DETAILED ACTION
1. This communication is in response to the request for continued examination filed on March 27, 2025 and corresponding amendments filed on December 12, 2024 for Application No. 17/153,960 in which claims 1 and 4-17 are presented for examination.
Notice of Pre-AIA or AIA Status
2. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Continued Examination Under 37 CFR 1.114
3. A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 03/27/2025 has been entered.
Response to Arguments
4. The amendments filed on December 12, 2024 have been considered. Claims 1 and 9 have been amended. Claims 2 and 3 have been cancelled. Thus, Claims 1 and 4-17 are pending and presented for examination.
5. Applicant’s arguments filed December 12, 2024 with respect to the 35 U.S.C. 103 rejection 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
6. 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.
7. Claims 1 and 4-15 are rejected under 35 U.S.C. 103 as being unpatentable over Akdeniz et al. (hereinafter Akdeniz) (US PG-PUB 20230068386), in view of Sattler et al. (hereinafter Sattler) (“Robust and Communication-Efficient Federated Learning from Non-IID Data”).
Regarding Claim 1, Akdeniz teaches a method implemented using one or more processors (Akdeniz, Abstract, “The apparatus of an edge computing node, a system, a method and a machine-readable medium. The apparatus includes a processor to perform rounds of federated machine learning training including: […]”, thus, a method implemented using one or more processors is disclosed), the method comprising:
obtaining data from one or more data sources that are available in a given domain among a plurality of domains, wherein the data is in a domain-specific form that is specific to the given domain (Akdeniz, Par. [0039], "The edge cloud 110 is located much closer to the endpoint (consumer and producer) data sources 160 (e.g., autonomous vehicles 161, user equipment 162, business and industrial equipment 163, video capture devices 164, drones 165, smart cities and building devices 166, sensors and loT devices 167, etc.) than the cloud data center 130.", thus, data is obtained from one or more data sources that are available in a given domain (i.e., autonomous vehicles, user equipment, business/industrial equipment, video devices, drones, smart city devices, sensors, IoT devices, etc.). This is further supported by Figure 1. Similarly, Figure 2 label 205 also depicts the computational use cases for each set of domain-specific data & Par. [0200] also describes the different use cases (text classification, image recognition, etc.));
processing the data using one or more trained machine learning models (Akdeniz, Abstract, “The apparatus includes a processor to perform rounds of federated machine learning training including: processing client reports from a plurality of clients of the edge computing network; selecting a candidate set of clients from the plurality of clients for an epoch of the federated machine learning training; causing a global model to be sent to the candidate set of clients; and performing the federated machine learning training on the candidate set of clients”, thus, data is processed in a federated learning system using one or more trained machine learning models), wherein the one or more trained machine learning models include:
a domain-specific set of weights that is tailored to the given domain (Akdeniz, Par. [0412], “Therefore, some embodiments propose to utilize stochastic loss based sampling and correction using the bias factor at the client after each local update. In other words, when each client performs τ local updates before reporting the weight updates to the server, each client updates its local weight matrix as set forth in Equation (BS4): […]”, thus, each client (associated with different domains as supported by Par. [0039]) maintains their own local weights according to local client data, “tailored” to said domain), and
at least first and second global sets of weights that are shared across a plurality of domains of a federated learning system (Akdeniz, Par. [0167], “At operation 1433, the central server 1493 aggregates the received weights to obtain a new global weight. At operation 1436, the local weights and aggregated weights are shared between the client computing nodes 1491 and the central server 1493. Operations 1427-1436 are repeated until the model converges.”, therefore, global weights are computed/updated by aggregating local client weights. This process is iterative and performed across all layers – hence, the global weights aggregated by the server include at least first and second set of global weights shared across the federated learning system), wherein the domain-specific set of weights transforms a weight output by the first global set of weights to a transformed weight input to the second global set of weights (Akdeniz, Par. [0412-0413], “The local weight matrix of each client is therefore adjusted or biased based on the updated gradient estimate for each client k. After the server receives K updates from the K selected clients out of the total N clients, the server computes the global weight as the average of these weights as set forth in Equation (BS5):”, therefore, the domain-specific set of weights/local weights transform a weight output by the first global set of weights (as disseminated by the server to all client nodes) by adjusting bias (See Applicant’s specification Par. [0033] which mentions that a function for biasing a feature map includes such a “transformation” of weight output) to produce a biased/transformed weight which is then sent back to the server to be input as a second global set of weights during the iterative training process); and
providing, at one or more output components, an outcome of the processing (Akdeniz, Par. [0053], “In FIG. 3 , various client endpoints 310 (in the form of mobile devices, computers, autonomous vehicles, business computing equipment, industrial processing equipment) exchange requests and responses that are specific to the type of endpoint network aggregation.”, therefore, an outcome of processing/response is provided for each client endpoint, using components specific to each type of endpoint network),
wherein the global weights are trained using a plurality of gradients computed at the plurality of domains of the federated learning system, and the domain-specific weights are learned using local gradients computed within the given domain (Akdeniz, Par. [0466], “This algorithm is especially suitable for federated learning, as different clients can compute their local gradients, and the MEC server may then aggregate all local gradients to produce the global updated model, since the gradient for the loss function for all clients through their datasets may be given by Equation (DQ2) below: […]”, therefore, global weights are trained using aggregated local gradients across edge nodes whereas domain-specific weights are learned using local gradients computed for a given client), and wherein the domain- specific set of weights is isolated from the first and second global sets of weights during training (See introduction of Sattler reference below for explicit recitation of wherein the domain-specific set of weights is isolated from the first and second global set of weights during training).
While Akdeniz generally discloses that local/domain-specific weights are shared between client computing nodes and only sent to the global model for aggregation (See Akdeniz Par. [0167]), Akdeniz does not explicitly disclose wherein the domain-specific set of weights is isolated from the first and second global sets of weights during training
However, Sattler teaches wherein the domain- specific set of weights is isolated from the first and second global sets of weights during training (Sattler, Pg. 2, “Fig. 1: Federated Learning with a parameter server. Illustrated is one communication round of distributed SGD: a) Clients synchronize with the server. b) Clients compute a weight update independently based on their local data. c) Clients upload their local weight updates to the server, where they are averaged to produce the new master model.”, therefore, domain-specific client weights are independently updated based on local data and thus are isolated from the first/second global sets of weights during training – they are only shared once uploaded to the server for aggregation)
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the methods for federated learning, as disclosed by Akdeniz to include wherein the domain-specific set of weights is isolated from the first and second global sets of weights during training, as disclosed by Sattler. One of ordinary skill in the art would have been motivated to make this modification to enable robust privacy-preserving collaborative learning, such that local weights and data are isolated from global weights during training (Sattler, Pg. 1, “This leaves us facing the following dilemma: How are we going to make use of the rich combined data of millions of IoT devices for training deep learning models if this data can not be stored at a centralized location? Federated Learning resolves this issue as it allows multiple parties to jointly train a deep learning model on their combined data, without any of the participants having to reveal their data to a centralized server [10]. This form of privacy-preserving collaborative learning is achieved by following a simple three step protocol illustrated in Fig. 1”).
Regarding Claim 4, Akdeniz in view of Sattler teaches the method of claim 1, wherein the domain-specific weights correspond to an affine transform of the weight output by the first global set of weights (Akdeniz, Par. [0412], “Although, this approach adjusts bias for the weight after 1 round, when the number of local computations or local rounds τ is greater than 1, the gradient estimate at each client after each local update is biased. Therefore, some embodiments propose to utilize stochastic loss based sampling and correction using the bias factor at the client after each local update. In other words, when each client performs τ local updates before reporting the weight updates to the server, each client updates its local weight matrix as set forth in Equation (BS4): w k,t+τ =w k,t+τ−1 −η*g k,t+τ *p k /q k (BS4) where η is the learning rate. The local weight matrix of each client is therefore adjusted or biased based on the updated gradient estimate for each client k”, therefore, the domain-specific weights corresponds to an affine transform of the weight output by adjusting according to a bias term)
Regarding Claim 5, Akdeniz in view of Sattler teaches the method of claim 1, wherein one or more of the trained machine learning models comprises a convolutional neural network (Sattler, Pg. 8, “To cover a broad spectrum of learning problems we evaluate on differently sized convolutional and recurrent neural networks for the relevant Federated Learning tasks of image classification and speech recognition: […]”, thus, a convolutional neural network is disclosed).
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of claim 1, as disclosed by Akdeniz in view of Sattler to include wherein one or more of the trained machine learning models comprises a convolutional neural network, as disclosed by Sattler. One of ordinary skill in the art would have been motivated to make this modification to enable the use of convolutional neural networks which may be utilized for a wide variety of complex tasks in different domains, such as image classification and speech recognition (Sattler, Pg. 8, “We evaluate our proposed communication protocol on four different learning tasks and compare it’s performance to FederatedAveraging and signSGD in a wide a variety of different Federated Learning environments. Models and Datasets: To cover a broad spectrum of learning problems we evaluate on differently sized convolutional and recurrent neural networks for the relevant Federated Learning tasks of image classification and speech recognition:”).
Regarding Claim 6, Akdeniz in view of Sattler teaches the method of claim 1, wherein the domain-specific set of weights and the first and second global sets of weights are incorporated into a single trained machine learning model of the one or more trained machine learning models during the processing (Akdeniz, Par. [0122], "Federated learning, where a global model is trained with coordination with a federation of client computing nodes/client nodes/clients while keeping the training data local at the clients is one of the problems under consideration herein. The federated learning protocol iteratively allows clients to download a centrally trained artificial intelligence/machine-learning model (or model) from a server, such as a MEC server, an edge server or a cloud server, update it with their own data and upload the model updates (such as a gradient update) back to the server. The model updates may include updates weight values for nodes of the NN model, for instance. The server then aggregates updates from multiple clients to update the global model", thus, the domain-specific weights and global weights may be incorporated into a federated learning system which maintains a global model).
Regarding Claim 7, Akdeniz in view of Sattler teaches the method of claim 1, wherein the domain-specific set of weights and the first and second global sets of weights are learned during combined training of one or more of the trained machine learning models (Akdeniz, Par. [0122], "Federated learning, where a global model is trained with coordination with a federation of client computing nodes/client nodes/clients while keeping the training data local at the clients is one of the problems under consideration herein.", thus, the weights are learned during combined training of the federated learning models).
Regarding Claim 8, Akdeniz in view of Sattler teaches the method of claim 1, wherein two or more of the obtaining, processing, and providing are performed by a computing device associated with the given domain (Akdeniz, Par. [0082], "In a more detailed example, FIG. 8 illustrates a block diagram of an example of components that may be present in an edge computing node 850 for implementing the techniques (e.g., operations, processes, methods, and methodologies) described herein.", thus, the aforementioned limitations are performed by a computing device, edge computing node, associated with the given client and according domain).
Regarding Claim 9, Akdeniz teaches a method for federated learning using one or more processors of a federated learning system (Akdeniz, Abstract, “The apparatus of an edge computing node, a system, a method and a machine-readable medium. The apparatus includes a processor to perform rounds of federated machine learning training including: […]”, thus, a method implemented using one or more processors of a federated learning system is disclosed), the method comprising:
obtaining data from one or more data sources that are available in a given domain among a plurality of domains of the federated learning system, wherein the data is in a domain-specific form that is specific to the given domain (Akdeniz, Par. [0039], "The edge cloud 110 is located much closer to the endpoint (consumer and producer) data sources 160 (e.g., autonomous vehicles 161, user equipment 162, business and industrial equipment 163, video capture devices 164, drones 165, smart cities and building devices 166, sensors and loT devices 167, etc.) than the cloud data center 130.", thus, data is obtained from one or more data sources that are available in a given domain (i.e., autonomous vehicles, user equipment, business/industrial equipment, video devices, drones, smart city devices, sensors, IoT devices, etc.). This is further supported by Figure 1. Similarly, Figure 2 label 205 also depicts the computational use cases for each set of domain-specific data & Par. [0200] also describes the different use cases (text classification, image recognition, etc.));
providing one or more machine learning models (Akdeniz, Abstract, “The apparatus includes a processor to perform rounds of federated machine learning training including: processing client reports from a plurality of clients of the edge computing network; selecting a candidate set of clients from the plurality of clients for an epoch of the federated machine learning training; causing a global model to be sent to the candidate set of clients; and performing the federated machine learning training on the candidate set of clients”, thus, data is processed in a federated learning system using one or more machine learning models), wherein the one or more machine learning models include:
at least first and second global sets of weights that are shared across a plurality of domains of the federated learning system (Akdeniz, Par. [0167], “At operation 1433, the central server 1493 aggregates the received weights to obtain a new global weight. At operation 1436, the local weights and aggregated weights are shared between the client computing nodes 1491 and the central server 1493. Operations 1427-1436 are repeated until the model converges.”, therefore, global weights are computed/updated by aggregating local client weights. This process is iterative and performed across all layers – hence, the global weights aggregated by the server include at least first and second set of global weights shared across the federated learning system), and
a domain-specific set of weights (Akdeniz, Par. [0412], “Therefore, some embodiments propose to utilize stochastic loss based sampling and correction using the bias factor at the client after each local update. In other words, when each client performs τ local updates before reporting the weight updates to the server, each client updates its local weight matrix as set forth in Equation (BS4): […]”, thus, each client (associated with different domains as supported by Par. [0039]) maintains their own local weights according to local client data, “tailored” to said domain), wherein the domain-specific set of weights transforms a weight output by the first global set of weights to a transformed weight input to the second global set of weights(Akdeniz, Par. [0412-0413], “The local weight matrix of each client is therefore adjusted or biased based on the updated gradient estimate for each client k. After the server receives K updates from the K selected clients out of the total N clients, the server computes the global weight as the average of these weights as set forth in Equation (BS5):”, therefore, the domain-specific set of weights/local weights transform a weight output by the first global set of weights (as disseminated by the server to all client nodes) by adjusting bias (See Applicant’s specification Par. [0033] which mentions that a function for biasing a feature map includes such a “transformation” of weight output) to produce a biased/transformed weight which is then sent back to the server to be input as a second global set of weights during the iterative training process);
isolating the domain-specific set of weights from the first and second global sets of weights (See introduction of Sattler reference below for explicit recitation of wherein the domain-specific set of weights is isolated from the first and second global set of weights);
processing the data to train the one or more machine learning models while the domain-specific set of weights is isolated from the first and second global sets of weights (Akdeniz, Figure 14, which depicts the processing of data between the client compute node(s) and the central server to train the one or more machine learning models. For recitation of isolating the domain-specific set of weights from the first/second global sets of weights, see introduction of Sattler reference below); and
based on one or more outcomes of the processing, training the one or more machine learning models (Akdeniz, Par. [0167], “At operation 1427, the client computing nodes 1491 compute model updates by training over the embedded training data {circumflex over (X)}(i) and raw labels Y(i). At operation 1430, the client computing nodes 1491 return the updated weights to the central server 1493. At operation 1433, the central server 1493 aggregates the received weights to obtain a new global weight.”, thus, the one or more machine learning models are trained),
wherein the global weights are trained using a plurality of gradients computed at the plurality of domains of the federated learning system, and the domain-specific weights are learned using local gradients computed within the given domain (Akdeniz, Par. [0466], “This algorithm is especially suitable for federated learning, as different clients can compute their local gradients, and the MEC server may then aggregate all local gradients to produce the global updated model, since the gradient for the loss function for all clients through their datasets may be given by Equation (DQ2) below: […]”, therefore, global weights are trained using aggregated local gradients across edge nodes whereas domain-specific weights are learned using local gradients computed for a given client), and wherein the domain-specific set of weights is isolated from the first and second global sets of weights during training (See introduction of Sattler reference below for explicit recitation of wherein the domain-specific set of weights is isolated from the first and second global set of weights).
Akdeniz does not explicitly disclose:
isolating the domain-specific set of weights from the first and second global sets of weights
wherein the domain-specific set of weights is isolated from the first and second global sets of weights during training
However, Sattler teaches:
isolating the domain-specific set of weights from the first and second global sets of weights (Sattler, Pg. 2, “Fig. 1: Federated Learning with a parameter server. Illustrated is one communication round of distributed SGD: a) Clients synchronize with the server. b) Clients compute a weight update independently based on their local data. c) Clients upload their local weight updates to the server, where they are averaged to produce the new master model.”, therefore, domain-specific client weights are independently updated based on local data and thus are isolated from the first/second global sets of weights during training – they are only shared once uploaded to the server for aggregation)
wherein the domain-specific set of weights is isolated from the first and second global sets of weights during training (Sattler, Pg. 2, “Fig. 1: Federated Learning with a parameter server. Illustrated is one communication round of distributed SGD: a) Clients synchronize with the server. b) Clients compute a weight update independently based on their local data. c) Clients upload their local weight updates to the server, where they are averaged to produce the new master model.”, therefore, domain-specific client weights are independently updated based on local data and thus are isolated from the first/second global sets of weights during training – they are only shared once uploaded to the server for aggregation)
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the methods for federated learning, as disclosed by Akdeniz to include wherein the domain-specific set of weights is isolated from the first and second global sets of weights during training, as disclosed by Sattler. One of ordinary skill in the art would have been motivated to make this modification to enable robust privacy-preserving collaborative learning, such that local weights and data are isolated from global weights during training (Sattler, Pg. 1, “This leaves us facing the following dilemma: How are we going to make use of the rich combined data of millions of IoT devices for training deep learning models if this data can not be stored at a centralized location? Federated Learning resolves this issue as it allows multiple parties to jointly train a deep learning model on their combined data, without any of the participants having to reveal their data to a centralized server [10]. This form of privacy-preserving collaborative learning is achieved by following a simple three step protocol illustrated in Fig. 1”).
Regarding Claim 10, Akdeniz in view of Sattler teaches the method of claim 9, wherein the training includes alternating between updating the first and second global sets of weights and updating the domain-specific set of weights (Akdeniz, Par. [0167], “At operation 1430, the client computing nodes 1491 return the updated weights to the central server 1493. At operation 1433, the central server 1493 aggregates the received weights to obtain a new global weight. At operation 1436, the local weights and aggregated weights are shared between the client computing nodes 1491 and the central server 1493. Operations 1427-1436 are repeated until the model converges.", thus, the global weight is only updated after receiving the already updated local weights – hence, the training includes alternating between updating the global weights and domain-specific weights).
Regarding Claim 11, Akdeniz in view of Sattler teaches the method of claim 10, wherein the first and second global sets of weights are held constant during training of the domain-specific set of weights, and the domain-specific set of weights are held constant during training of the first and second global sets of weights (Sattler, Pg. 2, “Fig. 1: Federated Learning with a parameter server. Illustrated is one communication round of distributed SGD: a) Clients synchronize with the server. b) Clients compute a weight update independently based on their local data. c) Clients upload their local weight updates to the server, where they are averaged to produce the new master model.”, therefore, domain-specific client weights are independently updated based on local data and thus are isolated from the first/second global sets of weights during training – they are only shared once uploaded to the server for aggregation. Hence, weights are held constant during training of the other corresponding weights).
The reasons of obviousness have been noted in the rejection of Claim 9 above and applicable herein.
Regarding Claim 12, Akdeniz in view of Sattler teaches the method of claim 10, wherein updating the first and second global sets of weights includes:
computing a local gradient for the first and second global sets of weights using the data obtained from the one or more data sources available in the given domain (Akdeniz, Par. [0610], "Each client may compute a local gradient from I.sub.i*(t*) number of raw data points. Therefore, in each epoch, each client randomly picks l.sub.i* (t*) raw data points out of 1.sub.i raw data points available at the client, such that each data point has equal likelihood of selection given by", therefore, a local gradient is computed for each client and according domain); and
transmitting data indicative of the local gradient to a federated learning central server (Akdeniz, Par. [0610], "The client computes its local gradient and uploads its local gradient to the server. The expected value of total local gradients received by the MEC server within the deadline time t* is given by Equation (DR18):", thus, each local gradient is uploaded to the federated learning central server), wherein the federated learning central server uses the local gradient and other local gradients computed in other domains participating in the federated learning to train the first and second global sets of weights (Akdeniz, Par. [0466], “This algorithm is especially suitable for federated learning, as different clients can compute their local gradients, and the MEC server may then aggregate all local gradients to produce the global updated model, since the gradient for the loss function for all clients through their datasets may be given by Equation (DQ2) below:", thus, the local gradients are used to train the global weights and produce a corresponding global model).
Regarding Claim 13, Akdeniz in view of Sattler teaches the method of claim 9, wherein one or more of the machine learning models comprises a convolutional neural network (Sattler, Pg. 8, “To cover a broad spectrum of learning problems we evaluate on differently sized convolutional and recurrent neural networks for the relevant Federated Learning tasks of image classification and speech recognition: […]”, thus, a convolutional neural network is disclosed).
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of claim 9, as disclosed by Akdeniz in view of Sattler to include wherein one or more of the trained machine learning models comprises a convolutional neural network, as disclosed by Sattler. One of ordinary skill in the art would have been motivated to make this modification to enable the use of convolutional neural networks which may be utilized for a wide variety of complex tasks in different domains, such as image classification and speech recognition (Sattler, Pg. 8, “We evaluate our proposed communication protocol on four different learning tasks and compare it’s performance to FederatedAveraging and signSGD in a wide a variety of different Federated Learning environments. Models and Datasets: To cover a broad spectrum of learning problems we evaluate on differently sized convolutional and recurrent neural networks for the relevant Federated Learning tasks of image classification and speech recognition:”).
Regarding Claim 14, Akdeniz in view of Sattler teaches the method of claim 9, wherein the domain-specific weights correspond to a differentiable function (Akdeniz, Par. [0185], “As is common in NN training, updates are made to the NN using gradient descent, where the partial gradient on each data point is obtained using a forward pass on the da ta followed by a backward pass. Additionally, we use the standard mean squared error (MSE) loss for training. Furthermore, a common strategy in NN training is stochastic gradient descent (SGD), where training is performed in batches of data and for each batch, the model is updated after a forward backward pass on the batch.", thus, the domain-specific weights may correspond with the use of stochastic gradient descent - further supported by Par. [0186]).
Regarding Claim 15, Akdeniz in view of Sattler teaches a system comprising one or more processors and memory storing instructions that, in response to execution by the one or more processors (Akdeniz, Par. [0084], “The processor 852 may communicate with a system memory 854 over an interconnect 856 (e.g., a bus) through an interconnect interface 853 of the processor.”, thus, a system (as depicted by Figure 8) comprising one or more processors and memory storing instructions is disclosed), cause the one or more processors to perform the method of claim 1 (See rejection of Claim 1 above).
8. Claims 16 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Akdeniz et al. (hereinafter Akdeniz) (US PG-PUB 20230068386), in view of Sattler et al. (hereinafter Sattler) (“Robust and Communication-Efficient Federated Learning from Non-IID Data”), further in view of Daykin et al. (hereinafter Daykin) (US PG-PUB 20210097381).
Regarding Claim 16, Akdeniz in view of Sattler teaches the method of claim 1.
Akdeniz in view of Sattler does not explicitly disclose wherein the plurality of domains are a plurality of different hospital systems.
However, Daykin teaches wherein the plurality of domains are a plurality of different hospital systems (Daykin, Par. [0050], “FIG. 3 shows an example of four institutions. In other embodiments, any number of institutions may be connected. The institutions may form a federation for performing a federated learning method in accordance with an embodiment. In the present embodiment, the institutions are hospitals. In other embodiments, at least some of the institutions may comprise, for example, companies or universities.”, therefore, the plurality of domains in the federated learning system may comprise a plurality of different hospital systems).
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of claim 1, as disclosed by Akdeniz in view of Sattler to include wherein the plurality of domains are a plurality of different hospital systems, as disclosed by Daykin. One of ordinary skill in the art would have been motivated to make this modification to enable the use of federated learning within the medical domain, including different hospital systems, enabling efficient and accurate model training without data transfer – hence protecting data and/or medical privacy (Daykin, Par. [0023-0024], “Federated learning may be of importance in tasks where large amounts of data cannot be easily transferred to a single location. This may include the medical domain. In the medical domain, datasets are typically sensitive. It may be challenging to move datasets out of the hospital or institution that generated them due to legal, political, technical, and/or financial reasons. Federated learning seeks to train a model without any data transfer.”).
Claim 17 recites substantially the same limitations as Claim 16, therefore it is rejected under the same rationale.
Conclusion
9. Any inquiry concerning this communication or earlier communications from the examiner should be directed to Devika S Maharaj whose telephone number is (571)272-0829. The examiner can normally be reached Monday - Thursday 8:30am - 5:30pm.
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, Alexey Shmatov can be reached on (571)270-3428. 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.S.M./Examiner, Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123