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 .
Specification
Applicant is reminded of the proper language and format for an abstract of the disclosure.
The abstract should be in narrative form and generally limited to a single paragraph on a separate sheet within the range of 50 to 150 words in length. The abstract should describe the disclosure sufficiently to assist readers in deciding whether there is a need for consulting the full patent text for details.
The language should be clear and concise and should not repeat information given in the title. It should avoid using phrases which can be implied, such as, “The disclosure concerns,” “The disclosure defined by this invention,” “The disclosure describes,” etc. In addition, the form and legal phraseology often used in patent claims, such as “means” and “said,” should be avoided.
The abstract of the disclosure is objected to because it is not narrative in form and is legal form as it a copy of claim 1. A corrected abstract of the disclosure is required and must be presented on a separate sheet, apart from any other text. See MPEP § 608.01(b).
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-10 and 16-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claims recite a mental process of observation, judgement and evaluation. This judicial exception is not integrated into a practical application because and does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional limitation are extra-solution activity is well-understood, routine and convention in combination with generic computer hardware used to execute the abstract idea. See the analysis below for further details.
Claim 1
Step 1: The claim recites an apparatus, therefore, it falls into a statutory category.
Step 2A Prong 1: The claim recites, inter alia:
Determining the information about the first neural network based on the indication information. (This is mental step of observation, consideration and judgement wherein a user looks at information and makes a determination from it.)
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
at least one processing circuit; and one or more memories coupled to the at least one processing circuit and storing programming instructions for execution by the at least one processing circuit to cause the apparatus to: (This is use generic computer hardware to execute the abstract idea above, see MPEP 2106.05(f).)
obtaining indication information from a first device, wherein the indication information indicates information about a first neural network model, (This is extra solution activity sending and receiving data, which is data collection, see MPEP 2106.05(g).) the first neural network model comprises a dedicated layer network and a common layer network, the dedicated layer network is dedicated to the first neural network model, and the common layer network is a common network of the first neural network model and a second neural network model; (This is cited a high level of generality and result in use a neural network as tool to implement the abstract idea, see MPEP 2106.05(f).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination of generic computer functions that are implemented to perform the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “obtaining indication information from a first device, wherein the indication information indicates information about a first neural network model,” amount to transmitting data and well-understood, routine and conventional and does not amount to significantly more. See MPEP 2106.06(d)(II) wherein it cites “The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity. i. Receiving or transmitting data over a network, e.g., using the Internet to gather data”. The limitations of “the first neural network model comprises a dedicated layer network and a common layer network, the dedicated layer network is dedicated to the first neural network model, and the common layer network is a common network of the first neural network model and a second neural network model; at least one processing circuit; and one or more memories coupled to the at least one processing circuit and storing programming instructions for execution by the at least one processing circuit to cause the apparatus to” amount to using generic computer hardware to execute the abstract idea, see MPEP 2106.05(f). When viewing the claim as a whole it does not amount to significantly more than the abstract idea.
The additional elements as disclosed above in combination of the abstract idea are not sufficient to amount to significantly more than the judicial exception as they are well, understood, routine and conventional activity as disclosed in combination of generic computer functions that are implemented to perform the disclosed abstract idea above.
Claim 2
Step 2A Prong 1: The claim recites, inter alia:
The claim inherits the abstract idea of claim 1.
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
wherein the information about the first neural network model comprises at least one of the following: neural network cascaded structure information, common layer network structure information, a common layer network parameter, dedicated layer network structure information, or a dedicated layer network parameter. (This amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination with the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “wherein the information about the first neural network model comprises at least one of the following: neural network cascaded structure information, common layer network structure information, a common layer network parameter, dedicated layer network structure information, or a dedicated layer network parameter.” amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).
Claim 3
Step 2A Prong 1: The claim recites, inter alia:
The claim inherits the abstract idea of claim 1.
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
wherein that the information about the first neural network model comprises the neural network cascaded structure information comprises: the information about the first neural network model indicates the neural network cascaded structure; or the information about the first neural network model indicates a configured neural network cascaded structure from a neural network cascaded structure candidate set. (This amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination with the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “wherein that the information about the first neural network model comprises the neural network cascaded structure information comprises: the information about the first neural network model indicates the neural network cascaded structure; or the information about the first neural network model indicates a configured neural network cascaded structure from a neural network cascaded structure candidate set.” amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).
Claim 4
Step 2A Prong 1: The claim recites, inter alia:
The claim inherits the abstract idea of claim 1.
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
wherein that the information about the first neural network model comprises the common layer network structure information comprises: the information about the first neural network model indicates the common layer network structure; or the information about the first neural network model indicates a configured common layer network structure from a common layer network structure candidate set. (This amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination with the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “wherein that the information about the first neural network model comprises the common layer network structure information comprises: the information about the first neural network model indicates the common layer network structure; or the information about the first neural network model indicates a configured common layer network structure from a common layer network structure candidate set.” amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).
Claim 5
Step 2A Prong 1: The claim recites, inter alia:
The claim inherits the abstract idea of claim 1.
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
wherein that the information about the first neural network model comprises common layer network parameter information comprises: the information about the first neural network model indicates the common layer network parameter; or the information about the first neural network model indicates a configured common layer network parameter from a common layer network parameter candidate set. (This amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination with the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “wherein that the information about the first neural network model comprises common layer network parameter information comprises: the information about the first neural network model indicates the common layer network parameter; or the information about the first neural network model indicates a configured common layer network parameter from a common layer network parameter candidate set.” amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).
Claim 6
Step 2A Prong 1: The claim recites, inter alia:
The claim inherits the abstract idea of claim 1.
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
wherein that the information about the first neural network model comprises the dedicated layer network structure information comprises: the information about the first neural network model indicates the dedicated layer network structure; or the information about the first neural network model indicates a configured dedicated layer network structure from a dedicated layer network structure candidate set. (This amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination with the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “wherein that the information about the first neural network model comprises the dedicated layer network structure information comprises: the information about the first neural network model indicates the dedicated layer network structure; or the information about the first neural network model indicates a configured dedicated layer network structure from a dedicated layer network structure candidate set.” amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).
Claim 7
Step 2A Prong 1: The claim recites, inter alia:
The claim inherits the abstract idea of claim 1.
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
wherein that the information about the first neural network model comprises dedicated layer network parameter information comprises: the information about the first neural network model indicates the dedicated layer network parameter; or the information about the first neural network model indicates a configured dedicated layer network parameter from a dedicated layer network parameter candidate set. (This amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination with the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “herein that the information about the first neural network model comprises dedicated layer network parameter information comprises: the information about the first neural network model indicates the dedicated layer network parameter; or the information about the first neural network model indicates a configured dedicated layer network parameter from a dedicated layer network parameter candidate set.” amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).
Claim 8
Step 2A Prong 1: The claim recites, inter alia:
The claim inherits the abstract idea of claim 1.
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
receiving configuration information from the first device, wherein the configuration information indicates one or more of the following content: a neural network cascaded structure information candidate set, a common layer network structure information candidate set, a common layer network parameter candidate set, a dedicated layer network structure information candidate set, or the dedicated layer network parameter candidate set. (This amount to data collection which extra-solution activity, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination with the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “receiving configuration information from the first device, wherein the configuration information indicates one or more of the following content: a neural network cascaded structure information candidate set, a common layer network structure information candidate set, a common layer network parameter candidate set, a dedicated layer network structure information candidate set, or the dedicated layer network parameter candidate set.” amount to transmitting data and well-understood, routine and conventional and does not amount to significantly more. See MPEP 2106.06(d)(II) wherein it cites “The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity. i. Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Claim 9 and 19
Step 2A Prong 1: The claim recites, inter alia:
The claim inherits the abstract idea of claim 1 and 16.
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
Sending (receiving – claim 19) auxiliary information to the first (second – claim 19) device, wherein the auxiliary information indicates a wireless system parameter for a communication between the first device and a second device. (This amount to data collection which extra-solution activity, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination with the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “sending auxiliary information to the first device, wherein the auxiliary information indicates a wireless system parameter for a communication between the first device and a second device.” amount to transmitting data and well-understood, routine and conventional and does not amount to significantly more. See MPEP 2106.06(d)(II) wherein it cites “The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity. i. Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Claim 10 and 20
Step 2A Prong 1: The claim recites, inter alia:
The claim inherits the abstract idea of claim 1 and 16.
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
sending capability information to the first device, wherein the capability information indicates one or more pieces of the following information: (1) whether to support using a neural network to replace or implement a function of a communication module; (2) whether to support a neural network structure in which the common layer network and the dedicated layer network are cascaded; (3) whether to support receiving of at least one of dedicated layer network structure information or dedicated layer network parameter through signaling; (4) stored dedicated layer network structure information predefined in a protocol or pre-configured; (5) a stored dedicated layer network parameter predefined in the protocol or pre-configured; (6) a memory space used to store neural network cascaded structure information, common layer network structure information, a common layer network parameter, the dedicated layer network structure information, and/or the dedicated layer network parameter; or (7) computing power information used to run the neural network. (This amount to data collection which extra-solution activity, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination with the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of sending capability information to the first device, wherein the capability information indicates one or more pieces of the following information: (1) whether to support using a neural network to replace or implement a function of a communication module; (2) whether to support a neural network structure in which the common layer network and the dedicated layer network are cascaded; (3) whether to support receiving of at least one of dedicated layer network structure information or dedicated layer network parameter through signaling; (4) stored dedicated layer network structure information predefined in a protocol or pre-configured; (5) a stored dedicated layer network parameter predefined in the protocol or pre-configured; (6) a memory space used to store neural network cascaded structure information, common layer network structure information, a common layer network parameter, the dedicated layer network structure information, and/or the dedicated layer network parameter; or (7) computing power information used to run the neural network.” amount to transmitting data and well-understood, routine and conventional and does not amount to significantly more. See MPEP 2106.06(d)(II) wherein it cites “The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity. i. Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Claim 16
Step 1: The claim recites an apparatus, therefore, it falls into a statutory category.
Step 2A Prong 1: The claim recites, inter alia:
determining indication information, wherein the indication information indicates information about a first neural network model, the first neural network model comprises a dedicated layer network and a common layer network, the dedicated layer network is dedicated to the first neural network model, and the common layer network is a common network of the first neural network model and a second neural network model; (This is mental step of observation, consideration and judgement wherein a user looks at looks at a neural network and determines information about it, where it is the number of layers, weights, type of neural network or anything else.)
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
at least one processing circuit; and one or more memories coupled to the at least one processing circuit and storing programming instructions for execution by the at least one processing circuit to cause the apparatus to: (This is using generic computer hardware to execute the abstract idea above, see MPEP 2106.05(f).)
sending the indication information to a second device. (This is extra solution activity sending and receiving data, which is data collection, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination of generic computer functions that are implemented to perform the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “sending the indication information to a second device.” amount to transmitting data and well-understood, routine and conventional and does not amount to significantly more. See MPEP 2106.06(d)(II) wherein it cites “The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity. i. Receiving or transmitting data over a network, e.g., using the Internet to gather data”. The additional limitations of “at least one processing circuit; and one or more memories coupled to the at least one processing circuit and storing programming instructions for execution by the at least one processing circuit to cause the apparatus to:” amount to using generic computer hardware to execute the abstract idea, see MPEP 2106.05(f).
The additional elements as disclosed above in combination of the abstract idea are not sufficient to amount to significantly more than the judicial exception as they are well, understood, routine and conventional activity as disclosed in combination of generic computer functions that are implemented to perform the disclosed abstract idea above.
Claim 17
Step 2A Prong 1: The claim recites, inter alia:
The claim inherits the abstract idea of claim 16.
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
wherein the information about the first neural network model comprises at least one of the following: neural network cascaded structure information, common layer network structure information, a common layer network parameter, dedicated layer network structure information, or a dedicated layer network parameter. (This amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination with the disclosed abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “wherein the information about the first neural network model comprises at least one of the following: neural network cascaded structure information, common layer network structure information, a common layer network parameter, dedicated layer network structure information, or a dedicated layer network parameter.” amount a particular type of the data to be used or manipulated as such is extra-solution activity, see MPEP 2106.05(g).
Claim 18
Step 2A Prong 1: The claim recites, inter alia:
The claim inherits the abstract idea of claim 16.
Step 2A Prong 2:
This judicial exception is no integrated into a practical application. Aside from the limitations above, the claim recites:
sending configuration information to the second device, wherein the configuration information indicates one or more of the following content: a neural network cascaded structure information candidate set, a common layer network structure information candidate set, a common layer network parameter candidate set, a dedicated layer network structure information candidate set, or a dedicated layer network parameter candidate set. (This is extra solution activity sending and receiving data, which is data collection, see MPEP 2106.05(g).)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination with the abstract idea above.
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional elements of “sending configuration information to the second device, wherein the configuration information indicates one or more of the following content: a neural network cascaded structure information candidate set, a common layer network structure information candidate set, a common layer network parameter candidate set, a dedicated layer network structure information candidate set, or a dedicated layer network parameter candidate set.” amount to transmitting data and is well-understood, routine and conventional and does not amount to significantly more. See MPEP 2106.06(d)(II) wherein it cites “The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity. i. Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
The additional elements as disclosed above in combination of the abstract idea are not sufficient to amount to significantly more than the judicial exception as they are well, understood, routine and conventional activity as disclosed in combination with the abstract idea above.
Claim Rejections - 35 USC § 102
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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim 1-8 and 16-18 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Arivazhagan et al. (“Federated Learning with Personalization Layers” – hereinafter Arivazhagan).
In regards to claim 1, Arivazhagan discloses a communication apparatus, comprising:
at least one processing circuit and one or more memories coupled to the at least one processing circuit and storing programming instructions for execution by the at least one processing circuit to cause the apparatus to perform the following: (Arivazhagan page 1 section 1 teaches mobile devices having hardware and software that execute machine learning and communicate with a parameter sever. As it is mobile devices, it thus has at least one processing circuit and memory.)
obtaining indication information from a first device, (Arivazhagan’s federated learning (FL) setup uses a parameter server (cloud) that coordinates training with client devices as cite on page 1 section 1 wherein it cites “…execute machine learning (ML) model training computations on device with locally available data and occasional communication with an aggregating parameter server. This approach … is termed as federated learning.”. Also see figure 1caption which cites “The base layers are shared with the parameter server while the personalization layers are kept private by each device.”. Thus, each client device obtains parameters of the base layers from the parameter server, i.e., obtains “indication information” from a first device (the parameter server).
wherein the indication information indicates information about a first neural network model, (Arivazhagan is expressly about deep feedforward neural networks, see page 1 abstract which cites “This paper proposes FedPer, a base + personalization layer approach for federated training of deep feed forward neural networks, which can combat the ill-effects of statistical heterogeneity.”. Also see Figure 1 and page 3 Section 3.1 which states “Consider the setup shown in Figure 1 where all user devices share the same base layers and have unique personalization layers constituting the deep feed forward neural network models. We denote the number of base and personalized layers on each client by positive integers KB and KP respectively and let there be N user devices.” Thus, the base layer weight matrices, WB, and the per-user personalized layer weight matrices WP define each user’s neural network model. Further section 3.1 and equations 1-3 teaches the parameters sent from server to client (base-layer weights WB) are therefore information about that client’s neural network model. the client uses them as part of its model definition. Thus the “indication information” indicates information (weights/structure) about a first neural network model at the client.)
the first neural network model comprises a dedicated layer network and a common layer network, (Arivazhagan page 2 section 1.1 cites “We propose to capture personalization aspects in federated learning by viewing deep learning models as base + personalization layers as illustrated in Figure 1. Our training algorithm comprises of the base layers being trained by federated averaging (or some variant thereof) and personalization layers being trained only from local data…” and Figure 1 caption cites “All user devices share a set of base layers with same weights (colored blue) and have distinct personalization layers that can potentially adapt to individual data.”, these two citations teach a first neural network model comprising dedicated layer network (red layer in model A, green layer in model B, and yellow layer in model N) with common layer network as blue in all the models.)
the dedicated layer network is dedicated to the first neural network model, (Arivazhagan teaches personalization layers are distinct and private per device, see Figure 1 caption which cites “All user devices share a set of base layers with same weights… and have distinct personalization layers that can potentially adapt to individual data. The base layers are shared with the parameter server while the personalization layers are kept private by each device.”. Thus, for user j, the personalization layers WPj are specific to that user’s model, not shared with others. Those personalization layers therefore form a dedicated layer network dedicated to the first (user-specific) neural network model.)
and the common layer network is a common network of the first neural network model and a second neural network model; and (Arivazhagan figure 1 shows the blue layers are common to all models, thus common layer is common in both the first and second neural network.)
determining the information about the first neural net- work model based on the indication information. (Arivazhagan page section 1.1 teaches Base layers are trained collaboratively using FedAvg, wherein it cites “Our training algorithm comprises of the base layers being trained by federated averaging… and personalization layers being trained only from local data…”. Also see Algorithm.1 and algorithm 2 on page 4 which teaches, the parameter server aggregates client updates to produce updated base layer parameters, WB. These updated base parameters are shared back with clients (indication information), and each client combines the received base parameters with its local personalization parameters, WPj, and runs local stochastic gradient descent (SGD) to update its model. Thus, each client determines its current neural network model (the tuple of base and personalization weights) based on the indication information (server-sent base parameters) plus local.)
In regards to claim 2, Arivazhagan discloses the information about the first neural network model comprises at least one of the following:
neural network cascaded structure information (Arivazhagan teaches neural network cascade structure in figure 1 as it shows the common layers are followed by dedicated/personalized layers), common layer network structure information (Arivazhagan section 3.1 teaches the structure of the shared base layers, i.e., which layers form the base, their ordering (WB,1 … WB,KB), and their activation functions aB,1 … aB,KB, , thus you base layers orders and thus structure.), a common layer network parameter, (Arivazhagan section 3.1 teaches the actual weight values of the base layers, as WB = (WB,KB, …, WB,1), which are trained via federated averaging and “shared with the parameter server.”) dedicated layer network structure information, (Arivazhagan section 3.1 teaches the structure of the personalization layer structure for user j, which is the sequence of personalization layers (WPj,1 … WPj,KP ) and their activation functions (aPj,1 … aPj,KP) which comes after the common layers in figure 1.); or a dedicated layer network parameter. (Arivazhagan section 3.1 teaches the personalization weights WPj = (WPj,KP, …, WPj,1), which are “trained only from local data with stochastic gradient descent.)
In regards to claim 3, Arivazhagan discloses the apparatus according to claim 2, wherein that the information about the first neural network model comprises the neural network cascaded structure information comprises:
the information about the first neural network model indicates the neural network cascaded structure; or the information about the first neural network model indicates a configured neural network cascaded structure from a neural network cascaded structure candidate set. (See Arivazhagan section 3.1 and algorithms 1 and 2 that teaches weight matrix of base and personalized layers, which contains cascade information as it has the order and weight associated with each layer. Also see figure 1. Further see section 4.3 and figure 4 that teaches configuration of the cascade structure candidate set where figure 4 teaches accuracy based on the number blocks in the personalized layer, wherein it tests 1,2,3 and 4))
In regards to claim 4, Arivazhagan discloses the apparatus according to claim 2, wherein that the information about the first neural network model comprises the common layer network structure information comprises:
the information about the first neural network model indicates the common layer network structure; or the information about the first neural network model indicates a configured common layer network structure from a common layer network structure candidate set. (Arivazhagan section 3.1 teaches the structure of the shared base layers, i.e., which layers form the base, their ordering (WB,1 … WB,KB), and their activation functions aB,1 … aB,KB, , thus you base layers orders and thus structure.)
In regards to claim 5, Arivazhagan discloses the apparatus according to claim 2, wherein that the information about the first neural network model comprises common layer network parameter information comprises:
the information about the first neural network model indicates the common layer network parameter; or the information about the first neural network model indicates a configured common layer network parameter from a common layer network parameter candidate set. (Arivazhagan section 3.1 teaches the actual weight values of the base layers, as WB = (WB,KB, …, WB,1), which are trained via federated averaging and “shared with the parameter server.”)
In regards to claim 6, Arivazhagan discloses the apparatus according to claim 2, wherein that the information about the first neural network model comprises the dedicated layer network structure information comprises:
the information about the first neural network model indicates the dedicated layer network structure; or the information about the first neural network model indicates a configured dedicated layer network structure from a dedicated layer network structure candidate set. : (Arivazhagan section 3.1 teaches the structure of the personalization layer structure for user j, which is the sequence of personalization layers (WPj,1 … WPj,KP ) and their activation functions (aPj,1 … aPj,KP) which comes after the common layers in figure 1.)
In regards to claim 7, Arivazhagan discloses the apparatus according to claim 2, wherein that the information about the first neural network model comprises dedicated layer network parameter information comprises:
the information about the first neural network model indicates the dedicated layer network parameter; or the information about the first neural network model indicates a configured dedicated layer network parameter from a dedicated layer network parameter candidate set. (Arivazhagan section 3.1 teaches the personalization weights WPj = (WPj,KP, …, WPj,1),
In regards to claim 8, Arivazhagan discloses the apparatus according to claim 1, wherein the programming instructions, when executed by the at least one processing circuit, cause the apparatus perform the following:
receiving configuration information from the first device, wherein the configuration information indicates one or more of the following contents: a neural network cascaded structure information candidate set, a common layer net- work structure information candidate set, a common layer network parameter candidate set, a dedicated layer network structure information candidate set, or the dedicated layer network parameter candidate set. (See Arivazhagan section 3.1 and algorithms 1 and 2 that teaches weight matrix of base and personalized layers, which contains cascade information as it has the order and weight associated with each layer. Also see figure 1. Further see section 4.3 and figure 4 that teaches configuration of the cascade structure candidate set where figure 4 teaches accuracy based on the number blocks in the personalized layer, wherein it tests 1,2,3 and 4.)
In regards to claim 16, Arivazhagan discloses a communication apparatus, comprising:
at least one processing circuit; and one or more memories coupled to the at least one processing circuit and storing programming instructions for execution by the at least one processing circuit to cause the apparatus perform the following: (Arivazhagan page 1 section 1 teaches mobile devices having hardware and software that execute machine learning and communicate with a parameter sever. As it is mobile devices, it thus has at least one processing circuit and memory.)
determining indication information, wherein the indication information indicates information about a first neural network model, the first neural network model comprises a dedicated layer network and a common layer network, the dedicated layer network is dedicated to the first neural network model, and the common layer network is a common network of the first neural network model and a second neural network model; and (Arivazhagan is expressly about deep feedforward neural networks, see page 1 abstract which cites “This paper proposes FedPer, a base + personalization layer approach for federated training of deep feed forward neural networks, which can combat the ill-effects of statistical heterogeneity.”. This teaches the first neural network model has both dedicated and common layer, and the common layers being in both a first and second neural networks in figure 1. Also see Figure 1 and page 3 Section 3.1 which states “Consider the setup shown in Figure 1 where all user devices share the same base layers and have unique personalization layers constituting the deep feed forward neural network models. We denote the number of base and personalized layers on each client by positive integers KB and KP respectively and let there be N user devices.” Thus, the base layer weight matrices, WB, and the per-user personalized layer weight matrices WP define each user’s neural network model. Further section 3.1, equations 1-3, and algorithm 1 and 2 in section 3.2 teaches the parameters sent from server to client (base-layer weights WB) are therefore information about that client’s neural network model, the client uses them as part of its model definition. Thus the “indication information” indicates information (weights/structure) about a first neural network model at the client.)
sending the indication information to a second device. (Arivazhagan section 3.1, equations 1-3 and section 3.2 algorithm 1 and 2 teaches the parameters sent from server to client (base-layer weights WB)
In regards to claim 17, Arivazhagan discloses the apparatus according to claim 16, wherein the information about the first neural network model comprises at least one of the following: neural network cascaded structure information (Arivazhagan teaches neural network cascade structure in figure 1 as it shows the common layers are followed by dedicated/personalized layers), common layer network structure information (Arivazhagan section 3.1 teaches the structure of the shared base layers, i.e., which layers form the base, their ordering (WB,1 … WB,KB), a common layer network parameter (Arivazhagan section 3.1 teaches the actual weight values of the base layers, as WB = (WB,KB, …, WB,1), which are trained via federated averaging and “shared with the parameter server.”), dedicated layer network structure information (Arivazhagan section 3.1 teaches the structure of the personalization layer structure for user j, which is the sequence of personalization layers (WPj,1 … WPj,KP ) and their activation functions (aPj,1 … aPj,KP) which comes after the common layers in figure 1.), or a dedicated layer network parameter. (Arivazhagan section 3.1 teaches the personalization weights WPj = (WPj,KP, …, WPj,1), which are “trained only from local data with stochastic gradient descent.)
In regards to claim 18, Arivazhagan discloses the apparatus according to claim 16, wherein the programming instructions, when executed by the at least one processing circuit, the processor is further configured to execute instructions stored in the memory to cause the apparatus perform the following: sending configuration information to the second device, wherein the configuration information indicates one or more of the following content: a neural network cascaded structure information candidate set, a common layer network structure information candidate set, a common layer network parameter candidate set, a dedicated layer network structure information candidate set, or a dedicated layer network parameter candidate set. (See Arivazhagan section 3.1 and algorithms 1 and 2 that teaches weight matrix of base and personalized layers, which contains cascade information as it has the order and weight associated with each layer. Also see figure 1. Further see section 4.3 and figure 4 that teaches configuration of the cascade structure candidate set where figure 4 teaches accuracy based on the number blocks in the personalized layer, wherein it tests 1,2,3 and 4.)
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.
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.
Claims 9 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Arivazhagan et al. (“Federated Learning with Personalization Layers” – hereinafter Arivazhagan) above, and further in view of Mittal et al. (US 10,911,123 B2 – hereinafter Mittal)
In regards to claim 9, Arivazhagan discloses the apparatus according to claim 1, but fails to explicitly disclose wherein the programming instructions, when executed by the at least one processing circuit, cause the apparatus perform the following: sending auxiliary information to the first device, wherein the auxiliary information indicates a wireless system parameter for a communication between the first device and a second device.
Mittal discloses sending auxiliary information to the first device, wherein the auxiliary information indicates a wireless system parameter for a communication between the first device and a second device. (Mittal teaches channel state information, CSI, feedback from a UE (user equipment/second device) to a base station (first device). The abstract states that, after processing a received reference signal, the UE: “generates a combinatorial codeword representing the set of tap indices and reports the combinatorial codeword for the set of tap indices as part of a CSI feedback report.”. This teaches a CSI feedback report which is report describing the radio channel between the base station (first device) and the UE (second device). In NR/LTE, CSI feedback is sent from UE to BS and used by the BS to configure the downlink/uplink communication. Also see column 12 lines 22-30 which teaches the CSI report being generated by the edge device (user equipment) and sent the base station (server), wherein it cites “To support spatial multiplexing and MU-MIMO, the remote unit 105 provides CSI feedback 125 to the base unit 110 using Type-II codebook compression using multistage quantization. The remote unit 105 generates a set of modified channel matrix (HsB W 1), where (HsB) is an estimate of channel matrices for a set of sub-bands and (W1) is the beam space matrix. The remote unit 105 also generates a set of singular vector coefficients (W2) from the modified channel matrix.”.)
It would have been obvious to one of ordinary skill before the earliest effective filing date of the claimed invention to modify the teachings of Arivazhagan with that of Mittal in order to allow for sending auxiliary information between to devices wherein the auxiliary information indicates a wireless system parameter for communication between two devices as both reference deal with systems that use base stations/servers along with edge devices. The benefit of doing so is it allows for the better communication between the two devices by allowing them to pass communication parameters that allow for optimization the communication process.
In regards to claim 19, it apparatus embodiment of claim 9 with similar limitations and thus rejected using the same reasoning as that in claim 9.
Claims 10 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Arivazhagan et al. (“Federated Learning with Personalization Layers” – hereinafter Arivazhagan) above, and further in view of Prakash et al. (US2019/013893 A1 – hereinafter Prakash).
In regards to claim 10, Arivazhagan discloses the apparatus according to claim 1, but fails to further disclose wherein the programming instructions, when executed by the at least one processing circuit, cause the apparatus perform the following: sending capability information to the first device, wherein the capability information indicates one or more pieces of the following information:
(1) whether to support using a neural network to replace or implement a function of a communication module;
(2) whether to support a neural network structure in which the common layer network and the dedicated layer network are cascaded;
(3) whether to support receiving of at least one of dedicated layer network structure information or dedicated layer network parameter through signaling;
(4) stored dedicated layer network structure information predefined in a protocol or pre-configured;
(5) a stored dedicated layer network parameter predefined in the protocol or pre-configured;
(6) a memory space used to store neural network cascaded structure information, common layer network structure information, a common layer network para- meter, the dedicated layer network structure information, and/or the dedicated layer network parameter; or
(7) computing power information used to run the neural network.
Prakash discloses wherein the programming instructions, when executed by the at least one processing circuit, cause the apparatus perform the following: sending capability information to the first device, wherein the capability information indicates one or more pieces of the following information: (7) computing power information used to run the neural network. (Prakash para. [0035] teaches federated learning wherein a central server communicates with a plurality of edge devices wherein it cites “Recently, federated learning has been proposed for distributed GD computation, where learning takes place by a federation of client compute nodes that are coordinated by a central server. Each client compute node fetches a global model, updates the global model using its local data, and communicates the updated model to the central server.”. Para. [0040] teaches task or jobs, which are machine learning computations, given to edge devices based on operational parameters, which are device capabilities or computer power information, wherein it cites “… the computations are balanced across the plurality of edge compute nodes based on statistical knowledge of various operational parameters including, but not limited to, link quality, processing speed, and battery life.”. Then in claim 15 Prakash teaches sending and receiving capability information indicating computer power information used to run machine learning computations, wherein it cites “…wherein the operational parameters of the corresponding compute nodes include network conditions experienced by the corresponding compute nodes and compute node capabilities of the corresponding compute nodes, and the communication circuitry is arranged to receive, from the corresponding compute nodes, an indication of the operational parameters of the corresponding compute nodes, and wherein the compute node capabilities include one or more of a processor speed, memory utilization, memory or storage size, link adaptation capabilities, available battery power, a battery power budget, an average computation time per workload, and an achievable data rate per channel usage.”. Also, para. [0022] teaches the process is applicable to neural networks wherein it cites “Although the embodiments herein are discussed in terms of GD algorithms for linear regression, the distributed training embodiments discussed herein are applicable to more complex ML algorithms such as deep neural networks and the like.”.
It would have been obvious to one of ordinary skill in the art before earliest effective filing date of the claimed invention to modify the teachings of Arivazhagan with that of the Prakash in order to allow for sending the computing power capabilities of devices in the federated learning system as both references deal with federated learning and the benefit of doing so it allows for creating a more efficient network by ensuring that edge devices in the network are actually capable of performing the machine learning tasks assigned to them.
In regards to claim 20, it apparatus embodiment of claim 10 with similar limitations and thus rejected using the same reasoning as that in claim 10.
Claims 11-13 are rejected under 35 U.S.C. 103 as being unpatentable over Claussen et al. (US 2020/0254609 A1 – hereinafter Claussen) and further in view of Arivazhagan et al. (“Federated Learning with Personalization Layers” – hereinafter Arivazhagan).
In regards to claim 11, Claussen discloses a communication apparatus, comprising: at least one processing circuit; (Claussen fig. 4 element 420 teaches processors.) and one or more memories coupled to the at least one processing circuit and storing programming instructions for execution by the at least one processing circuit to a processor and a memory, wherein the memory is coupled to the processor, and the processor is configured to execute instructions stored in the memory to cause the apparatus perform the following: (Claussen fig. 4 shows memory (element 430, 442, and 441 couple to the processor).
obtaining first indication information, wherein the first indication information indicates common information of a first neural network model, the first neural network model comprises a dedicated layer network and a common layer network, and the common information of the first neural network model comprises information about the common layer network; (Claussen figure. 1d teaches a first neural network model (Complete Neural Network, element 115) comprising a common layer network (Machine-Specific Module, element 105) and a dedicated layer network (Task-Specific Module, element 110). Then para. [0019] teaches the modules have weights and structure of the neural network wherein it cites “The term "module," as used herein refers to an individual neural network layer or a group of neural network layers. For each module, the information that needs to be stored and exchanged is composed by the weights and the network shape, such as the numbers of nodes per layer. This information could be managed, for example, via Extensible Markup Language (XML) files or user-defined binary files.” Thus, it has information about both common and dedicated layers of the neural network.)
obtaining second indication information from the first device, wherein the second indication information indicates information about the dedicated layer network. (Claussen figure. 1D teaches obtaining the second indication information (task-specific module, element 110) from the a first device (element 150, task planning computer.)
However, Claussen does not explicitly disclose wherein the first indication information is obtained from a first device.
Arivazhagan discloses a communication apparatus wherein a first indication information is obtained from a first device. (Arivazhagan’s federated learning (FL) setup uses a parameter server (cloud) that coordinates training with client devices as cite on page 1 section 1 wherein it cites “…execute machine learning (ML) model training computations on device with locally available data and occasional communication with an aggregating parameter server. This approach … is termed as federated learning.”. Also see figure 1caption which cites “The base layers are shared with the parameter server while the personalization layers are kept private by each device.”. Thus, each client device obtains parameters of the base layers from the parameter server, i.e., obtains “indication information” from a first device (the parameter server).)
It would have been obvious to one of ordinary skill in the art before earlies effective filing date of the claimed invention to modify the teachings of Claussen with that of Arivazhagan in order to allow for sending or obtaining first indication information about common layers from a first device as both references deal with transferring weights and neural network parameters between devices and the benefit of doing so it allows fast starting of devices by allowing for transferring neural networks and setups between devices.
In regards to claim 12, Claussen in view of Arivazhagan discloses the apparatus according to claim 11, wherein the information about the common layer network comprises at least one of common layer network structure information or a common layer network parameter. (Arivazhagan section 3.1 teaches the structure of the shared base layers (common base layers), which layers form the base, their ordering (WB,1 … WB,KB), and their activation functions aB,1 … aB,KB, , thus you base layers orders and thus structure. Arivazhagan also teaches a common layer network parameter in section 3.1, wherein it teaches the actual weight values of the base layers, as WB = (WB,KB, …, WB,1), which are trained via federated averaging and shared with the parameter server.)
In regards to claim 13, Claussen in view of Arivazhagan discloses the apparatus according to claim 11, wherein the information about the dedicated layer network comprises at least one of the following: neural network cascaded structure information, dedicated layer network structure information, or a dedicated layer network parameter. (Claussen para. [0019] teaches the modules have weights and structure of the neural network wherein it cites “The term "module," as used herein refers to an individual neural network layer or a group of neural network layers. For each module, the information that needs to be stored and exchanged is composed by the weights and the network shape, such as the numbers of nodes per layer. This information could be managed, for example, via Extensible Markup Language (XML) files or user-defined binary files.” Thus, it has information about both common and dedicated layers of the neural network, wherein it has weights (parameters) and structure (shape).)
Claim 14 is rejected under 35 U.S.C. 103 as being unpatentable over Claussen et al. (US 2020/0254609 A1 – hereinafter Claussen) in view of Arivazhagan et al. (“Federated Learning with Personalization Layers” – hereinafter Arivazhagan) and further in view of Tabet et al. (US 9,872,286 B2 – hereinafter Tabet).
In regards to claim 14, Claussen in view of Arivazhagan disclose the apparatus according to claim 11, wherein indication information is transmitted between devices. (See Arivazhagan fig. 1 on page 2 that teaches data shared between a server and mobile devices, also see page 10 section A.3. first paragraph that cites
PNG
media_image1.png
102
953
media_image1.png
Greyscale
, also Claussen para. [0020] teaches transmitting data (task-specific module, element 110) between the computer (element 150) and the robot (120) in figure. 1D. )
However, Claussen in view of Arivazhagan does not disclose explicitly disclose wherein the first indication information is carried in a broadcast message, a system message, or a common message.
Tabet disclose receiving information from carried in a broadcast, system or common message. (Tabet claim cites “wherein the base station is configured to: broadcast first system information blocks (SIBs) encoded using a first coding rate and a first identifier; and broadcast second SIBs encoded using a second coding rate that is lower than the first coding rate and a second identifier, wherein a compressed SIB type in the second SIBs includes only a portion of the information included in the same non-compressed SIB type in the first SIBs and wherein the second SIBs are usable by a class of user equipment devices (UEs) having a limited link budget to determine access parameters for the base station.”. This teaches a base station sending message to edge devices (user equipment) containing indication information carried in system messages.)
It would have been obvious to one of ordinary skill in the art before the earliest effective filing date of the claimed invention to modify the teachings of the Claussen in view of Arivazhagan with that of Tabet in order to allow for sending information using system messages as the reference deal with sending information between devices. The benefit of doing so is it provides essential network configuration, mobility parameters and access rules for the mobile devices to connect and maintain communication.
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Claussen et al. (US 2020/0254609 A1 – hereinafter Claussen) in view of Arivazhagan et al. (“Federated Learning with Personalization Layers” – hereinafter Arivazhagan) and further in view of Yuan et al. (US 2023/0030642 A1 – hereinafter Yuan).
In regards to claim 15, Claussen in view of Arivazhagan disclose the apparatus according to claim 11, but fails to disclose wherein the second indication information is carried in radio resource control (RRC) signaling specific to a second device, a medium access control control element (MAC CE), or downlink control information (DCI).
Yuan disclose wherein information is carried in radio resource control (RRC) signaling specific to a second device, a medium access control control element (MAC CE), or downlink control information (DCI). (Yuan para. [0032] cites “Specifically, the base station 102 includes a communication component 141 in communication with the UE 104 and having a downlink assignment index (DAI) indication component 143 configured to generate DAI information 145 for transmission with a single downlink control information (DCI) 147 that schedules resources for monitoring by the UE 104 on a plurality of CCs.”, this teaches a bases station generates control information (DAI), encodes it in DCI, and transmit to the user devices.)
It would have been obvious to one of ordinary skill in the art before the earliest effective filing date of the claimed invention to modify the teachings of the Claussen in view of Arivazhagan with that of Yuan in order to allow for sending information using DCI as all the reference deal with passing information between a base station/server to edge devices. The benefit of doing so it allow for faster scheduling of the between devices.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PAULINHO E SMITH whose telephone number is (571)270-1358. The examiner can normally be reached Mon-Fri. 10AM-6PM CST.
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, Abdullah Kawsar can be reached at 571-270-3169. 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.
/PAULINHO E SMITH/Primary Examiner, Art Unit 2127