machine learning mode is unavailable due to at least one of: an absence of input data required by the primary machine learning mode, a deactivation of the primary machine learning mode, or a measurement error, or a connectivity error; (He Paragraph 0024; “The device status may indicate a current resource capacity of the wireless communication device, measured by parameters including, for example, the battery power level, a measure of a processor load, a measure of memory use, a measure of a network connection quality, and/or other types of measures of the resource capacity of the wireless communication device. Different machine learning models may be associated with different expected resource use.” Examiner notes that the fallback operating mode (different machine learning model) is applied when the primary machine learning mode is unavailable due to connectivity error (bad connection quality))
initiating, with the UE, a communication session [for an initial exploration and tuning of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability]; (He Fig 9 and Paragraph 0119; “FIG. 9, signal flow 900 may include a user opening application 410 via user interface 910 (signal 912).” Examiner notes that a communication session is initiated with the UE as shown in fig 9)
PNG
media_image1.png
487
643
media_image1.png
Greyscale
activating at least one machine learning based functionality among the plurality of machine learning based functionalities, wherein the activation comprises a network-initiated activation comprising: (He Paragraph 0119; “FIG. 9 is a diagram of an exemplary signal flow 900 according to an implementation described herein… For example, assume the user activates a financial application and the financial application selects, in addition to authenticating the user via a login process, to perform behavioral authentication using a machine learning module trained to perform behavioral authentication based on the user's behavior patterns. Thus, application 410 may call an API to request a machine learning service to perform behavioral authentication via smart engine 420 (signal 914). Smart engine 420 may then obtain device status data from device status module 430 (signals 916 and 918)” Examiner notes that Fig 9 shows activating at least one machine learning based functionality amount the plurality of machine learning based functionalities (perform behavioral authentication using a machine learning module), wherein the activation comprises a network initiated activation (signal flow))
selecting, by the processor based on radio resource management needs of the UE, the at least one machine learning based functionality; (He Paragraph 0119; “Smart engine 420 may then determine whether to perform the requested machine learning process based on the obtained device status data (block 920)... [0120] Smart engine 420 may then select a machine learning module 470 (signal 930) and then select an option for the selected machine learning module 470 by accessing ML options DB 460 (signal 940) and using the information in ML options DB 470 relating to the available options to select the options for machine learning module 470 (signal 942).” Examiner notes that smart engine selects the at least one machine learning based functionality (machine learning module associated with application) based on radio resource management needs of the UE (device status))
causing transmission of a first activation request to the UE to activate the selected at least one machine learning based functionality; (He Paragraph 0119; “the user activates a financial application and the financial application selects, in addition to authenticating the user via a login process, to perform behavioral authentication using a machine learning module” Examiner refers to previous mapping to show that step 912 open application is causing a transmission of a first activation request to the UE (UE device interface on machine learning system is used to communicate with UE devices) to activate the selected at least one machine learning based functionality (to perform behavioral authentication using a machine learning module))
and receiving, from the UE, a first activation request response, wherein in a first instance when the at least one machine learning based functionality is activated, the first activation request response comprises an indication of the activation, and (He Paragraph 0121; “machine learning module 470 may provide the result of the behavioral authentication to application 410 (signal 960) and application 410 may provide the results to the user via user interface 910, informing the user that the user has been authenticated (signal 962).” Examiner notes that the UE sends a first activation request response (result of the behavioral authentication) to the user interface on machine learning system, wherein a first instance when the at least one machine learning based functionality is activated (the returned result shows ML functionality is activated), the first activation request response comprises an indication of the activation (the returned result is an indication of activation))
in a second instance when the at least one machine learning based functionality failed to activate, the first activation request response [comprises a cause for the failure, the cause comprising at least one of: an indication of an alternative preferred machine learning based functionality, an indication to use a fallback operating mode, an indication of insufficient battery power, or an indication of a change in channel radio conditions that exceeds a channel condition threshold]; (He Paragraph 0128; “If it is determined that the battery level is not greater than zero (block 1130—NO), a low accuracy model may be selected (block 1140) and fewer data sources may be selected (block 1150). As an example, smart engine 420 may select a particular machine learning module 470 associated with a less computationally complex model… Furthermore, smart engine 420 may select fewer data sources… For example, if performing a behavioral authentication machine learning process, smart engine 420 may select to use a touchscreen data source and may select not to use an accelerometer data source.” Examiner notes that when the at least one machine learning based functionality failed to activate (when high accuracy model cannot be used), the first activation request response is sent (result of application using the low accuracy model))
He does not teach each machine learning entity is configured for at least one of a handover (HO) prediction or a quality of service (QoS) prediction,
However, Tullberg does teach each machine learning entity is configured for at least one of a handover (HO) prediction or a quality of service (QoS) prediction, (Tullberg Paragraph 0072; “The dedicated machine learning model may accurately learn the propagation environment and the prediction of handover of UE to neighboring cells.” Examiner notes that machine learning entity (machine learning model) is configured for handover prediction (prediction of handover))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He and Tullberg. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. One of ordinary skill would have motivation to combine He and Tullberg to improve the over performance of the system by implementing a machine learning method “a machine learning method for cell handover reduces overhead, delay and handover failures. This improves the overall performance of the system and aid users close to the cell edge.” (Tullberg Paragraph 0025).
He in view of Tullberg does not teach session for an initial exploration and tuning of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability
However McCourt does teach session for an initial exploration and tuning of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability (McCourt Paragraph 0076; “S220 may function to dynamically set and/or configure exploration and exploitation parameters for a tuning associated with each partial tuning task of a multi-task tuning request. Exploration parameters preferably enable the tuning service to identify potential hyperparameter values for a given model.” Examiner notes that session for initial exploration and tuning (configure exploration and exploitation parameters for a tuning) of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability (given model; tuning hypermeters of model that machine learning based functionalities is executed with))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, and McCourt. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. One of ordinary skill would have motivation to combine He, Tullberg, and McCourt to improve the overall performance of a model with saving computational resources “The embodiments of the present application, however, provide an intelligent optimization platform that functions to optimize hyperparameters and/or parameters of any type of model with significantly fewer evaluation thereby saving computational resources while greatly improving an overall performance of a model.” (McCourt Paragraph 0033).
He in view of Tullberg in further view of McCourt does not teach response comprises a cause for the failure, the cause comprising at least one of: an indication of an alternative preferred machine learning based functionality, an indication to use a fallback operating mode, an indication of insufficient battery power, or an indication of a change in channel radio conditions that exceeds a channel condition threshold
receiving via the transceiver, from the UE, a change request [to change the activated at least one machine learning based functionality], wherein the change request comprises a cause indication for the change, the cause indication comprising at least one of the following:
the activated at least one machine [learning based functionality] cannot be run anymore due to at least one of: low battery power, low computational power, and detected changes in input data;
in response to receiving the change request: transmitting, via the transceiver, a machine learning functionality deactivation reply to the UE, the machine learning functionality deactivation reply comprising a confirmation of a change to the activated at least one machine learning based functionality corresponding to the cause indication; and
However, Stauffer does teach response comprises a cause for the failure, the cause comprising at least one of: an indication of an alternative preferred machine learning based functionality, an indication to use a fallback operating mode, an indication of insufficient battery power, or an indication of a change in channel radio conditions that exceeds a channel condition threshold (Stauffer Paragraph 0004; “the user device can autonomously provide, to the base station, a request to enter a low-power mode based on local factors, such as a low battery power level or a high temperature of the UE. The base station can receive the request and dictate a change in the power mode of the UE to a low-power mode to reduce power consumption, conserve battery life, and/or decrease the UE temperature.” Examiner notes that UE cannot perform action, so a response (request) is sent to base station, cause for failure comprises an indication of insufficient battery power (low battery power))
receiving via the transceiver, from the UE, a change request [to change the activated at least one machine learning based functionality], wherein the change request comprises a cause indication for the change, the cause indication comprising at least one of the following: (Stauffer Paragraph 0004; “the user device can autonomously provide, to the base station, a request to enter a low-power mode based on local factors, such as a low battery power level or a high temperature of the UE. The base station can receive the request and dictate a change in the power mode of the UE to a low-power mode to reduce power consumption, conserve battery life, and/or decrease the UE temperature.” Examiner notes a base station receives from the UE (user device) a change request (a request to enter a low-power mode), wherein the change request comprises a cause indication (change request to enter lower power mode is a cause indication))
the activated at least one machine [learning based functionality] cannot be run anymore due to at least one of: low battery power, low computational power, and detected changes in input data; (Examiner refers to previous mapping to show that the activated machine (user device) cannot be run anymore due to low battery power (low battery power level));
in response to receiving the change request: transmitting, via the transceiver, a machine learning functionality deactivation reply to the UE, the machine learning functionality deactivation reply comprising a confirmation of a change to the activated at least one machine learning based functionality corresponding to the cause indication; and (Stauffer Paragraph 0034; “a downlink (DL) signal is detected that includes an FLPM acknowledgment (ACK) from the base station. For example, the UE 102 can detect a DL signal from the base station 104 via the wireless link 106. In at least one example, the DL signal includes instructions to cause the UE 102 to enter the low-power mode. The low-power mode may be the specifically requested low-power mode, such as the inactive mode.” Examiner notes that a machine learning functionality deactivation reply to the UE is transmitted (User device detects a downlink (DL) signal that includes an FLPM acknowledgment (ACK) from the base station); the reply comprises a confirmation of a change to the active machine learning based functionality corresponding to the cause indication (DL signal includes instructions to cause the UE 102 to enter the low-power mode corresponding to low battery power level))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, and Stauffer. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, and Stauffer to reduce power consumption, conserve battery life, and decrease UE temperature “The base station can receive the request and dictate a change in the power mode of the UE to a low-power mode to reduce power consumption, conserve battery life, and/or decrease the UE temperature. Alternatively, the request can be sent based on a user input selecting a particular power mode and the base station can honor that request by causing the UE to enter the particular power mode. In this way, flexibility is provided to the UE (and the user of the UE) to change the power mode of the UE to reduce power consumption.” (Stauffer Paragraph 0004).
He in view of Tullberg in further view of McCourt in further view of Stauffer does not teach in response to the activation, causing transmission of a machine learning inference reporting configuration message to the UE, the machine learning inference reporting configuration message comprising configuration information for at least one of a reporting periodicity or a content of an inference report;
However, Vu does teach in response to the activation, causing transmission of a machine learning inference reporting configuration message to the UE, the machine learning inference reporting configuration message comprising configuration information for at least one of a reporting periodicity or a content of an inference report; (Vu Paragraph 0068; “The figure includes a client 902 with a client inference engine 904 and a machine learning inference service 910 with a server inference engine 912. The client inference engine 904 includes CM 216 and server inference engine 912 includes SM 218. The client inference engine 904 receives client input data 906, performs an inference on CM 216 and sends the output, outputK, to the server inference engine 912. The server inference engine uses outputK to perform an inference on SM 218 and sends the results, outputO, to the client inference engine which reports the inference results 908 to the user.” Examiner notes server inference engine causes transmission/sends a machine learning interface reporting configuration to the UE (sends the results, outputO, to the client inference engine which reports the inference results 908 to the user), the machine learning inference reporting configuration message comprises content of an inference report (inference results 908))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, and Vu. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Vu teaches a method and system for training a neural network. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, and Vu to improve the performance of training to obtain an improved neural network model “Another advantage is that the workload of training of or inference using the neural network can be balanced between a client and a server training/inference engine, thereby improving the performance of training of or inference using the neural network. Yet another advantage is that intermediate data sent between a client training/inference engine and a server training/inference engine in the training/inference process does not include sensitive information, which helps to maintain the privacy of the original data. Yet another advantage is that bandwidth used over a communication link between the client training/inference engine and the server training/inference engine can be reduced, thereby improving the performance of the training of or inference using the neural network. Yet another advantage is that power consumption of edge devices in a cloud containing the server training engine can be reduced by assigning a large portion of the computation to the cloud.” (Vu Paragraph 0004).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Vu does not teach continuously monitoring a quality of inference reports received from the UE,
However, Orozco does teach continuously monitoring a quality of inference reports received from the UE, (Orozco Column 11 Line 20; “Data structures 102 may be periodically accessed to determine the accuracy of the response outputs 118 stored therein, such as at a time when a data structure 102 is received from a computing device.” Examiner notes that a quality of inference reports (accuracy of the response outputs) received from the UE (received from a computing device) is continuously monitored (periodically accessed to determine))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Vu, and Orozco. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Vu, and Orozco to make the system more efficient by reducing the time and computational resources used “storing certain inputs and corresponding outputs in a data structure for future use, then using the outputs stored in the data structure rather than recomputing an output fora subsequent request may significantly reduce the time and computational resources used to generate a response.” (Orozco Column 4 Line 1).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Vu in further view of Orozco does not teach and based on the monitored quality, performing a Radio Resource Management (RRM) action;
de-allocating network resources previously allocated for receiving the inference reports from the UE.
However, Faxer does teach and based on the monitored quality, performing a Radio Resource Management (RRM) action; (Faxer Paragraph 0088; “if a certain terminal device 300 has good radio conditions and strong uplink signals, or it has a relatively lower QoS requirement, then, a small adjacent interference might be considered as less severe. The scheduling restriction (as defined by the radio resource management action) might be applied per signal/channel basis.” Examiner notes that based on the monitored quality (QoS/Quality of Service), a RRM action is performed (radio resource management action is applied))
de-allocating network resources previously allocated for receiving the inference reports from the UE. (Faxer Paragraph 0088; “if a certain terminal device 300 has good radio conditions and strong uplink signals, or it has a relatively lower QoS requirement, then, a small adjacent interference might be considered as less severe. The scheduling restriction (as defined by the radio resource management action) might be applied per signal/channel basis.” Examiner notes that the network resources previously allocated for receiving the inference reports from the UE is de allocated (the scheduling restriction is applied per signal/channel basis))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Vu, Orozco, and Faxer. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. Faxer teaches a method for cross link interference handling. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Vu, Orozco, and Faxer to improve the performance of services and reduce interference “One or more of the various embodiments improve the performance of OTT services provided to UE 530 using OTT connection 550, in which wireless connection 570 forms the last segment. More precisely, the teachings of these embodiments may reduce interference, due to improved classification ability of airborne UEs which can generate significant interference.” (Faxer Paragraph 0151).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Vu in further view of Orozco in further view of Faxer does not teach a number of detected errors has exceeded a predetermined threshold within a given time window;
However, Belghoul does teach a number of detected errors has exceeded a predetermined threshold within a given time window; (Belghoul Paragraph 0119; “The UEs may additionally send error metric(s), other meta data, and/or time stamps.” Examiner notes that sending error metrics along with time stamps is an indication that a number of detected errors has exceeded a predetermined threshold within a given time window; predetermined threshold can be if a single error has occurred);
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Vu, Orozco, Faxer, and Belghoul. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. Faxer teaches a method for cross link interference handling. Belghoul teaches Techniques for optimizing and deploying convolutional neural network (CNN) machine learning models for inference using integrated graphics processing units are described. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Vu, Orozco, Faxer, and Belghoul to implement the techniques for shorter latency, less burden on network bandwidth, and better privacy protection for users “it is becoming more and more desirable to execute the model inference directly at the edge devices for shorter latency, less burden of the network bandwidth, and better privacy protection to the users.” (Belghoul Column 1 Line 18).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Vu in further view of Orozco in further view of Faxer in further view of Belghoul does not teach a request to restart the machine learning based assistance capability;
and a request to switch to another configured machine learning based assistance;
However, Wang does teach a request to restart the machine learning based assistance capability; (Wang Page 23 Paragraph 4; “In response, in some embodiments, the user, via the user device 702, can transmit a request to the model training system 120 to modify the machine learning model being trained (for example, transmit a modification request)… For example, the model training system 120 can cause the virtual machine instance 722 to optionally delete an existing ML training container 730, create and initialize a new ML training container 730 using some or all of the information included in the request, and execute the code 737 stored in the new ML training container 730 to restart the machine learning model training process.” Examiner notes that the user device requests to restart the machine learning based assistance capability through restarting the training process of the machine learning model.);
and a request to switch to another configured machine learning based assistance; (Wang Page 23 Paragraph 4; “In response, in some embodiments, the user, via the user device 702, can transmit a request to the model training system 120 to modify the machine learning model being trained (for example, transmit a modification request). The request can include a new or modified container image, a new or modified algorithm, new or modified hyperparameter(s), and/or new or modified information describing the computing machine on which to train a machine learning model. The model training system 120 can modify the machine learning model accordingly. For example, the model training system 120 can cause the virtual machine instance 722 to optionally delete an existing ML training container 730, create and initialize a new ML training container 730 using some or all of the information included in the request, and execute the code 737 stored in the new ML training container 730 to restart the machine learning model training process.” Examiner notes that the request to restart the machine learning based assistance capability is also a request to switch to another configured machine learning based assistance/newly trained machine learning model);
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Vu, Orozco, Faxer, Belghoul, and Wang. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. Faxer teaches a method for cross link interference handling. Belghoul teaches Techniques for optimizing and deploying convolutional neural network (CNN) machine learning models for inference using integrated graphics processing units are described. Wang teaches sending out requests to restart and switch machine learning models based on cause indications to optimize the model to run on integrated graphics processing units. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Vu, Orozco, Faxer, Belghoul, and Wang to have a user equipment that comprises machine learning capabilities to perform machine learning functionalities and to restart and switch machine learning models based on cause indications “Thus, a user device 702 can refer to trained machine learning model(s) stored in the ML scoring container(s) 750 using the endpoint. This allows for the network address of an ML scoring container 750 to change without causing the user operating the user device 702 to change the way in which the user refers to a trained machine learning model.” (Wang Page 23 Paragraph 3).
Claim(s) 76 and 83 are rejected under 35 U.S.C. 103 as being unpatentable over Dayong He et al; US 20190086988 A1 filed on Sep 18, 2017 (hereinafter “He”) in view of Hugo Tullberg et al; US 20220322195 A1 filed on Jun 19, 2019 (hereinafter “Tullbeg”) in further view of Michael McCourt et al; US 20200019888 A1 filed on Jul 15, 2019 (hereinafter “McCourt”) in further view of Erik Richard Stauffer et al; US 20210029643 A1 filed on Apr 17, 2019 (hereinafter “Stauffer”) in further view of Michelle Felt et al; US 9794786 B2 filed on Sep 24, 2015 (hereinafter “Felt”) in further view of Lan Vu et al; US 20200372360 A1 filed on May 20, 2019 (hereinafter “Vu”) in further view of Miguel Angel Hernandez Orozco et al; US 11182305 B1 filed on Sep 30, 2019 (hereinafter “Orozco”) in further view of Sebastian Faxer et al; US 20220303108 A1 filed on Fun 19, 2019 (hereinafter “Faxer”) in further view of Belghoul; Farouk et al; US 20190132708 A1 filed on Sep 28, 2018 (hereinafter “Belghoul”) in further view of Wang; Leyuan et al; US 11797876 B1 filed on Jun 26, 2019 (hereinafter “Wang”).
Regarding Claim 76, He teaches An apparatus comprising: a processor; a transceiver; and a memory comprising computer-executable instructions that, when executed by the processor, cause the apparatus to perform the following operations: (He Paragraph 0044; “UE device 110 and machine learning system 140 may each include one or more devices 200 or components of device 200. As shown in FIG. 2, device 200 may include a bus 210, a processor 220, a memory 230, an input device 240, an output device 250, and a communication interface 260.” He Paragraph 0049; “Communication interface 260 may include a transceiver that enables device 200 to communicate with other devices and/or systems via wireless communications” He Claim 20; “A non-transitory computer-readable memory device storing instructions executable by a processor…” Examiner notes that a apparatus (machine learning system 140) comprises a processor (processor 220), a transceiver (communication interface 260), and a memory comprising executable instructions)
receiving, via the transceiver, from a User Equipment (UE), UE capabilities comprising a machine learning based assistance capability, wherein the machine learning based assistance capability comprises a plurality of machine learning based functionalities, each machine learning based functionality of the plurality of machine learning based functionalities comprising a machine learning entity and at least one machine learning mode associated with the machine learning entity, (He Paragraph 0025; “Different machine learning modules may use different algorithms and/or different parameters for the algorithms to make decisions, predictions, and/or inferences. The smart engine may make adjustments to data sources and/or machine learning models/classifiers/algorithms used by a particular machine learning module based on the determined device status.” He Paragraph 0056; “Smart engine 330 may control the operation of machine learning processes on UE device 110. Smart engine 330 may obtain the device status of UE device 110… applications currently running on UE device 110, machine learning processes scheduled to run… whether to communicate with machine learning system 140 to perform a machine learning process” He Paragraph 0054; “Applications 310 may include applications running on UE device 110 that may make use of machine learning. For example, applications 310 may include … another type of application that may utilize machine learning.” He Paragraph 0061; “Machine learning modules 350 may include machine learning modules for particular applications. For example, a first machine learning module may be trained to perform behavior classification, a second machine learning module may be trained to perform object recognition in images” Examiner notes that machine learning system receives from a UE (smart engine in UE communicates with machine learning system) UE capabilities (applications) comprising a machine learning based assistance capability (applications make use of machine learning) comprises a plurality of machine learning based functionalities (plurality of applications include various functionalities as seen in Paragraph 0054); each machine learning based functionality comprises a machine learning entity (machine learning module for particular application) and at least one machine learning mode associated with machine learning entity (different machine learning models used in machine learning module))
the at least one machine learning mode of each machine learning based functionality comprises a fallback operating mode or a primary machine learning mode, (He Paragraph 0024; “Different machine learning models may be associated with different expected resource use. For example, a first machine learning model may use less battery power, processor time, memory, and/or network bandwidth, etc., and a second machine learning model may use more battery power, processor time, memory, and/or network bandwidth, etc.” Examiner notes that the at least one machine learning mode (machine learning model) of each machine learning based functionality comprises a fallback operating mode (a first machine learning model may use less battery power, processor time, memory, and/or network bandwidth, etc.) or a primary machine learning mode (a second machine learning model may use more battery power, processor time, memory, and/or network bandwidth, etc))
and the fallback operating mode applies when the primary machine learning mode is unavailable due to at least one of: an absence of input data required by the primary machine learning mode, a deactivation of the primary machine learning mode, or a measurement error, or a connectivity error; (He Paragraph 0024; “The device status may indicate a current resource capacity of the wireless communication device, measured by parameters including, for example, the battery power level, a measure of a processor load, a measure of memory use, a measure of a network connection quality, and/or other types of measures of the resource capacity of the wireless communication device. Different machine learning models may be associated with different expected resource use.” Examiner notes that the fallback operating mode (different machine learning model) is applied when the primary machine learning mode is unavailable due to connectivity error (bad connection quality))
initiating, with the UE, a communication session [for an initial exploration and tuning of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability]; (He Fig 9 and Paragraph 0119; “FIG. 9, signal flow 900 may include a user opening application 410 via user interface 910 (signal 912).” Examiner notes that a communication session is initiated with the UE as shown in fig 9)
PNG
media_image1.png
487
643
media_image1.png
Greyscale
activating at least one machine learning based functionality among the plurality of machine learning based functionalities, wherein the activation comprises one of: a network-initiated activation comprising: (He Paragraph 0119; “FIG. 9 is a diagram of an exemplary signal flow 900 according to an implementation described herein… For example, assume the user activates a financial application and the financial application selects, in addition to authenticating the user via a login process, to perform behavioral authentication using a machine learning module trained to perform behavioral authentication based on the user's behavior patterns. Thus, application 410 may call an API to request a machine learning service to perform behavioral authentication via smart engine 420 (signal 914). Smart engine 420 may then obtain device status data from device status module 430 (signals 916 and 918)” Examiner notes that Fig 9 shows activating at least one machine learning based functionality amount the plurality of machine learning based functionalities (perform behavioral authentication using a machine learning module), wherein the activation comprises a network initiated activation (signal flow))
selecting, by the processor based on radio resource management needs of the UE, the at least one machine learning based functionality; (He Paragraph 0119; “Smart engine 420 may then determine whether to perform the requested machine learning process based on the obtained device status data (block 920)... [0120] Smart engine 420 may then select a machine learning module 470 (signal 930) and then select an option for the selected machine learning module 470 by accessing ML options DB 460 (signal 940) and using the information in ML options DB 470 relating to the available options to select the options for machine learning module 470 (signal 942).” Examiner notes that smart engine selects the at least one machine learning based functionality (machine learning module associated with application) based on radio resource management needs of the UE (device status))
causing transmission of a first activation request to the UE to activate the selected at least one machine learning based functionality; (He Paragraph 0119; “the user activates a financial application and the financial application selects, in addition to authenticating the user via a login process, to perform behavioral authentication using a machine learning module” Examiner refers to previous mapping to show that step 912 open application is causing a transmission of a first activation request to the UE (UE device interface on machine learning system is used to communicate with UE devices) to activate the selected at least one machine learning based functionality (to perform behavioral authentication using a machine learning module))
and receiving, from the UE, a first activation request response, wherein in a first instance when the at least one machine learning based functionality is activated, the first activation request response comprises an indication of the activation, and (He Paragraph 0121; “machine learning module 470 may provide the result of the behavioral authentication to application 410 (signal 960) and application 410 may provide the results to the user via user interface 910, informing the user that the user has been authenticated (signal 962).” Examiner notes that the UE sends a first activation request response (result of the behavioral authentication) to the user interface on machine learning system, wherein a first instance when the at least one machine learning based functionality is activated (the returned result shows ML functionality is activated), the first activation request response comprises an indication of the activation (the returned result is an indication of activation))
in a second instance when the at least one machine learning based functionality failed to activate, the first activation request response [comprises a cause for the failure, the cause comprising at least one of: an indication of an alternative preferred machine learning based functionality, an indication to use a fallback operating mode, an indication of insufficient battery power, or an indication of a change in channel radio conditions that exceeds a channel condition threshold]; (He Paragraph 0128; “If it is determined that the battery level is not greater than zero (block 1130—NO), a low accuracy model may be selected (block 1140) and fewer data sources may be selected (block 1150). As an example, smart engine 420 may select a particular machine learning module 470 associated with a less computationally complex model… Furthermore, smart engine 420 may select fewer data sources… For example, if performing a behavioral authentication machine learning process, smart engine 420 may select to use a touchscreen data source and may select not to use an accelerometer data source.” Examiner notes that when the at least one machine learning based functionality failed to activate (when high accuracy model cannot be used), the first activation request response is sent (result of application using the low accuracy model))
He does not teach each machine learning entity is configured for at least one of a handover (HO) prediction or a quality of service (QoS) prediction,
However, Tullberg does teach each machine learning entity is configured for at least one of a handover (HO) prediction or a quality of service (QoS) prediction, (Tullberg Paragraph 0072; “The dedicated machine learning model may accurately learn the propagation environment and the prediction of handover of UE to neighboring cells.” Examiner notes that machine learning entity (machine learning model) is configured for handover prediction (prediction of handover))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He and Tullberg. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. One of ordinary skill would have motivation to combine He and Tullberg to improve the over performance of the system by implementing a machine learning method “a machine learning method for cell handover reduces overhead, delay and handover failures. This improves the overall performance of the system and aid users close to the cell edge.” (Tullberg Paragraph 0025).
He in view of Tullberg does not teach session for an initial exploration and tuning of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability
However McCourt does teach session for an initial exploration and tuning of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability (McCourt Paragraph 0076; “S220 may function to dynamically set and/or configure exploration and exploitation parameters for a tuning associated with each partial tuning task of a multi-task tuning request. Exploration parameters preferably enable the tuning service to identify potential hyperparameter values for a given model.” Examiner notes that session for initial exploration and tuning (configure exploration and exploitation parameters for a tuning) of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability (given model; tuning hypermeters of model that machine learning based functionalities is executed with))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, and McCourt. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. One of ordinary skill would have motivation to combine He, Tullberg, and McCourt to improve the overall performance of a model with saving computational resources “The embodiments of the present application, however, provide an intelligent optimization platform that functions to optimize hyperparameters and/or parameters of any type of model with significantly fewer evaluation thereby saving computational resources while greatly improving an overall performance of a model.” (McCourt Paragraph 0033).
He in view of Tullberg in further view of McCourt does not teach response comprises a cause for the failure, the cause comprising at least one of: an indication of an alternative preferred machine learning based functionality, an indication to use a fallback operating mode, an indication of insufficient battery power, or an indication of a change in channel radio conditions that exceeds a channel condition threshold
receiving via the transceiver, from the UE, a change request [to change the activated at least one machine learning based functionality], wherein the change request comprises a cause indication for the change, the cause indication comprising at least one of the following:
the activated at least one machine [learning based functionality] cannot be run anymore due to at least one of: low battery power, low computational power, and detected changes in input data;
in response to receiving the change request: transmitting, via the transceiver, a machine learning functionality deactivation reply to the UE, the machine learning functionality deactivation reply comprising a confirmation of a change to the activated at least one machine learning based functionality corresponding to the cause indication; and
However, Stauffer does teach response comprises a cause for the failure, the cause comprising at least one of: an indication of an alternative preferred machine learning based functionality, an indication to use a fallback operating mode, an indication of insufficient battery power, or an indication of a change in channel radio conditions that exceeds a channel condition threshold (Stauffer Paragraph 0004; “the user device can autonomously provide, to the base station, a request to enter a low-power mode based on local factors, such as a low battery power level or a high temperature of the UE. The base station can receive the request and dictate a change in the power mode of the UE to a low-power mode to reduce power consumption, conserve battery life, and/or decrease the UE temperature.” Examiner notes that UE cannot perform action, so a response (request) is sent to base station, cause for failure comprises an indication of insufficient battery power (low battery power))
receiving via the transceiver, from the UE, a change request [to change the activated at least one machine learning based functionality], wherein the change request comprises a cause indication for the change, the cause indication comprising at least one of the following: (Stauffer Paragraph 0004; “the user device can autonomously provide, to the base station, a request to enter a low-power mode based on local factors, such as a low battery power level or a high temperature of the UE. The base station can receive the request and dictate a change in the power mode of the UE to a low-power mode to reduce power consumption, conserve battery life, and/or decrease the UE temperature.” Examiner notes a base station receives from the UE (user device) a change request (a request to enter a low-power mode), wherein the change request comprises a cause indication (change request to enter lower power mode is a cause indication))
the activated at least one machine [learning based functionality] cannot be run anymore due to at least one of: low battery power, low computational power, and detected changes in input data; (Examiner refers to previous mapping to show that the activated machine (user device) cannot be run anymore due to low battery power (low battery power level));
in response to receiving the change request: transmitting, via the transceiver, a machine learning functionality deactivation reply to the UE, the machine learning functionality deactivation reply comprising a confirmation of a change to the activated at least one machine learning based functionality corresponding to the cause indication; and (Stauffer Paragraph 0034; “a downlink (DL) signal is detected that includes an FLPM acknowledgment (ACK) from the base station. For example, the UE 102 can detect a DL signal from the base station 104 via the wireless link 106. In at least one example, the DL signal includes instructions to cause the UE 102 to enter the low-power mode. The low-power mode may be the specifically requested low-power mode, such as the inactive mode.” Examiner notes that a machine learning functionality deactivation reply to the UE is transmitted (User device detects a downlink (DL) signal that includes an FLPM acknowledgment (ACK) from the base station); the reply comprises a confirmation of a change to the active machine learning based functionality corresponding to the cause indication (DL signal includes instructions to cause the UE 102 to enter the low-power mode corresponding to low battery power level))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, and Stauffer. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, and Stauffer to reduce power consumption, conserve battery life, and decrease UE temperature “The base station can receive the request and dictate a change in the power mode of the UE to a low-power mode to reduce power consumption, conserve battery life, and/or decrease the UE temperature. Alternatively, the request can be sent based on a user input selecting a particular power mode and the base station can honor that request by causing the UE to enter the particular power mode. In this way, flexibility is provided to the UE (and the user of the UE) to change the power mode of the UE to reduce power consumption.” (Stauffer Paragraph 0004).
He in view of Tullberg in further view of McCourt in further view of Stauffer does not teach or a UE-initiated activation comprising:
receiving, from the UE, a second activation request to activate the at least one machine [learning based functionality]; and
causing transmission of a second activation request response to the UE to approve the activation;
However, Felt does teach or a UE-initiated activation comprising: (Felt Column 17 Line 41; “FIG. 11 illustrates an exemplary mobile wireless communication device activation method 1100.” Examiner notes that Fig 11 shows a UE initiated activation)
receiving, from the UE, a second activation request to activate the at least one machine [learning based functionality]; and (Felt Column 17 Line 52; “a system receives a request to initiate an activation process to activate a mobile wireless communication device… the system receives the request by way of the mobile wireless communication device that is not currently activated on the mobile wireless communication network.” Examiner notes that the system receives from the UE (mobile wireless communication device) a second activation request to activate the at least one machine (request to initiate an activation process to activate a mobile wireless communication device))
causing transmission of a second activation request response to the UE to approve the activation; (Felt Column 17 Line 64; “the system establishes an activation session between the authenticated mobile wireless communication device” Felt Column 18 Line 5; “the system activates the authenticated mobile wireless communication device on the mobile wireless communication network” Examiner notes that the system causes a transmission of a second activation request response (activation request to activate the communication device through activation session) to the UE (communication device) to approve the activation (establishing an activation session is approving of the activation))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, and Felt. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, and Felt to reduce the amount of resources committed, reduce the chance of errors occurring, and increase convenience/flexibility in activation process “systems and methods described herein may reduce the amount of resources (e.g., equipment and/or human resources) that a service provider must commit to support activation of mobile devices, may reduce the chance of errors occurring during activation of mobile devices, and/or may provide increased convenience and/or flexibility in activation processes.” (Felt Column 3 Line 15).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Felt does not teach in response to the activation, causing transmission of a machine learning inference reporting configuration message to the UE, the machine learning inference reporting configuration message comprising configuration information for at least one of a reporting periodicity or a content of an inference report;
However, Vu does teach in response to the activation, causing transmission of a machine learning inference reporting configuration message to the UE, the machine learning inference reporting configuration message comprising configuration information for at least one of a reporting periodicity or a content of an inference report; (Vu Paragraph 0068; “The figure includes a client 902 with a client inference engine 904 and a machine learning inference service 910 with a server inference engine 912. The client inference engine 904 includes CM 216 and server inference engine 912 includes SM 218. The client inference engine 904 receives client input data 906, performs an inference on CM 216 and sends the output, outputK, to the server inference engine 912. The server inference engine uses outputK to perform an inference on SM 218 and sends the results, outputO, to the client inference engine which reports the inference results 908 to the user.” Examiner notes server inference engine causes transmission/sends a machine learning interface reporting configuration to the UE (sends the results, outputO, to the client inference engine which reports the inference results 908 to the user), the machine learning inference reporting configuration message comprises content of an inference report (inference results 908))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Felt, and Vu. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. Vu teaches a method and system for training a neural network. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Felt, and Vu to improve the performance of training to obtain an improved neural network model “Another advantage is that the workload of training of or inference using the neural network can be balanced between a client and a server training/inference engine, thereby improving the performance of training of or inference using the neural network. Yet another advantage is that intermediate data sent between a client training/inference engine and a server training/inference engine in the training/inference process does not include sensitive information, which helps to maintain the privacy of the original data. Yet another advantage is that bandwidth used over a communication link between the client training/inference engine and the server training/inference engine can be reduced, thereby improving the performance of the training of or inference using the neural network. Yet another advantage is that power consumption of edge devices in a cloud containing the server training engine can be reduced by assigning a large portion of the computation to the cloud.” (Vu Paragraph 0004).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Felt in further view of Vu does not teach continuously monitoring a quality of inference reports received from the UE,
However, Orozco does teach continuously monitoring a quality of inference reports received from the UE, (Orozco Column 11 Line 20; “Data structures 102 may be periodically accessed to determine the accuracy of the response outputs 118 stored therein, such as at a time when a data structure 102 is received from a computing device.” Examiner notes that a quality of inference reports (accuracy of the response outputs) received from the UE (received from a computing device) is continuously monitored (periodically accessed to determine))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, and Orozco. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, and Orozco to make the system more efficient by reducing the time and computational resources used “storing certain inputs and corresponding outputs in a data structure for future use, then using the outputs stored in the data structure rather than recomputing an output fora subsequent request may significantly reduce the time and computational resources used to generate a response.” (Orozco Column 4 Line 1).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Felt in further view of Vu in further view of Orozco does not teach and based on the monitored quality, performing a Radio Resource Management (RRM) action;
de-allocating network resources previously allocated for receiving the inference reports from the UE.
However, Faxer does teach and based on the monitored quality, performing a Radio Resource Management (RRM) action; (Faxer Paragraph 0088; “if a certain terminal device 300 has good radio conditions and strong uplink signals, or it has a relatively lower QoS requirement, then, a small adjacent interference might be considered as less severe. The scheduling restriction (as defined by the radio resource management action) might be applied per signal/channel basis.” Examiner notes that based on the monitored quality (QoS/Quality of Service), a RRM action is performed (radio resource management action is applied))
de-allocating network resources previously allocated for receiving the inference reports from the UE. (Faxer Paragraph 0088; “if a certain terminal device 300 has good radio conditions and strong uplink signals, or it has a relatively lower QoS requirement, then, a small adjacent interference might be considered as less severe. The scheduling restriction (as defined by the radio resource management action) might be applied per signal/channel basis.” Examiner notes that the network resources previously allocated for receiving the inference reports from the UE is de allocated (the scheduling restriction is applied per signal/channel basis))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, and Faxer. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. Faxer teaches a method for cross link interference handling. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, and Faxer to improve the performance of services and reduce interference “One or more of the various embodiments improve the performance of OTT services provided to UE 530 using OTT connection 550, in which wireless connection 570 forms the last segment. More precisely, the teachings of these embodiments may reduce interference, due to improved classification ability of airborne UEs which can generate significant interference.” (Faxer Paragraph 0151).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Felt in further view of Vu in further view of Orozco in further view of Faxer does not teach a number of detected errors has exceeded a predetermined threshold within a given time window;
However, Belghoul does teach a number of detected errors has exceeded a predetermined threshold within a given time window; (Belghoul Paragraph 0119; “The UEs may additionally send error metric(s), other meta data, and/or time stamps.” Examiner notes that sending error metrics along with time stamps is an indication that a number of detected errors has exceeded a predetermined threshold within a given time window; predetermined threshold can be if a single error has occurred);
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, Faxer, and Belghoul. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. Faxer teaches a method for cross link interference handling. Belghoul teaches Techniques for optimizing and deploying convolutional neural network (CNN) machine learning models for inference using integrated graphics processing units are described. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, Faxer, and Belghoul to implement the techniques for shorter latency, less burden on network bandwidth, and better privacy protection for users “it is becoming more and more desirable to execute the model inference directly at the edge devices for shorter latency, less burden of the network bandwidth, and better privacy protection to the users.” (Belghoul Column 1 Line 18).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Felt in further view of Vu in further view of Orozco in further view of Faxer in further view of Belghoul does not teach a request to restart the machine learning based assistance capability;
and a request to switch to another configured machine learning based assistance;
However, Wang does teach a request to restart the machine learning based assistance capability; (Wang Page 23 Paragraph 4; “In response, in some embodiments, the user, via the user device 702, can transmit a request to the model training system 120 to modify the machine learning model being trained (for example, transmit a modification request)… For example, the model training system 120 can cause the virtual machine instance 722 to optionally delete an existing ML training container 730, create and initialize a new ML training container 730 using some or all of the information included in the request, and execute the code 737 stored in the new ML training container 730 to restart the machine learning model training process.” Examiner notes that the user device requests to restart the machine learning based assistance capability through restarting the training process of the machine learning model.);
and a request to switch to another configured machine learning based assistance; (Wang Page 23 Paragraph 4; “In response, in some embodiments, the user, via the user device 702, can transmit a request to the model training system 120 to modify the machine learning model being trained (for example, transmit a modification request). The request can include a new or modified container image, a new or modified algorithm, new or modified hyperparameter(s), and/or new or modified information describing the computing machine on which to train a machine learning model. The model training system 120 can modify the machine learning model accordingly. For example, the model training system 120 can cause the virtual machine instance 722 to optionally delete an existing ML training container 730, create and initialize a new ML training container 730 using some or all of the information included in the request, and execute the code 737 stored in the new ML training container 730 to restart the machine learning model training process.” Examiner notes that the request to restart the machine learning based assistance capability is also a request to switch to another configured machine learning based assistance/newly trained machine learning model);
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, Faxer, Belghoul, and Wang. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. Faxer teaches a method for cross link interference handling. Belghoul teaches Techniques for optimizing and deploying convolutional neural network (CNN) machine learning models for inference using integrated graphics processing units are described. Wang teaches sending out requests to restart and switch machine learning models based on cause indications to optimize the model to run on integrated graphics processing units. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, Faxer, Belghoul, and Wang to have a user equipment that comprises machine learning capabilities to perform machine learning functionalities and to restart and switch machine learning models based on cause indications “Thus, a user device 702 can refer to trained machine learning model(s) stored in the ML scoring container(s) 750 using the endpoint. This allows for the network address of an ML scoring container 750 to change without causing the user operating the user device 702 to change the way in which the user refers to a trained machine learning model.” (Wang Page 23 Paragraph 3).
Regarding Claim 83, He teaches A system comprising: a network node; a processor; a transceiver; and a memory comprising computer-executable instructions that, when executed by the processor, cause the network node to perform the following operations: (He Paragraph 0044; “UE device 110 and machine learning system 140 may each include one or more devices 200 or components of device 200. As shown in FIG. 2, device 200 may include a bus 210, a processor 220, a memory 230, an input device 240, an output device 250, and a communication interface 260.” He Paragraph 0049; “Communication interface 260 may include a transceiver that enables device 200 to communicate with other devices and/or systems via wireless communications” He Claim 20; “A non-transitory computer-readable memory device storing instructions executable by a processor…” Examiner notes that a apparatus (machine learning system 140) comprises a processor (processor 220), a transceiver (communication interface 260), and a memory comprising executable instructions)
receiving, via the transceiver, from a User Equipment (UE), UE capabilities comprising a machine learning based assistance capability, wherein the machine learning based assistance capability comprises a plurality of machine learning based functionalities, each machine learning based functionality of the plurality of machine learning based functionalities comprising a machine learning entity and at least one machine learning mode associated with the machine learning entity, (He Paragraph 0025; “Different machine learning modules may use different algorithms and/or different parameters for the algorithms to make decisions, predictions, and/or inferences. The smart engine may make adjustments to data sources and/or machine learning models/classifiers/algorithms used by a particular machine learning module based on the determined device status.” He Paragraph 0056; “Smart engine 330 may control the operation of machine learning processes on UE device 110. Smart engine 330 may obtain the device status of UE device 110… applications currently running on UE device 110, machine learning processes scheduled to run… whether to communicate with machine learning system 140 to perform a machine learning process” He Paragraph 0054; “Applications 310 may include applications running on UE device 110 that may make use of machine learning. For example, applications 310 may include … another type of application that may utilize machine learning.” He Paragraph 0061; “Machine learning modules 350 may include machine learning modules for particular applications. For example, a first machine learning module may be trained to perform behavior classification, a second machine learning module may be trained to perform object recognition in images” Examiner notes that machine learning system receives from a UE (smart engine in UE communicates with machine learning system) UE capabilities (applications) comprising a machine learning based assistance capability (applications make use of machine learning) comprises a plurality of machine learning based functionalities (plurality of applications include various functionalities as seen in Paragraph 0054); each machine learning based functionality comprises a machine learning entity (machine learning module for particular application) and at least one machine learning mode associated with machine learning entity (different machine learning models used in machine learning module))
the at least one machine learning mode of each machine learning based functionality comprises a fallback operating mode or a primary machine learning mode, (He Paragraph 0024; “Different machine learning models may be associated with different expected resource use. For example, a first machine learning model may use less battery power, processor time, memory, and/or network bandwidth, etc., and a second machine learning model may use more battery power, processor time, memory, and/or network bandwidth, etc.” Examiner notes that the at least one machine learning mode (machine learning model) of each machine learning based functionality comprises a fallback operating mode (a first machine learning model may use less battery power, processor time, memory, and/or network bandwidth, etc.) or a primary machine learning mode (a second machine learning model may use more battery power, processor time, memory, and/or network bandwidth, etc))
and the fallback operating mode applies when the primary machine learning mode is unavailable due to at least one of: an absence of input data required by the primary machine learning mode, a deactivation of the primary machine learning mode, or a measurement error, or a connectivity error; (He Paragraph 0024; “The device status may indicate a current resource capacity of the wireless communication device, measured by parameters including, for example, the battery power level, a measure of a processor load, a measure of memory use, a measure of a network connection quality, and/or other types of measures of the resource capacity of the wireless communication device. Different machine learning models may be associated with different expected resource use.” Examiner notes that the fallback operating mode (different machine learning model) is applied when the primary machine learning mode is unavailable due to connectivity error (bad connection quality))
initiating, with the UE, a communication session [for an initial exploration and tuning of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability]; (He Fig 9 and Paragraph 0119; “FIG. 9, signal flow 900 may include a user opening application 410 via user interface 910 (signal 912).” Examiner notes that a communication session is initiated with the UE as shown in fig 9)
PNG
media_image1.png
487
643
media_image1.png
Greyscale
activating at least one machine learning based functionality among the plurality of machine learning based functionalities, (He Paragraph 0119; “FIG. 9 is a diagram of an exemplary signal flow 900 according to an implementation described herein… For example, assume the user activates a financial application and the financial application selects, in addition to authenticating the user via a login process, to perform behavioral authentication using a machine learning module trained to perform behavioral authentication based on the user's behavior patterns. Thus, application 410 may call an API to request a machine learning service to perform behavioral authentication via smart engine 420 (signal 914). Smart engine 420 may then obtain device status data from device status module 430 (signals 916 and 918)” Examiner notes that Fig 9 shows activating at least one machine learning based functionality amount the plurality of machine learning based functionalities (perform behavioral authentication using a machine learning module), wherein the activation comprises a network initiated activation (signal flow))
He does not teach each machine learning entity is configured for at least one of a handover (HO) prediction or a quality of service (QoS) prediction,
However, Tullberg does teach each machine learning entity is configured for at least one of a handover (HO) prediction or a quality of service (QoS) prediction, (Tullberg Paragraph 0072; “The dedicated machine learning model may accurately learn the propagation environment and the prediction of handover of UE to neighboring cells.” Examiner notes that machine learning entity (machine learning model) is configured for handover prediction (prediction of handover))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He and Tullberg. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. One of ordinary skill would have motivation to combine He and Tullberg to improve the over performance of the system by implementing a machine learning method “a machine learning method for cell handover reduces overhead, delay and handover failures. This improves the overall performance of the system and aid users close to the cell edge.” (Tullberg Paragraph 0025).
He in view of Tullberg does not teach session for an initial exploration and tuning of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability
However McCourt does teach session for an initial exploration and tuning of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability (McCourt Paragraph 0076; “S220 may function to dynamically set and/or configure exploration and exploitation parameters for a tuning associated with each partial tuning task of a multi-task tuning request. Exploration parameters preferably enable the tuning service to identify potential hyperparameter values for a given model.” Examiner notes that session for initial exploration and tuning (configure exploration and exploitation parameters for a tuning) of one or more of the plurality of machine learning based functionalities included within the received machine learning based assistance capability (given model; tuning hypermeters of model that machine learning based functionalities is executed with))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, and McCourt. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. One of ordinary skill would have motivation to combine He, Tullberg, and McCourt to improve the overall performance of a model with saving computational resources “The embodiments of the present application, however, provide an intelligent optimization platform that functions to optimize hyperparameters and/or parameters of any type of model with significantly fewer evaluation thereby saving computational resources while greatly improving an overall performance of a model.” (McCourt Paragraph 0033).
He in view of Tullberg in further view of McCourt does not teach response comprises a cause for the failure, the cause comprising at least one of: an indication of an alternative preferred machine learning based functionality, an indication to use a fallback operating mode, an indication of insufficient battery power, or an indication of a change in channel radio conditions that exceeds a channel condition threshold
receiving via the transceiver, from the UE, a change request [to change the activated at least one machine learning based functionality], wherein the change request comprises a cause indication for the change, the cause indication comprising at least one of the following:
the activated at least one machine [learning based functionality] cannot be run anymore due to at least one of: low battery power, low computational power, and detected changes in input data;
in response to receiving the change request: transmitting, via the transceiver, a machine learning functionality deactivation reply to the UE, the machine learning functionality deactivation reply comprising a confirmation of a change to the activated at least one machine learning based functionality corresponding to the cause indication; and
However, Stauffer does teach response comprises a cause for the failure, the cause comprising at least one of: an indication of an alternative preferred machine learning based functionality, an indication to use a fallback operating mode, an indication of insufficient battery power, or an indication of a change in channel radio conditions that exceeds a channel condition threshold (Stauffer Paragraph 0004; “the user device can autonomously provide, to the base station, a request to enter a low-power mode based on local factors, such as a low battery power level or a high temperature of the UE. The base station can receive the request and dictate a change in the power mode of the UE to a low-power mode to reduce power consumption, conserve battery life, and/or decrease the UE temperature.” Examiner notes that UE cannot perform action, so a response (request) is sent to base station, cause for failure comprises an indication of insufficient battery power (low battery power))
receiving via the transceiver, from the UE, a change request [to change the activated at least one machine learning based functionality], wherein the change request comprises a cause indication for the change, the cause indication comprising at least one of the following: (Stauffer Paragraph 0004; “the user device can autonomously provide, to the base station, a request to enter a low-power mode based on local factors, such as a low battery power level or a high temperature of the UE. The base station can receive the request and dictate a change in the power mode of the UE to a low-power mode to reduce power consumption, conserve battery life, and/or decrease the UE temperature.” Examiner notes a base station receives from the UE (user device) a change request (a request to enter a low-power mode), wherein the change request comprises a cause indication (change request to enter lower power mode is a cause indication))
the activated at least one machine [learning based functionality] cannot be run anymore due to at least one of: low battery power, low computational power, and detected changes in input data; (Examiner refers to previous mapping to show that the activated machine (user device) cannot be run anymore due to low battery power (low battery power level));
in response to receiving the change request: transmitting, via the transceiver, a machine learning functionality deactivation reply to the UE, the machine learning functionality deactivation reply comprising a confirmation of a change to the activated at least one machine learning based functionality corresponding to the cause indication; and (Stauffer Paragraph 0034; “a downlink (DL) signal is detected that includes an FLPM acknowledgment (ACK) from the base station. For example, the UE 102 can detect a DL signal from the base station 104 via the wireless link 106. In at least one example, the DL signal includes instructions to cause the UE 102 to enter the low-power mode. The low-power mode may be the specifically requested low-power mode, such as the inactive mode.” Examiner notes that a machine learning functionality deactivation reply to the UE is transmitted (User device detects a downlink (DL) signal that includes an FLPM acknowledgment (ACK) from the base station); the reply comprises a confirmation of a change to the active machine learning based functionality corresponding to the cause indication (DL signal includes instructions to cause the UE 102 to enter the low-power mode corresponding to low battery power level))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, and Stauffer. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, and Stauffer to reduce power consumption, conserve battery life, and decrease UE temperature “The base station can receive the request and dictate a change in the power mode of the UE to a low-power mode to reduce power consumption, conserve battery life, and/or decrease the UE temperature. Alternatively, the request can be sent based on a user input selecting a particular power mode and the base station can honor that request by causing the UE to enter the particular power mode. In this way, flexibility is provided to the UE (and the user of the UE) to change the power mode of the UE to reduce power consumption.” (Stauffer Paragraph 0004).
He in view of Tullberg in further view of McCourt in further view of Stauffer does not teach wherein the activation comprise a UE-initiated activation comprising:
receiving, from the UE, a second activation request to activate the at least one machine [learning based functionality]; and
causing transmission of a second activation request response to the UE to approve the activation;
However, Felt does teach or a UE-initiated activation comprising: (Felt Column 17 Line 41; “FIG. 11 illustrates an exemplary mobile wireless communication device activation method 1100.” Examiner notes that Fig 11 shows a UE initiated activation)
receiving, from the UE, a second activation request to activate the at least one machine [learning based functionality]; and (Felt Column 17 Line 52; “a system receives a request to initiate an activation process to activate a mobile wireless communication device… the system receives the request by way of the mobile wireless communication device that is not currently activated on the mobile wireless communication network.” Examiner notes that the system receives from the UE (mobile wireless communication device) a second activation request to activate the at least one machine (request to initiate an activation process to activate a mobile wireless communication device))
causing transmission of a second activation request response to the UE to approve the activation; (Felt Column 17 Line 64; “the system establishes an activation session between the authenticated mobile wireless communication device” Felt Column 18 Line 5; “the system activates the authenticated mobile wireless communication device on the mobile wireless communication network” Examiner notes that the system causes a transmission of a second activation request response (activation request to activate the communication device through activation session) to the UE (communication device) to approve the activation (establishing an activation session is approving of the activation))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, and Felt. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, and Felt to reduce the amount of resources committed, reduce the chance of errors occurring, and increase convenience/flexibility in activation process “systems and methods described herein may reduce the amount of resources (e.g., equipment and/or human resources) that a service provider must commit to support activation of mobile devices, may reduce the chance of errors occurring during activation of mobile devices, and/or may provide increased convenience and/or flexibility in activation processes.” (Felt Column 3 Line 15).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Felt does not teach in response to the activation, causing transmission of a machine learning inference reporting configuration message to the UE, the machine learning inference reporting configuration message comprising configuration information for at least one of a reporting periodicity or a content of an inference report;
However, Vu does teach in response to the activation, causing transmission of a machine learning inference reporting configuration message to the UE, the machine learning inference reporting configuration message comprising configuration information for at least one of a reporting periodicity or a content of an inference report; (Vu Paragraph 0068; “The figure includes a client 902 with a client inference engine 904 and a machine learning inference service 910 with a server inference engine 912. The client inference engine 904 includes CM 216 and server inference engine 912 includes SM 218. The client inference engine 904 receives client input data 906, performs an inference on CM 216 and sends the output, outputK, to the server inference engine 912. The server inference engine uses outputK to perform an inference on SM 218 and sends the results, outputO, to the client inference engine which reports the inference results 908 to the user.” Examiner notes server inference engine causes transmission/sends a machine learning interface reporting configuration to the UE (sends the results, outputO, to the client inference engine which reports the inference results 908 to the user), the machine learning inference reporting configuration message comprises content of an inference report (inference results 908))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Felt, and Vu. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. Vu teaches a method and system for training a neural network. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Felt, and Vu to improve the performance of training to obtain an improved neural network model “Another advantage is that the workload of training of or inference using the neural network can be balanced between a client and a server training/inference engine, thereby improving the performance of training of or inference using the neural network. Yet another advantage is that intermediate data sent between a client training/inference engine and a server training/inference engine in the training/inference process does not include sensitive information, which helps to maintain the privacy of the original data. Yet another advantage is that bandwidth used over a communication link between the client training/inference engine and the server training/inference engine can be reduced, thereby improving the performance of the training of or inference using the neural network. Yet another advantage is that power consumption of edge devices in a cloud containing the server training engine can be reduced by assigning a large portion of the computation to the cloud.” (Vu Paragraph 0004).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Felt in further view of Vu does not teach continuously monitoring a quality of inference reports received from the UE,
However, Orozco does teach continuously monitoring a quality of inference reports received from the UE, (Orozco Column 11 Line 20; “Data structures 102 may be periodically accessed to determine the accuracy of the response outputs 118 stored therein, such as at a time when a data structure 102 is received from a computing device.” Examiner notes that a quality of inference reports (accuracy of the response outputs) received from the UE (received from a computing device) is continuously monitored (periodically accessed to determine))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, and Orozco. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, and Orozco to make the system more efficient by reducing the time and computational resources used “storing certain inputs and corresponding outputs in a data structure for future use, then using the outputs stored in the data structure rather than recomputing an output fora subsequent request may significantly reduce the time and computational resources used to generate a response.” (Orozco Column 4 Line 1).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Felt in further view of Vu in further view of Orozco does not teach and based on the monitored quality, performing a Radio Resource Management (RRM) action;
de-allocating network resources previously allocated for receiving the inference reports from the UE.
However, Faxer does teach and based on the monitored quality, performing a Radio Resource Management (RRM) action; (Faxer Paragraph 0088; “if a certain terminal device 300 has good radio conditions and strong uplink signals, or it has a relatively lower QoS requirement, then, a small adjacent interference might be considered as less severe. The scheduling restriction (as defined by the radio resource management action) might be applied per signal/channel basis.” Examiner notes that based on the monitored quality (QoS/Quality of Service), a RRM action is performed (radio resource management action is applied))
de-allocating network resources previously allocated for receiving the inference reports from the UE. (Faxer Paragraph 0088; “if a certain terminal device 300 has good radio conditions and strong uplink signals, or it has a relatively lower QoS requirement, then, a small adjacent interference might be considered as less severe. The scheduling restriction (as defined by the radio resource management action) might be applied per signal/channel basis.” Examiner notes that the network resources previously allocated for receiving the inference reports from the UE is de allocated (the scheduling restriction is applied per signal/channel basis))
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, and Faxer. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. Faxer teaches a method for cross link interference handling. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, and Faxer to improve the performance of services and reduce interference “One or more of the various embodiments improve the performance of OTT services provided to UE 530 using OTT connection 550, in which wireless connection 570 forms the last segment. More precisely, the teachings of these embodiments may reduce interference, due to improved classification ability of airborne UEs which can generate significant interference.” (Faxer Paragraph 0151).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Felt in further view of Vu in further view of Orozco in further view of Faxer does not teach a number of detected errors has exceeded a predetermined threshold within a given time window;
However, Belghoul does teach a number of detected errors has exceeded a predetermined threshold within a given time window; (Belghoul Paragraph 0119; “The UEs may additionally send error metric(s), other meta data, and/or time stamps.” Examiner notes that sending error metrics along with time stamps is an indication that a number of detected errors has exceeded a predetermined threshold within a given time window; predetermined threshold can be if a single error has occurred);
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, Faxer, and Belghoul. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. Faxer teaches a method for cross link interference handling. Belghoul teaches Techniques for optimizing and deploying convolutional neural network (CNN) machine learning models for inference using integrated graphics processing units are described. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, Faxer, and Belghoul to implement the techniques for shorter latency, less burden on network bandwidth, and better privacy protection for users “it is becoming more and more desirable to execute the model inference directly at the edge devices for shorter latency, less burden of the network bandwidth, and better privacy protection to the users.” (Belghoul Column 1 Line 18).
He in view of Tullberg in further view of McCourt in further view of Stauffer in further view of Felt in further view of Vu in further view of Orozco in further view of Faxer in further view of Belghoul does not teach a request to restart the machine learning based assistance capability;
and a request to switch to another configured machine learning based assistance;
However, Wang does teach a request to restart the machine learning based assistance capability; (Wang Page 23 Paragraph 4; “In response, in some embodiments, the user, via the user device 702, can transmit a request to the model training system 120 to modify the machine learning model being trained (for example, transmit a modification request)… For example, the model training system 120 can cause the virtual machine instance 722 to optionally delete an existing ML training container 730, create and initialize a new ML training container 730 using some or all of the information included in the request, and execute the code 737 stored in the new ML training container 730 to restart the machine learning model training process.” Examiner notes that the user device requests to restart the machine learning based assistance capability through restarting the training process of the machine learning model.);
and a request to switch to another configured machine learning based assistance; (Wang Page 23 Paragraph 4; “In response, in some embodiments, the user, via the user device 702, can transmit a request to the model training system 120 to modify the machine learning model being trained (for example, transmit a modification request). The request can include a new or modified container image, a new or modified algorithm, new or modified hyperparameter(s), and/or new or modified information describing the computing machine on which to train a machine learning model. The model training system 120 can modify the machine learning model accordingly. For example, the model training system 120 can cause the virtual machine instance 722 to optionally delete an existing ML training container 730, create and initialize a new ML training container 730 using some or all of the information included in the request, and execute the code 737 stored in the new ML training container 730 to restart the machine learning model training process.” Examiner notes that the request to restart the machine learning based assistance capability is also a request to switch to another configured machine learning based assistance/newly trained machine learning model);
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, Faxer, Belghoul, and Wang. He teaches a method for managing machine learning involving mobile devices. Tullberg teaches a method for use in a network node for predicting handover using machine learning. McCourt teaches method for accelerated tuning of hyperparameters. Stauffer teaches techniques and systems for 5G NR fast low-power mode. Felt teaches a mobile wireless communications device activation system. Vu teaches a method and system for training a neural network. Orozco teaches a system for reducing computations performed for iterative processes. Faxer teaches a method for cross link interference handling. Belghoul teaches Techniques for optimizing and deploying convolutional neural network (CNN) machine learning models for inference using integrated graphics processing units are described. Wang teaches sending out requests to restart and switch machine learning models based on cause indications to optimize the model to run on integrated graphics processing units. One of ordinary skill would have motivation to combine He, Tullberg, McCourt, Stauffer, Felt, Vu, Orozco, Faxer, Belghoul, and Wang to have a user equipment that comprises machine learning capabilities to perform machine learning functionalities and to restart and switch machine learning models based on cause indications “Thus, a user device 702 can refer to trained machine learning model(s) stored in the ML scoring container(s) 750 using the endpoint. This allows for the network address of an ML scoring container 750 to change without causing the user operating the user device 702 to change the way in which the user refers to a trained machine learning model.” (Wang Page 23 Paragraph 3).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DANIEL DUC TRAN whose telephone number is (571)272-6870. The examiner can normally be reached Mon-Fri 8:00-5:00 EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Viker Lamardo can be reached at (571) 270-5871. 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.
/D.D.T./Examiner, Art Unit 2147
/VIKER A LAMARDO/Supervisory Patent Examiner, Art Unit 2147
Prosecution Insights