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 .
Information Disclosure Statement
The reference CN118102475A has not been considered in that the quality of the text of the document is not clear. The examiner was able to obtain a more legible copy of the document, and have it machine translated for review and added to the record. Should this not be the correct document, Applicant is requested to alleviate such under a new IDS.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim(s) 1, 5, 8, 12, 15 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over MORADI (Publication 2023/0351205) in view of PENG (Publication 2023/0222356) and NASR-AZADANI (Publication 2022/0012089).
As to claim 1, MORADI teaches a method, comprising: adaptively distributing artificial intelligence (AI) workloads across a federated inference infrastructure (via federated learning across a subset of mobile communications devices) based on real-time assessments of device capabilities, network conditions, and workload requirements; and distributing inference demands, to distribute tasks while supporting mobile device heterogeneity (see [0016, 0027, 0020-0022, 0023-0024, 0028-0032, 0034-0035).
However, MORADI does not teach augmenting rule-based decision by adaptively partitioning artificial intelligence (AI) workloads across a federated inference infrastructure
PENG teaches a federated learning environment wherein federated tasks are adaptively partitioned across the federated inference infrastructure and partitioned inference demands to distribute tasks while supporting mobile device heterogeneity (see Figures 1 and 4, note steps S405 in Figure 4 and [0064-0070, 0075-0081 and 0089]). It would be obvious to one of ordinary skill in the art before the effective filing of the claimed invention to apply the technique of PENG in the environment of MORADI in order to adequate deploy the federated learning task to the mobile device [0007].
However, MORADI and PENG does not teach the schedule is a rule-based such it is augmented adaptively based on real time assessments of device capabilities, network conditions and workload requirements.
NASR-AZADANI teaches a known AI distribution system that is rule based wherein augmenting of rule-based decisions (twin machine learning and AI system wherein its knowledge for distribution can be altered or modified based on assignments) by adaptively distributing artificial intelligence (AI) workloads (AI job / task) which based on real-time assessments of device capabilities, network conditions, and workload requirements (based on target metrics for AI task compared to different hardware platforms (e.g. compare the power usage, cost, latency and accuracy of machine learning models on FPGA, GPU, CPU, etc.); and distributing (via scheduling and determining assignments of AI job / task) inference demands, to distribute tasks while supporting device heterogeneity (having diverse hardware architectures) (abstract; [0016, 0031-0032, 0036-0038, 0041-0042, 0050-0051, 0059-0061]) . It would be obvious that one applies the well-known technique of NAS-AZADANI to the MORADI-PENG environment in order to self-improve or evolve over time via probabilistic programming in the distributing of AI operations [0018].
As to claim 5, NASR-AZADANI teaches employing fault-tolerance strategies for the federated inference infrastructure, including mechanisms for detecting and handling device failures, network disruptions and well as characteristics of foundation models (unpredicted uncertainties, for example unpredicted hardware failures and inaccurate resource predictions) ([0060, 0062]). Refer to claim 1 for the motivation to combine.
As to claims 8 and 12, reference is made to a system that corresponds to the method of claims 1 and 5 and is therefore rejected based on the rejection of claims 1 and 5 above. Further, claim 8 further details the system comprises memory and a processor coupled to the memory, wherein the processor performs the operations.
As to claims 15 and 19, reference is made to a computer program product that corresponds to the method of claims 1 and 5 and is therefore rejected based on the rejection of claims 1 and 5 above. Further claim 15 further details the computer program product further comprising a computer readable storage medium, wherein code stored in the computer readable storage medium when executed by a processor performs the operations.
Claims 2-4, 9-11 and 16-18 are rejected under 35 U.S.C. 103 as being unpatentable over MORADI (Publication 2023/0351205) in view of PENG (Publication 2023/0222356) and NASR-AZADANI (Publication 2022/0012089) as applied to claim 1 above, and further in view of CHEN (Publication 2024/0086699).
As to claim 2, while MORADI and NAS-AZADANI substantially teaches acquiring resource characteristics for federated training scheduling to devices to ensure efficient distribution of tasks and employing machine learning techniques to adaptively adjust workload partitioning based on historical data and real-time feedback (Moradi - [0016, 0027, 0020-0022, 0023-0024, 0028-0032, 0034-0035); NAS-AZADANI -abstract; [0016, 0031-0032, 0036-0038, 0041-0042, 0050-0051, 0059-0061]), the combination does not teach a SoC of such.
CHEN teaches analyzing computational capabilities of each mobile device, including available Central Processing Unit (CPU), Graphics Processing Unit (GPU), memory resources, and specialized AI inference System on Chips (SoCs); continuously monitoring network conditions, including bandwidth and latency ([0029-0033; 0065-0067, 0069-0075).
Therefore, it would be obvious to combine the teachings of CHEN to the teachings of MORADI-PENG and NAS-AZADANI in order to dynamically determining hardware capability for processing federated learning [0033].
As to claim 3, CHEN teaches dynamically balancing workload distribution across the federated inference infrastructure to prevent overloading and maximize processing efficiency by analyzing SoC characteristics ([0029-0033; 0065-0067, 0069-0075). It would be obvious to one of ordinary skill in the art before the effective filing of the claimed invention that by iteratively indicating the devices capabilities wherein the model for federated learning can be changed based on the updated hardware capabilities accomplishes the intended result of preventing overloading and maximizing processing efficiency. See also citations from Morandi, Peng and Nasr-Azadani used in parent claims. Refer to claim 2 for the motivation to combine.
As to claim 4, NASR-AZADANI teaches employing rule-based decision and Large AI models to predict future workload demands and recommend task migration, wherein tasks are dynamically reassigned from overloaded devices to underutilized ones, ensuring optimal resource utilization across a federated network [0038, claim 1]. It would be obvious to one of ordinary skill in the art before the effective filing of the claimed invention that by obtaining a new forecast regarding model assignments to resources of nodes, one obviously achieves the dynamically reassignment from one device to another device to achieve the goal of ensuring optimal resource utilization across a federated network. See also citations from Morandi, Peng and Nasr-Azadani wherein the model forecast is used for assignments / change of assignments as paragraphs are used in parent claims. It would be obvious that one applies the well-known technique of NAS-AZADANI to the MORADI-PENG-CHEN environment in order to self-improve or evolve over time via probabilistic programming in the distributing of AI operations [0018].
As to claims 9-11, reference is made to a system that corresponds to the method of claims 2-4 and is therefore rejected based on the rejection of claims 2-4 above.
As to claims 16-18, reference is made to a computer program product that corresponds to the method of claims 2-4 and is therefore rejected based on the rejection of claims 2-4 above.
Claims 6-7, 13-14 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over MORADI (Publication 2023/0351205) in view of PENG (Publication 2023/0222356) and NASR-AZADANI (Publication 2022/0012089) as applied to claim 1 above, and further in view of “CodedPaddedFL and CodedSecAgg: Straggler Mitigation and Secure Aggregation in Federated Learning” by SCHLEGEL et al.
As to claim 6, MORANDI, PENG and NASR-AZADANI substantially teaches the invention. However, the combination does not teach employing replication and redundancy to mitigate device failures.
SCHLEGEL teaches a federated learning technique that implements rule-based decision around task replication and employing redundancy to mitigate impact of device failures or network disruptions (pg. 2013-2014, Introduction, “FEDERATED learning (FL) [3], [4], [5] is a distributed learning paradigm that trains an algorithm across multiple devices without exchanging the training data directly, thus limiting the privacy leakage and reducing the communication load. More precisely, FL enables multiple devices to collaboratively learn a global model under the coordination of a central server. At each epoch, the devices train a local model on their local data and send the locally trained models to the central server. The central server aggregates the local models to update the global model, which is sent to the devices for the next epoch of the training. FL has been used in real-world applications, e.g., for medical data [6], text predictions on mobile devices [7], or by Apple to personalize Siri…. Training over many heterogeneous devices can be detrimental to the overall latency due to the effect of so-called stragglers, i.e., devices that take exceptionally long to finish their tasks due to random phenomena such as processes running in the background and memory access. Dropouts, which can be seen as an extreme case of straggling, may also occur…. In this paper, borrowing tools from coded distributed computing and edge computing, we propose two novel FL schemes, referred to as CodedPaddedFL and CodedSecAgg, that provide resiliency against straggling devices (and hence dropouts) by introducing redundancy on the devices’ local data. Both schemes can be divided into two phases. In the first phase, the devices share an encoded version of their data with other devices. In the second phase, the devices and the central server iteratively and collaboratively train a global model. The proposed schemes achieve significantly lower training latency than state-of-the-art schemes.”; see also pg. 2019-2020, VI Coded Secure Aggregation).
Therefore, it would be obvious to one of ordinary skill in the art before the effective filing of the claimed invention to apply the known techniques of SCHLEGEL in the system of MORANDI, PENG and NASR-AZADANI to mitigate the effect of strangling devices by introducing redundancy on the devices data across the network (abstract).
As to claim 7, NASR-AZADANI teaches utilizing machine learning to predict potential failures and proactively mitigating the potential failures to maintain uninterrupted AI inference (unpredicted uncertainties, for example unpredicted hardware failures and inaccurate resource predictions. Further EN: “to maintain uninterrupted AI inference is intended usage of proactively mitigating and thus the systems factoring this in in determining the assignments meets the standards of maintaining uninterrupted AI inference) ([0060, 0062]). It would be obvious that one applies the well-known technique of NAS-AZADANI to the MORADI-PENG-SCHLEGEL environment in order to self-improve or evolve over time via probabilistic programming in the distributing of AI operations [0018].
As to claims 13-14, reference is made to a system that corresponds to the method of claims 6-7 and is therefore rejected based on the rejection of claims 6-7 above.
As to claim 20, reference is made to a computer program product that corresponds to the method of claim 6 and is therefore rejected based on the rejection of claim 6 above.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LEWIS ALEXANDER BULLOCK JR whose telephone number is (571)272-3759. The examiner can normally be reached Monday-Friday, 9:00-5:00 pm.
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, Cordelia Zecher can be reached at 571-272-7771. 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.
/LEWIS A BULLOCK JR/Supervisory Patent Examiner, Art Unit 2199
Prosecution Insights