DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
This office action is in response to amendment/reconsideration filed on 06/26/2026, the amendment/reconsideration has been considered. Claims 3, 4, 7, 8, 11, 18 and 22 have been amended. Claims 1-11, 13, 16-20 and 22-23 are pending for examination as cited below.
Response to Arguments
Applicant’s arguments with respect to claim(s) amendments filed on 6/26/2026 with respect to the applied art “Yip” is persuasive and have been considered. Upon further search and consideration, a newly applied art rejections is hereby cited as below.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-9, is/are rejected under 35 U.S.C. 103 as being unpatentable over Karjee et al. (Pub. No.: US 2022/0311678 A1), hereinafter “Kar” in view of Liu et al. (Pub. No.: US 2022/0012607 A1), hereinafter “Liu”.
As to claim 1. Kar discloses, A method performed by a user equipment (Kar, Abstract), the method comprising:
transmitting towards an application function, (AF) a request for splitting a machine learning (ML) inference process (Kar [0033], [0039], the IoT/UE selecting an edge and identifying a network, and then splitting the DNN and transmitting the second part to the edge.), wherein the request comprises any one or more of:
information about the UE, information about the ML inference process, and/or a request for information about a network to which the UE is connected (Kar, [0033], [0046, [0047], the specification describes using network condition information, bandwidth, historical inference time records, and benchmarking of layer inference times as inputs to select network and compute split ration); and
after transmitting the request for splitting the ML inference process, receiving split decision information indicating how to split the ML inference process, wherein the split decision information was transmitted by the AF (Kar, [0110], [0111], the IoT device computing the optimal split ration locally (DSC/E-DSC) and /or exchanging model/pipeline configuration via gRPC.).
Kar however is silent to disclose, an AF making and transmitting a split-decision back to the UE as a distinct network entity.
Liu discloses a similar concept in the same field of endeavor including, network/AF decisioning role (Liu, fig.4, step 403-408, [0043], the model scheduler is the decision-making entity: it generates the scheduling plan, compiles/model partitions the inference program, and dispatches the partitions to executers. Thus, it transmits the information implementing the split decision.).
Therefore, before the effective filing date of the instant application it would have been obvious to one of the skilled in the art to incorporate the teachings of “Liu” into those of “Kar” to provide a method for managing artificial intelligence model partitions for execution in an information processing system with edge computing resources. Such method uses the following steps. An intermediate representation of an artificial intelligence model is obtained. A computation graph is generated based on the intermediate representation. The computation graph is partitioned into a set of partitions. The method then schedules the set of partitions for respective execution on a set of computing devices in an edge computing environment, and causes deployment of the set of partitions respectively to the set of computing devices for execution in the edge computing environment.
As to claim 2. The combined system of Kar and Liu discloses the invention as in parent claim above including, wherein the information about the UE indicates a location of the UE and/or information about one or more resources available at the UE (Kar, [0033], [0034]), the information about the ML inference process indicates :
i) one or more requirements on resources needed for performing the ML inference process (Kar, [0033], [0034], resources available at UE: disclosed as device capability / low-cost computing device and use of computational latency / device inference time records.);
ii) a size of intermediate output data to be generated during the ML inference process (Kar, [0033], [0039], figs 4A and 6A-6C, output of each layer and transmitting layer outputs to edge.);
iii) a time duration needed for performing the ML inference process (Kar, [0033], [0039], fig. 7A-7F, inference time/per-layer timing/inference time records used to compute split ratio and optimal split point.); and/or
iv) an accuracy requirement of the ML inference process (Kar, [0033], [0039], various attributes significant to accuracy), and the information about the network indicates any one or more of:
a rate of uplink (UL) data transmission, a rate of downlink (DL) data transmission, a network latency, and/or a network reliability (Kar, [0033], [0034], figs.5B/7D, network metrics: bandwidth, network channel quality, available bandwidth, historical inference time records, and transmission time are explicitly used to select network and compute split.).
As to claim 3. The combined system of Kar and Liu discloses the invention as in parent claim above including, the method further comprising: based on the received split decision information, selecting a part of the ML inference process (Liu, [0022]); and
performing the selected part of the ML inference process (Kar, [0039] and [0110]-[0111]).
As to claim 4. The combined system of Kar and Liu discloses the invention as in parent claim above including, transmitting towards one or more network end points (NEs) ML sub-process data indicating a part of the ML inference process to be performed by said one or more NEs (Kar, [0039], splitting a plurality of layers of the DNN into a firs part and a second part based on the determined split ratio, and transmitting the second part to the selected at least one edge device through the identified network. [0110], [0111], the IoT device 605 sends the result621 of partial inference to the edge device 605. The edge device 605 executes the remaining PoseNet model inference and sends the result 615 back.).
As to claim 5. Is rejected for same rationale as applied to claim 1 above.
As to claim 6. Is rejected for same rationale as applied to claim 2 above.
As to claim 7. Is rejected for same rationale as applied to claim 3 above.
As to claim 8. Is rejected for same rationale as applied to claim 4 above.
As to claim 9. The combined system of Kar and Liu discloses the invention as in parent claim above including, wherein the NE information indicates: an amount of computational resources available at said one or more NEs, and/or end-to-end network performance between one or more pairs of NEs in case said one or more NEs includes more than one NE (Kar, [0033] selects at least one edge device from a plurality of devices within a communication range of the IoT devices based on network conditions and computational latency associated with the plurality of edge devices. [0034], the preferred network 105is identified based on available bandwidth and historical inference time records of a plurality of networks associated with the IoT device.).
Claim(s) 10 and 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Karjee et al. (Pub. No.: US 2022/0311678 A1), hereinafter “Kar” in view of Liu et al. (Pub. No.: US 2022/0012607 A1), hereinafter “Liu” and further in view of “3GPP TS 23.288 v16.40”, hereinafter “3GPP”.
As to claim 10. The combined system of Kar and Liu discloses the invention as in parent claim above. Kar and Liu however are silent to disclose explicitly, transmitting towards a network data analytics function (NWDAF) data indicating the NE information.
3GPP however discloses a similar concept in the same field of endeavor, including, transmitting towards a network data analytics function (NWDAF) data indicating the NE information (3GPP, section 4.2, page.8, fig.4.2-1, the specification defines the Nnf interface specifically for NWDAF data delivery subscription, cancellation, and request for particular context data reports. An NF that has subscribed data delivery transmits the relevant data/report to NWDAF.).
Therefore, before the effective filing date of the instant application it would have been obvious to one of the skilled in the art to incorporate the teachings of “3GPP” into those of “Kar and Liu” to provide a 5G system architecture that allows NWDAF to retrieve the management data from OAM by invoking OAM services.
As to claim 11. The combined system of Kar, Liu and 3GPP discloses the invention as in parent claim above including, wherein the data indicating the NE information is transmitted as a result of the NWDAF subscribing to the AF for the data or transmitting to the AF a request for the data (3GPP, Section 4.1, “General”; section 6.2.2.3 page.17 “procedure for data collection from AF via NEF”).
Claim(s) 13, 16, and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over “Kar” and “Liu” in view of Lee et al. (Pub. No.: US 20200322821 A1), hereinafter “Lee”.
As to claim 13. The combined system of Kar and Liu discloses the invention substantially as in claims above. Kar and Liu however are silent to disclose explicitly, further comprising receiving analytic data of said one or more types identified by said one or more analytic type identifiers, wherein the analytic data is generated based on the NE information.
Lee discloses a similar concept in the same field of endeavor including, receiving analytic data of said one or more types identified by said one or more analytic type identifiers, wherein the analytic data is generated based on the NE information. (Lee, [0062], [0063], The NWDAF (Network Data Analytics Function) provides analytics to 5GC NFs, and OAM. Analytics information are either statistical information of the past events, or predictive information.).
Therefore, before the effective filing date of the instant application it would have been obvious to one of the skilled in the art to incorporate the teachings of “Lee” into those of “Kar and Liu” to provide a network data collection method from application function device for network data analytic function is disclosed. The network data collection method includes transmitting a Naf_EventExposure_Subscribe message for a subscription of an event into an application function (AF) device; and receiving a Naf_EventExposure_Notify message from the AF device when the NWDAF subscribes the event.
As to claim 16. The combined system of Kar, Liu and Lee discloses the invention substantially including, wherein the analytic data indicates:
historical statistics and/or predictions regarding UL data transmission from the UE to each of said one or more NEs (Kar, [0034]);
historical statistics and/or predictions regarding packet delay on UL data transmission from the UE to each of said one or more NEs (Kar, [0033], [0034] and [0039]).;
historical statistics and/or predictions regarding packet loss rate on UL data transmission from the UE to each of said one or more NEs (Kar, [0049], AI model schedular encompasses the various parameter to find the optimal functionality based on the available infrastructure.), ; and/or
a quality of service (QoS) indicator indicating a predicted quality of service in case the ML inference process is split (Liu, [0049], AI model schedular.).
As to claim 17. The combined system of Kar, Liu and Lee discloses the invention substantially including, further comprising determining how to split the ML inference process based on the received analytic data, wherein determining how to split the ML inference process comprises determining:
a number of ML layers for performing a part of the ML inference process at the UE (Kar, fig. 6a-6c, 7a-7f, [0039], determines a split ratio and splits a plurality of DNN layers int a first part (device) and second part (edge)).;
one or more NE identifiers identifying said one or more NEs to perform a part of the ML inference process (Liu, fig.4, step 403-408, [0043], the model scheduler is the decision-making entity: it generates the scheduling plan, compiles/model partitions the inference program, and dispatches the partitions to executers. Thus, it transmits the information implementing the split decision.);
an ML layer identifier identifying an ML layer of which an operation corresponds to the last operation performed by the UE for the ML inference process (Kar, fig. 6a-6c, [0039], transmits layer outputs and partial inference);
an ML layer identifier identifying an ML layer of which an operation corresponds to the last operation performed by one of said one or more NEs (Kar, fig. 6a-6c, [0039], transmits layer outputs and partial inference); and/or
a time period for performing a part of the ML inference process at the UE (Kar, fig.5a,b, 7d-f, [0033], [0039], use inference time, transmission time, latency etc.).
As to claim 18-20 and 22-23, is/are rejected under 35 U.S.C. 103 as being unpatentable over Liu as applied above in view of Che et al. (Pub. No.: US 2020/0175361 A1), hereinafter “Che” and further in view of “3GPP” and further in view of “Che”.
As to claim 18. Liu discloses, a method performed by one or more network endpoints (NEs) (Liu, [0041-0044], [0051], EMC discloses, heterogeneous/IoT devices that provide inference input data, operate at edge location, and are associated with inference program execution.), the method comprising:
generating NE information about said one or more NEs (Liu, fig.3, steps 301-307, [0044] The application request includes information identifying the IoT devices and the inference program.).
Liu however is silent to disclose explicitly, transmitting towards an application function (AF) the generated NE information; and
performing a first part of an machine learning (ML) inference process, wherein the ML inference process is split into the first part and a second part based at least on the NE information.
3GPP discloses a similar concept in the same field of endeavor including, transmitting towards an application function (AF) the generated NE information (3GPP, Section 4.1, section 6.2.2.3, 3GPP teaches an AF as an information source and teaches AF related data collection directly or through NEF. “Data collection based on subscription to events provided by AMF, SMF, PCF, UDM, AF (directly or via NEF) and OAM.”).
Therefore, before the effective filing date of the instant application it would have been obvious to one of the ordinary skilled in the art to incorporate the teachings of “3GPP” into those of “Liu” to provide a 5G system architecture that allows NWDAF to retrieve the management data from OAM by invoking OAM services.
Liu and 3GPP however are silent to disclose explicitly, performing a first part of an machine learning (ML) inference process, wherein the ML inference process is split into the first part and a second part based at least on the NE information.
Che discloses a similar concept in the same field of endeavor including, performing a first part of an machine learning (ML) inference process (Che, [0024-0027], Che expressly teaches an edge device executing a first portion of a DNN inference/data flow graph up to a selected partition point.), wherein the ML inference process is split into the first part and a second part based at least on the NE information (Che, [0002], [0014], [0024-0027], Che teaches partitioning deep learning inference across the edge and cloud. The first portion runs at the edge device; results are transferred across an interconnect; and the second portion resumes at the cloud platform.).
Therefore, before the effective filing date of the instant application it would have been obvious to one of the ordinary skilled in the art to incorporate the teachings of “Che” into those of “Liu and 3GPP” to provide for improving the learning inference performance by partitioning the learning inference based on system fluctuations and available resources by parsing a trained neural network model of a neural network into a data flow graph with a plurality of nodes; generating a traversal order of the data flow graph; assigning a load level range to each edge device, an interconnect connecting the edge device and a cloud computing platform, and the cloud computing platform; profiling performance of each node over the load level range for the edge device and the cloud computing platform; and determining a partition point of the data flow graph based on the profiled performance of each node.
As to claim 19. The combined system of Liu, 3GPP and Che discloses the invention as in parent claim above including, wherein the NE information indicates: an amount of computational resources available at said one or more NEs, and/or end-to-end network performance between one or more pairs of NEs in case said one or more NEs includes more than one NE (Che, [0002], [0014], [0024-0027], the cloud-side portion of the distributed inference arrangement receives the result generated by the edge device’s execution of the first inference portion. The “results from the last node” executed at the edge device are the output of the fist partition/first part of the DNN data-flow graph. The reference further teaches, partitioning the DNN graph across edge and cloud. The intermediate ouput is transmitted precisely because the inference is split at the selected partition point.).
As to claim 20. The combined system of Liu, 3GPP and Che discloses the invention as in parent claim above including, transmitting towards one or more network end points (NEs) ML sub-process data indicating a part of the ML inference process to be performed by said one or more NEs (Che, [0021-0024], The system may include an edge device, an interconnect connecting the edge device and a cloud computing platform, and the cloud computing platform. The edge device and cloud platform are profiled according to discrete workload/resource levels. The selected split location therefore depends on the processing resource condition at the participating endpoint.).
As to claim 22, The combined system of Liu, 3GPP and Che discloses the invention as in parent claim above including, a method performed by a network data analytics function (NWDAF) (3GPP, section 4.1, fig.4.2-1), the method comprising:
receiving network endpoint (NE) information about one or more NEs, wherein the NE information was transmitted by an application function (AF) (3GPP section 4.1, section 6.2.2.3, NWDAF collects data associated with network entities, including data from Afs, 5GC NFs, and OAM. There reference’s AF/UE application /service data is a reasonable for information about network endpoints, subject to the claim’s definition of “NE information.”);
using at least the received NE information (3GPP, section 4.1, section 6.4 and 6.9),
generating analytic data for splitting a machine learning (ML) inference process (Che, [0021-0025], [0025-0029], Che generates performance-derived partition point information for split DNN inference. It’s selected partition point is analytics/decision data specifying how the inference process is divided between edge and cloud. Also see Liu, [0043], fig.3 and 4, step-307); and
transmitting towards the AF the generated analytic data, wherein the analytic data for splitting the ML inference process indicates (3GPP, section 6.3, “analytics exposure to AF” related analytics exposure procedure.);
historical statistics and/or predictions regarding uplink-UL, (UL) data transmission from a user equipment (UE) to each of said one or more NEs (3GPP, table 6.4.2-3, “UE level Network data from OAM related to the QoS profile.);
historical statistics and/or predictions regarding packet delay on UL data transmission from the UE to each of said one or more NEs (3GPP, Table 6.4.2-3, identifies observed packet delay in the UL direction as UE-level network data available for analytics.);
historical statistics and/or predictions regarding packet loss rate on UL data transmission from the UE to each of said one or more NEs (3GPP, section 6.4.2, table 6.4.2-3 QoS sustainability analytics. The reference supplies packet transmission and packet retransmission information, from which a skilled artisan could derive packet-loss performance. It also provides past QoS change statistics and future QoS change likelihood.); and/or
a quality of service (QoS) indicator indicating a predicted quality of service in case the ML inference process is split (3GPP, section 6.9, teaches past QoS statistics and future likelihood of QoS change. Che [0024-0025] and [0029-0031] teaches selecting the inference-split point using predicted/profiled edge, interconnect, and cloud conditions to minimize latency. Combining them yields a predicted QoS/latency indicator for selected split.).
As to claim 23. The combined system of Liu, 3GPP and Che discloses the invention as in parent claim above including, wherein the NE information (Che, [0021-0025]) indicates:
an amount of computational resources available at said one or more NEs (Che, [0021-0024], assigns workload/performance levels to the edge device and cloud platform. The workload levels represent available computing capacity or resource condition at the end point.),
and/or end-to-end network performance between one or more pairs of NEs in case said one or more NEs includes more than one NE (Che, [0024-0025] and [0029-0031], the partition point is selected using the combined profile of the edge device, interconnect and cloud platform. The combined profile includes the performance of the network path between the endpoint pair.).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Please see the attached PTO-892.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TAUQIR HUSSAIN whose telephone number is (571)270-1247. The examiner can normally be reached M-F 7:00 - 8:00 with IFP.
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, Vivek Srivastava can be reached at 571 272-7304. 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.
/Tauqir Hussain/Primary Examiner, Art Unit 2446
Prosecution Insights