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 Arguments
Applicant's arguments filed 07/13/2026 have been fully considered but they are not persuasive.
Regarding applicant’s remarks directed to the rejection of claims under 35 USC § 101, the applicant argues that the amended claims directed to a technical solution. Examiner respectfully agrees and withdraws the rejection of claims under 35 USC § 101.
Regarding applicant’s remarks directed to the rejection of claims under 35 USC § 102, the arguments are directed to newly amended limitations that were not previously examined by the examiner. Therefore, applicants arguments are rendered moot. The examiner refers to the rejection under 35 USC § 103 in the current office action for more details.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claim 1-30 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention. Claim 1 and analogous claims 10, 19 and 28 recites “selecting the first federated learning model, by the UE, from multiple federated learning models associated with different zones indicated by the zone topology graph, based on the zone membership.” While para. [00100] recites that the participating device (UE) may receive multiple models from one or more FL zone managers, examiner respectfully points out that the specification of the instant application lacks written description on selecting the first federated learning model from “multiple federated learning models associated with different zones indicated by the zone topology graph.” (“[00100] In some examples, the FL phone manager 802 may receive multiple models from one or more FL zone managers 852. That is, multiple models (e.g., federated learning models or applications) may be provided to the participating device 800. As an example, a first application may be a text prediction model and a second application may be a location-based advertising model. In such examples, the FL phone manager 802 may determine a training time for each model. In some examples, the participating device 800 may be associated with two different zone managers, where each zone server is associated with a different zone. Each zone server may transmit a different model. As another example, a single zone server may transmit two or more different zone models.”)
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.
Claim(s) 1-30 are rejected under 35 U.S.C. 103 as being unpatentable over Jiang, Xiaopeng, et al. "Federated meta-location learning for fine-grained location prediction." 2021 IEEE International Conference on Big Data (Big Data). IEEE, 2021. (“Jiang”) in view of Ye, Minghao, et al. "Federated traffic engineering with supervised learning in multi-region networks." 2021 IEEE 29th International Conference on Network Protocols (ICNP). IEEE, 2021. (“Ye”)
In regards to claim 1,
Jiang teaches A processor-implemented method,
(Jiang, Section I., “We benchmarked the model on Android phones, and the results demonstrate that both training and inference are feasible in terms of execution time and battery consumption.”)
Jiang teaches comprising: receiving, by a user equipment (UE), a zone determination function and a zone topology graph
(Jiang, Section III., “The fundamental information to predict location is travel direction and speed. The user movement preferences and road characteristics also help the prediction. The GPS trajectories of each user contain this information. FMLL on the phones processes the raw location data to generate the meta-location, which represents trajectories as relative points in an abstract 2D space [receiving, by a user equipment (UE), a zone determination function ie the process in FMLL that determines the abstract 2D space from raw location data and a zone topology graph ie the abstract 2D space representation]. This section presents the process of meta-location generation and its benefits.”)
Jiang teaches based on registering for a federated learning process for training a first federated learning model;
Examiner interprets the zone determination function in light of the specification, “[0029] According to aspects of the present disclosure, when a device registers to participate in federated learning, the device is provided with the latest zone topology graph along with a "zone determination function." This function accepts a set of parameters from the device and returns the zone information to which the device belongs.”
(Jiang, Section I B., “(1) Initialization. Newly participating phones are required to register with the server [based on registering for a federated learning process for training a first federated learning model] to ensure that the server knows when model gradients uploaded at different times come from the same user. This could further allow the server to remove potential malicious users who may inject fake data into the model. 1”)
Jiang teaches determining, by the UE, a zone membership in accordance with UE parameters and the zone determination function by applying the zone determination function to the UE parameters at the UE;
(Jiang, Section III B., “To generate the input sequences, FMLL splits the user trajectories into fixed-length sub-trajectories. The length in time of the trajectories is determined experimentally. Each sub-trajectory is transformed into a sequence of relative points in an abstract 2D space. The X and Y coordinates of relative points at time t are determined based on their offsets from the location at previous time step t-1. The location of the very first point in a trajectory session is excluded. A location offset is denoted as ∆Lt =< latt− latt−1, lont lont−1 >. An input sequence at time t that looks back−k steps is denoted as St = (∆Lt−k+1, ∆Lt−k+2, ..., ∆Lt−1, ∆Lt). In its training, FMLL considers all possible k-length sequences, including overlapping sequences.
PNG
media_image1.png
292
682
media_image1.png
Greyscale
The historic region occupancy matrices are extracted from a historic occupancy matrix of the entire space (e.g., a city) [determining, by the UE, a zone membership ie a cell in the historic region occupancy matrix in accordance with UE parameters ie raw location data (see Section III A.) and the zone determination function by applying the zone determination function to the UE parameters at the UE ie the process in FMLL that determines the abstract 2D space from raw location data]. FMLL divides the entire space into a grid of fixed-size cells, and each cell corresponds to an element in the historic occupancy matrix. Each element represents the number of visits of the user in its corresponding cell. The matrix represents the occupancy of a bounded region Rt with area A, which is centered at the physical location Lt at time t. Rt is divided into M ×M fixed-size grid-cells, where A and M are predefined constants based on the maximum speed of users and the desired spatial granularity for the prediction. Each historic region occupancy matrix Ht is a M × M matrix, and it is extracted from the historic occupancy matrix for the entire space. Once extracted, this matrix is a meta-location input that does not maintain any relation with the physical locations that it represents. A matrix can implicitly tell if a road exists in a given cell (i.e., non-zero value for the corresponding matrix element) and can also tell if adjacent cells form routes taken frequently by the user.”)
Jiang teaches and training the first federated learning model by the UE;
(Jiang, Section III B., “The raw location data of each user is processed on their phone to produce meta-location as two types of inputs for the prediction model: fixed-length sequences of relative points and historic region occupancy matrices of the space considered for prediction. The input sequences contain the speed and direction information of the user trajectories. The occupancy matrices record frequently visited places and the most likely trajectories between these places. The inputs are computed offline (e.g., when the phones are charging) and can be updated over time based on new data to enable re-training [training the first federated learning model by the UE].”)
However, Jiang does not explicitly teach the zone topology graph identifying zones and corresponding federated learning zone managers; selecting the first federated learning model, by the UE, from multiple federated learning models associated with different zones indicated by the zone topology graph, based on the zone membership; and communicating wirelessly, by the UE, a model update for the first federated learning model to a federated learning zone manager corresponding to the zone membership.
Ye teaches the zone topology graph identifying zones and corresponding federated learning zone managers;
PNG
media_image2.png
328
301
media_image2.png
Greyscale
(Ye, Section IV A. and fig. 3(a), “For each region, an intra-region encoder is applied to model and encode the regional topology and traffic information. The inputs to the intra-region encoder are node features and a node connectivity matrix, where the features for a given network node are a series of demands originated from that node, and the connectivity matrix indicates the neighbors of each node. The regional controller is represented as a virtual node with node features initialized as all ones, and all region nodes are assumed to be directly connected to this virtual node. Figure 3(a) shows a multi-region topology example [the zone topology graph identifying zones and corresponding federated learning zone managers; wherein Examiner notes the regional controller to be analogous to the FMLL controller of Jiang]..” )
Ye teaches selecting the first federated learning model, by the UE, from multiple federated learning models associated with different zones indicated by the zone topology graph, based on the zone membership;
(Ye, Section III., “Moreover, once training is done, the identical FedTe model would be deployed in each regional controller of the real network environment [selecting the first federated learning model, by the UE, from multiple federated learning models associated with different zones indicated by the zone topology graph, based on the zone membership; wherein the multiple identical FL models are associated with the different zones]. Then, multiple FedTe models perform in a distributed and collaborative manner to quickly obtain efficient local routing decisions towards global optimal performance.”)
Ye teaches and communicating wirelessly, by the UE, a model update for the first federated learning model to a federated learning zone manager corresponding to the zone membership.
PNG
media_image3.png
295
347
media_image3.png
Greyscale
(Ye, Section III., “FedTe is trained through a centralized offline training procedure as shown in Fig. 1(b). At first, regional network states are periodically collected and aggregated to a server to compute the global optimal routing solutions [communicating wirelessly, by the UE, a model update ie regional network states for the first federated learning model to a federated learning zone manager corresponding to the zone membership; see fig. 1(b) for edge devices transmitting to the controllers (zone managers)], which are converted to the corresponding cross-region traffic distributions. During the training phase, these cross-region traffic distributions are used as the training targets to guide the learning of FedTe.”)
Jiang and Ye are both considered to be analogous to the claimed invention because they are in the same field of federated learning and location embedding. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jiang to incorporate the teachings of Ye in order to apply methods of federated learning on divided regions in a regional controller level (wherein recall the occupancy matrix of Jiang is also a contiguous space (ex. A city) divided into cells (regions) (Jiang, Section III B., “The historic region occupancy matrices are extracted from a historic occupancy matrix of the entire space (e.g., a city). FMLL divides the entire space into a grid of fixed-size cells, and each cell corresponds to an element in the historic occupancy matrix.”)) as doing so achieves a scalable and efficient network management along with reduced computation complexity (Ye, Section I., “To achieve scalable and efficient network management, a large network is usually divided into multiple geographical regions, and each of the regions is controlled by a regional controller [19]–[21]. In this way, distributed TE can be applied to reduce computation complexity and routing update overhead, and to handle link failures promptly and efficiently.”)
In regards to claim 2,
Jiang and Ye teaches The method of claim 1,
Jiang teaches further comprising: tabulating training data based on the zone membership;
(Jiang, Section III B., “The raw location data of each user is processed on their phone to produce meta-location as two types of inputs for the prediction model: fixed-length sequences of relative points and historic region occupancy matrices of the space considered for prediction. The input sequences contain the speed and direction information of the user trajectories. The occupancy matrices record frequently visited places and the most likely trajectories between these places. The inputs are computed offline (e.g., when the phones are charging) and can be updated over time based on new data to enable re-training [tabulating training data ie converting the raw location data to a table ie historic region occupancy matrix based on the zone membership wherein the historic region occupancy matrix is based on zone membership (the region of Jiang is analogous to the zone of the instant application)].”)
PNG
media_image4.png
167
391
media_image4.png
Greyscale
Jiang teaches and communicating with a federated learning zone manager corresponding to the zone membership.
(Jiang, Section V., “The system architecture of our framework is shown in Figure 3. The framework software runs on a server and on the phones of the users, and it uses federated learning (FL) [2] for training across all users. The FMLL Controller on the phones mediates the communication between the server and the phones. The Meta-Location Generation module on the phones processes the physical location data and generates meta-location for training. The FMLL Training and Prediction module runs on the phones. This module performs local model training on the phones and then submits the model gradients to the server through the Controller [communicating with a federated learning zone manager ie the FMLL controller corresponding to the zone membership wherein the local models send gradients corresponding to the predicted ‘zone membership’]. The FMLL Aggregator module at the server aggregates the gradients of the local models into a global model, and then distributes this model to the phones. When the OS or apps need a prediction, the Training and Prediction module is invoked. The output of the prediction is a meta-location, which is then converted into a physical location, with help from Meta-Location Generation module.”)
PNG
media_image5.png
288
348
media_image5.png
Greyscale
In regards to claim 3,
Jiang and Ye teaches The method of claim 1,
Jiang teaches further comprising receiving an updated zone determination function from a zone partition keeper.
(Jiang, Section III B., “The raw location data of each user is processed on their phone to produce meta-location as two types of inputs for the prediction model: fixed-length sequences of relative points and historic region occupancy matrices of the space considered for prediction. The input sequences contain the speed and direction information of the user trajectories. The occupancy matrices record frequently visited places and the most likely trajectories between these places. The inputs are computed offline (e.g., when the phones are charging) and can be updated over time based on new data to enable re-training [receiving an updated zone determination function ie updated process in FMLL that determines the abstract 2D space from raw location data wherein the updated process is the same process on new data from a zone partition keeper wherein the zone partition keeper is interpreted to be each user].”)
In regards to claim 4,
Jiang and Ye teaches The method of claim 1,
Jiang teaches further comprising receiving an updated zone topology graph from a zone partition keeper,
(Jiang, Section III B., “The raw location data of each user is processed on their phone to produce meta-location as two types of inputs for the prediction model: fixed-length sequences of relative points and historic region occupancy matrices of the space considered for prediction. The input sequences contain the speed and direction information of the user trajectories. The occupancy matrices record frequently visited places and the most likely trajectories between these places. The inputs are computed offline (e.g., when the phones are charging) and can be updated over time based on new data to enable re-training [receiving an updated zone topology graph ie a recalculated zone topology graph as taught in claim 1 by Jiang in view of Ye from a zone partition keeper wherein the zone partition keeper is interpreted to be each user].”)
Ye teaches the updated zone topology graph indicating updated federated learning zone managers for each zone.
PNG
media_image2.png
328
301
media_image2.png
Greyscale
(Ye, Section IV A. and fig. 3(a), “For each region, an intra-region encoder is applied to model and encode the regional topology and traffic information. The inputs to the intra-region encoder are node features and a node connectivity matrix, where the features for a given network node are a series of demands originated from that node, and the connectivity matrix indicates the neighbors of each node. The regional controller is represented as a virtual node with node features initialized as all ones, and all region nodes are assumed to be directly connected to this virtual node. Figure 3(a) shows a multi-region topology example [the zone topology graph identifying zones and corresponding federated learning zone managers; wherein Examiner notes the regional controller to be analogous to the FMLL controller of Jiang]..” )
In regards to claim 5,
Jiang and Ye teaches The method of claim 1,
Jiang teaches further comprising periodically determining the zone membership.
(Jiang, Section IV, “The problem is defined based on the meta-location input and output, defined in Section III. Let St ∈ R2k be the size-k sequence of relative points at time t for a given user. Let Ht ∈ Z+M×M be the historic regional occupancy matrix of the same user, which is a square matrix of order M centered at the user location at time t. Our goal is to predict the relative location of this user Ŷt+i∈ Z2 M ×M for the future ith timestamp [periodically ie ith timestamp determining the zone membership].”)
In regards to claim 6,
Jiang and Ye teaches The method of claim 1,
Jiang teaches further comprising: storing sensor data associated with a parameter;
(Jiang, Section III A., “The raw location data is recorded by each phone using the embedded GPS sensor. Let Lt =< latt, lont > denote the latitude and longitude of a user at time t [storing sensor data ie raw location data recorded by the embedded GPS sensor associated with a parameter ex latitude and longitude].”)
Jiang teaches and determining data samples for training with respect to a specific zone, based on the parameter associated with the sensor data.
(Jiang, Section III B., “The raw location data of each user is processed on their phone to produce meta-location as two types of inputs for the prediction model: fixed-length sequences of relative points and historic region occupancy matrices of the space considered for prediction. The input sequences contain the speed and direction information of the user trajectories. The occupancy matrices record frequently visited places and the most likely trajectories between these places. The inputs are computed offline (e.g., when the phones are charging) and can be updated over time based on new data to enable re-training [determining data samples for training with respect to a specific zone, based on the parameter associated with the sensor data].”)
In regards to claim 7,
Jiang and Ye teaches The method of claim 1,
Jiang teaches further comprising determining the zone membership in response to a triggering event.
(Jiang, Section V, “When the OS or apps need a prediction [determining the zone membership in response to a triggering event; wherein the triggering event is a determined need for a prediction], the Training and Prediction module is invoked.”)
In regards to claim 8,
Jiang and Ye teaches The method of claim 1,
Jiang teaches further comprising selecting a second federated learning model for inference based on the zone membership.
(Jiang, Section III C., “C. Meta-Location Output for Prediction Model
The location to be predicted Lt+i is mapped into the region R [selecting a second federated learning model for inference ie a prediction model based on the zone membership ie an updated historic regional occupancy matrix provided new data; wherein Examiner interprets this specific schema for a prediction model to be selecting wherein it is further based on a ‘zone membership’]. FMLL builds a prediction matrix Yt+i, as shown below:
PNG
media_image6.png
50
603
media_image6.png
Greyscale
where yi,j,t+i is an element of Yt+i, and Ri,j (1 ≤ i, j ≤ M) is a cell in region R. The meta-location output is formulated as a categorical class rather than a numerical value, so that we can set the spatial granularity of the prediction as a constant. Another reason for using categories is that the historic region occupancy matrix does not contain information to predict with spatial granularity beyond the grid-cell size. Overall, the output is a relative grid-cell, which is translated into a physical grid-cell on the user’s phone.”)
In regards to claim 9,
Jiang and Ye teaches The method of claim 1,
Jiang teaches further comprising: storing, by the UE, federated learning data
(Jiang, Section V, “The Meta-Location Generation module on the phones processes the physical location data and generates meta-location for training. The FMLL Training and Prediction module runs on the phones [storing, by the UE, federated learning data ie gradients obtained from training]. This module performs local model training on the phones and then submits the model gradients to the server through the Controller.”)
Jiang teaches while switching from a first federated learning zone to a second federated learning zone, the switching occurring without network service;
(Jiang, Section III B., “The raw location data of each user is processed on their phone to produce meta-location as two types of inputs for the prediction model: fixed-length sequences of relative points and historic region occupancy matrices of the space considered for prediction. The input sequences contain the speed and direction information of the user trajectories. The occupancy matrices record frequently visited places and the most likely trajectories between these places. The inputs are computed offline (e.g., when the phones are charging) and can be updated over time based on new data [while switching from a first federated learning zone to a second federated learning zone wherein it is recorded that the user moves from one region to a new region (recorded in the historic region occupancy matrix), the switching occurring without network service wherein the inputs are computed offline] to enable re-training.
Jiang teaches and uploading the federated learning data to the federated learning zone manager after resuming network service.
(Jiang, Section V, “This module performs local model training on the phones and then submits the model gradients to the server through the Controller [uploading the federated learning data to the federated learning zone manager after resuming network service].”)
Base claims 10, 19 and 28 and corresponding dependents (11-18), (20-27) and (29-20) are rejected on the same rationale under 35 U.S.C. 102(a)(1) as analogous claim 1 and dependents (2-10) as they are substantially similar, respectively, Mutatis mutandis.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
NPL: Gao, Yujia, et al. "Federated region-learning for environment sensing in edge computing system." IEEE Transactions on Network Science and Engineering 7.4 (2020): 2192-2204. (“Gao”) discloses federated learning with customizes local models for each micro cloud wherein each micro cloud embodies different regional characteristics.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JASMINE THAI whose telephone number is (703)756-5904. The examiner can normally be reached M-F 8-4.
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, Michael Huntley can be reached at (303) 297-4307. 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.
/J.T.T./Examiner, Art Unit 2129
/MICHAEL J HUNTLEY/Supervisory Patent Examiner, Art Unit 2129
Prosecution Insights