NEURAL NETWORK TRAINING BASED ON CAPABILITY

§102 §103 §112

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 . Remarks This Office Action is in response to applicant’s amendment filed on July 22, 2026, under which claims 1-32 are pending and under consideration. Response to Arguments Applicant’s amendments have overcome the previous claim objections and the previous § 103 prior art rejections. However, upon further consideration, new grounds of rejection have been made under § 112(b) and § 103. Applicant’s arguments directed to the previous § 103 rejections have been fully considered but are moot under the new grounds of rejection, since they rely on new limitations that are addressed by a newly applied reference, Vitthaladevuni, as detailed in the rejections below. Claim Objections Claim 18 is objected to because of the following informalities: In claim 18, “CSA” should be “CSI”. Appropriate correction is required. Claim Rejections - 35 USC § 112 The following is a quotation of 35 U.S.C. 112(b): (b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention. The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph: The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention. Claims 1-32 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. In claims 1, 9, 15, 21, and 27, it is unclear whether “the CSI information” (two instances in the bottom half of the claim, appearing respectively in the last and third-to-last paragraph of the claim) refers to the CSI information recited in the phrase “…training configuration that identifies CSI information to be use by the UE” or instead refers to the CSI information recited in the phrase “to compress and decompress CSI information.” That is, since the term “CSI information” is used twice, there is a conflict of antecedent when the term “the CSI information” is later used, and it is also raises an issue of whether the two antecedent instances of “CSI information” are the same element or different elements. For purposes of examination, the Examiner believes that the two prior recitations of “CSI information” are intended to be distinct concept. Generally, compression and decompression are performed on CSI (which stands for “channel state information”), which is a term of art in telecom and 3GPP. On the other hand, the term “CSI information” (which seems to literally stand for “channel state information information”) seems to be used in a manner that broadly covers information that pertains to CSI information, such as the CSI report format. Based on this understanding, the Examiner interprets “the CSI information” in these claims (as well as in the dependent claims discussed below) as having the meaning of “the CSI information to be used by the UE” (i.e., referring to the first prior instance of this term), and interprets the second prior recitation of “CSI information” (in the phrase “to compress and decompress CSI information”) as a distinct element that generally refers to a particular type of data that is being compressed, such as CSI. Claims 3, 16, and 22 also use the term “the CSI information,” and are therefore given the same basis of rejection and interpretation described above. Additionally in claims 1, 9, 15, 21, and 27, the limitation of “the identified CSI information” in the second-to-last paragraph lacks antecedent basis because the previous part of the claim does not use the term “identify” or “identifying” in conjunction with “CSI information.” Instead, the claim recites “identify one or more CSI resources.” Therefore, the claim would provide antecedent basis for “the identified CSI resources” but not for “the identified CSI information.” For purposes of examination, and in light of the previous matter, the Examiner interprets the above phrase as having the meaning of “the CSI information to be used by the UE.” The above grounds of rejection given for the independent claims are also given to each of their dependent claims 2-8, 10-14, 16-20, 22-26, and 28-32, since these claims incorporate the indefinite recitations of their parent claims. Claim Rejections - 35 USC § 103 In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status. The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action: A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made. 1. Claims 1-6, 8-12, 14-19, 21-24, 26-29 and 31-32 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Wang et al. (US 2023/0259789 A1) (“Wang”) in view of Vitthaladevuni et al. (US 2023/0359886 A1) (“Vitthaladevuni”) (cited in the Office Action of 9/30/2025) and Timo et al. (US 2022/0149904 A1) (“Timo”). As to claim 1, Wang teaches one or more processors, comprising circuitry to: [[0033]: “The base station 120 also includes processor(s) 260 and computer-readable storage media 262.” [0042]: “The core network server 302…includes processor(s) 304 and computer-readable storage media 306.” Any one or combination of the processors described above may correspond to the instant limitation.] obtain information about one or more capabilities of a user equipment (UE) to train one or more neural networks [[0068]: “In aspects, the base station receives a UE capability information message (not illustrated) from each UE and selects the initial ML configuration based on a common UE capability between the UEs 111, 112, and 113. As one example, the UE capability information message includes ML capabilities (e.g., any one or more of: supported ML architectures, supported number of layers, available processing power, memory/storage capabilities, available power budget, fixed-point processing vs. floating-point processing, maximum kernel size capability, computation capability), and the base station 120 selects the initial ML configuration based on a common ML capability supported by the UEs 111, 112, and 113.” That is, the base station obtains the UE capabilities of each UE, which includes ML capabilities. In the present context, this is for training a neural network. See [0072]: “At 615, 616, and 617, the UEs 111, 112, and 113, respectively, receive the directions to form the DNN and then form the DNN using the initial ML configuration”; [0029]: “the base station 120 indicates, to the UE federated learning manager 220, to perform the training procedure”] to generate information about one or more fifth-generation (5G) signals; [The DNNs at the UE perform signal processing. See [0054]: “the DNNs 408 can perform any combination of extracting data embedded on the Rx signal, recovering binary data, correcting for data errors based on forward error correction applied at the transmitter block, extracting payload from frames and/or slots, and so forth.” The signal is a “5G” signal as described in [0021]: “The wireless links 130 may include one or more wireless links (e.g., radio links) or bearers implemented using…Fifth Generation New Radio (5G NR), and so forth.” [0025]: “The antennas 202 and the RF front end 204 can be tuned to, and/or be tunable to, one or more frequency bands defined by the 3GPP LTE and 5G NR communication standards and implemented by the LTE transceiver 206, and/or the 5G NR transceiver 208.”] determine, based at least in part on the one or more capabilities of the UE, a […] training configuration […] [0068]: “In aspects, the base station receives a UE capability information message (not illustrated) from each UE and selects the initial ML configuration based on a common UE capability between the UEs 111, 112, and 113. As one example, the UE capability information message includes ML capabilities (e.g., any one or more of: supported ML architectures, supported number of layers, available processing power, memory/storage capabilities, available power budget, fixed-point processing vs. floating-point processing, maximum kernel size capability, computation capability), and the base station 120 selects the initial ML configuration based on a common ML capability supported by the UEs 111, 112, and 113.” send, based at least one the […] training configuration, one or more instructions to the UE to train […(a model)] [[0071]: “At 610, the base station 120 directs each UE in a set of UEs (e.g., UE 111, UE 112, UE 113) to form a (respective) DNN using the initial ML configuration. To illustrate, the base station 120 transmits an indication of an index value of a neural network table to the UEs 111, 112, and 113.” [0073]: “At 620, the base station 120 requests each UE…to report updated ML information generated using a training procedure and input data local to the respective UE. …To illustrate, the base station implicitly requests the UE to report the updated ML information (and/or to perform the training procedure) by indicating one or more update conditions that specify rules or instructions on when to report the updated ML information.” [0074]: “In aspects, the base station 120 directs each UE to perform an online training procedure, such as an online training procedure that trains the DNNs while processing the wireless network communications.” That is, the indication of the neural network table values, the request to report updated ML information, and the request to perform online training collectively constitute one or more instructions to train the neural network, since the training is performed at least in part by the UE. These communications are based on the capability information, since they are based on the initial ML configuration that was selected “based on a common ML capability supported by the UEs” ([0068]).] Wang does not explicitly teach: (1) the training configuration being “a channel state information (CSI) training configuration that identifies CSI information to be used by the UE to train an autoencoder”; and (2) the instructions being to train “an autoencoder to compress and decompress CSI information”; and (3) “the one or more instructions including: at least one instruction to identify one or more CSI resources to be used by the UE according to the information, at least one instruction to measure CSI using the one or more CSI resources, and at least one instruction to report the identified CSI information using one or more formats indicated by the CSI instruction.” Vitthaladevuni teaches: a channel state information (CSI) training configuration that identifies CSI information to be used by the UE to train an autoencoder [This reference teaches a “training configuration” in [0181], quoted below: [0181] …The network entity 1104 may configure the UE 1102 with neural network training parameters for the UE 1102 to train the neural network configured for various procedures. The neural network training parameters may include a CSI reporting ID, a channel state reference signal identifier, a channel state reference signal ID, a component carrier ID, a BWP ID, a neural network ID, a first indication of at least one layer to be trained, a second indication of at least one layer to be frozen, a group of multiple layers to be trained, a subset of layers to be trained, or a combination thereof. The UE 1102 may receive the configuration of neural network training parameters from the network entity 1104, and train the neural network configured for various procedures, such as wireless channel compression, wireless channel measurement, wireless interference measurement, UE 1102 positioning, or wireless waveform determination at the UE 1102… See also [0171]: “In some aspects of the disclosure, the wireless network entity may send a message to the UE configuring the neural network training, dynamically.” That is the “training configuration” is described here as including neural network training parameters, which includes CSI information such as a CSI reporting ID. Additionally the network entity sends a training command. Additionally, see also [0186]: “At 1110, The network entity 1104 may transmit a training command in a wireless message to indicate to the UE 1102 to apply the configuration to train the neural network at the UE 1102. That is, the network entity 1104 may transmit a separate training command in a wireless message to the UE 1102.” Thus, the “message” that includes the configuration (including parameters) ([0171], as quoted above) and the “message” that includes a training command ([0171]-[0186]) collectively correspond to the “one or more instructions” sent to the UE. The limitation of the “autoencoder” is discussed below.]; “an autoencoder” [While this specific term does not appear in the reference, it is understood that the model described in FIG. 4A is an autoencoder, because it includes an encoder (CSI instance encoder 404 and CSI sequence encoder 406), and a decoder (CSI sequence decoder 414 and CSI sequence decoder 418), and is overall input/out from the encoder to decoder is “DL Channel Estimates 402” and “Re-constructed DL Channel Estimates” as described in detail in [0079]-[0082]. This configuration is the same as the configuration of the autoencoder shown in Figure 2 of the instant application, which similarly has a UE-side encoder and a BS-side decoder respectively for encoding and decoding the downlink channel. Furthermore, the autoencoder is trained. In general, the autoencoder is a NN as described in [0080]: “The CSI sequence encoder 406 may be based on a long short term memory (LSTM) network, whereas the CSI instance encoder 404 may be based on a feedforward network.” These layers are collectively referred to a “neural network” in the context of training. See also [0168]-[0169]: “Referring again to FIG. 4A, the CSI sequence encoder 406 may capture time-dependency (for example, the Doppler component), and the CSI instance encoder 404 may capture Largely time-invariant aspects of the channels (for example, number of reflector, scatterers, or angle that the beam reaches the antenna of the UE). … For example, in case the UE is stationary in a relatively stable environment, then the change to the channel due to the Doppler effect may be mild. In such a case, the UE may not need to train every layer of the neural network. Accordingly, the UE may train the time dependency capturing layers.” Note that while only the encoder is trained at the UE, the instant claim does not require both the encoder and decoder portions of the autoencoder to be trained only at the UE. Therefore, the training of either the encoder and/or decoder part of the autoencoder constitutes training the autoencoder.] “to compress and decompress CSI information” [[0168]: “The UE may train the neural network to perform one or more of a wireless channel compression at the UE”; [0098], quoted below: [0098] The neural network based CSI based on machine learning or a neural network, such as described in connection with FIG. 4A, may compress the downlink channel in a more comprehensive manner. For example, in Type II CSI, a sub-band size may be fixed for all sub-bands for which the UE reports CSI. For example, the sub-band granularity (e.g., sub-band) size may not be a function of a sub-band index within a bandwidth part (BWP). For some frequency bands, the sub-band size may provide more granularity that is needed. In other frequency bands, the sub-band size may not provide enough granularity. The neural network based CSI may address the problems of a fixed sub-band size by providing CSI over an entire channel, for example. The neural network based CSI may be configured to compress some sub-bands with greater or lesser accuracy… That is, the downlink channel corresponds to a particular type of “CSI information” that is being compressed. (Examiner’s Note: Because the term “CSI information” (literally channel state information information) is used broadly in the claim, this term also covers information pertaining to CSI information, e.g., information pertaining to the downlink channel including the configurations sent by the network that relate to CSI.) Decompression is performed as described in [0158]: “For example, the second device may decode the compressed data set using one or more decompression operations and reconstruction operations associated with a neural network, the one or more decompression and reconstruction operations being based at least in part on a set of features of the compressed data set to produce a reconstructed data set, as described above.” See also [0144]: “The decoding device may perform an Mx64Vfully connected operation to decompress the low dimension vector of size M into a spatial-temporal feature data set… The VxW convolution operation may decompress spatial features from a reduced dimension for each tap.”]; and “the one or more instructions including: at least one instruction to identify one or more CSI resources to be used by the UE according to the information, [[0205]: “The network entity may instruct the UE to train the neural network to perform one or more of a wireless channel compression at the UE, a wireless channel measurement at the UE, a wireless interference measurement at the UE, UE positioning, wireless waveform determination at the UE, etc.” [0181]: “The neural network training parameters may include a CSI reporting ID, a channel state reference signal identifier, a channel state reference signal ID…” [0074]: “The base station may configure the UE with a CSI resource configuration that indicates the type of reference signal, e.g., a non-zero power CSI-RS (NZP CSI-RS), SSB, CSI-IM resource, etc.” (See also [0074]-[0077] for additional example). That is, the “channel state reference signal identifier” and the “type of reference signal” are examples of “one or more CSI resources” to be used by the UE.] at least one instruction to measure CSI using the one or more CSI resources,” [[0079]: “As illustrated at 402, the encoding device 400 measures downlink channel estimates based on downlink signals from the base station, such as CSI-RS, SSB, CSI-IM resources, etc., that is input for encoding. A downlink channel estimate instance at time t is represented as H(t) and is provided to a CSI instance encoder 404 that encodes the single CSI instance for time t and outputs the encoded CSI instance for time t as m(t) to a CSI sequence encoder 406.” This is based on the reference signal. See [0074]: “Among other examples, a UE may perform measurements on downlink signals, such as reference signal, from a base station and may provide a CSI report…The base station may configure the UE to perform the CSI measurements, e.g., with a CSI measurement configuration. The base station may configure the UE with a CSI resource configuration that indicates the type of reference signal, e.g., a non-zero power CSI-RS (NZP CSI-RS), SSB, CSI-IM resource, etc.”] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of Wang with the teachings of Vitthaladevuni by implementing an autoencoder as the model being trained in the system of Wang and the technique of instructing a UE to train an autoencoder, so as to arrive at the above features of the claimed invention. Doing so would have enabled a UE to train a neural network to perform functions such as “a wireless channel compression at the UE, a wireless channel measurement at the UE, a wireless interference measurement at the UE, UE positioning, wireless waveform determination at the UE, etc.” (Vitthaladevuni, [0168]), The combination of references thus far does not explicitly teach the limitation of also indicating “at least one instruction to report the identified CSI information using one or more formats indicated by the CSI information.” Timo teaches “at least one instruction to report the identified CSI information using one or more formats indicated by the CSI information” [[0242]: “When a new UE joins a cell, the serving BS may configure the UE CSI reporting settings as part of the RRC (see Section 5.2 in 3GPP TS 38.214 V15.4.0).” [0254]: “Each Reporting Setting ReportConfig is associated with a single downlink BWP (higher layer parameter bandwidthPartId) and contains the reported parameter(s) for one CSI reporting band: CSI Type (I or II) if reported, codebook configuration including codebook subset restriction, time-domain behavior, frequency granularity for CQI and PMI, measurement restriction configurations, the strongest layer indicator (SLI), the reported L1-RSRP parameter(s), CRI, and SSBRI (SSB Resource Indicator). Each ReportConfig contains a ReportConfigID to identify the ReportConfig, a ReportConfigType to specify the time domain behavior of the report (either aperiodic, semi-persistent, or periodic), a ReportQuantity to indicate the CSI-related or L1-RSRP-related quantities to report, a ReportFreqConfiguration to indicate the reporting granularity in the frequency domain.” See also [0259]: “If configured for Type-I or Type-II CSI feedback, the ReportConfig contains a CodebookConfig that specifies configuration parameters for Type-1 and Type-II CSI feedback.” That is, a format as defined by parameters such as ReportQuantity, ReportFreqConfiguration, and CodebookConfig, which are part of instructions provided by the base station.] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far with the teachings of Timo by modifying the one or more instructions to comprise “at least one instruction to report the identified CSI information using one or more formats indicated by the CSI information.” The motivation for doing so would have been to provide information to a UE so as to configure it to provide CSI reporting in a manner desired by the base station, as suggested by Timo ([0242]: “When a new UE joins a cell, the serving BS may configure the UE CSI reporting settings as part of the RRC.”) As to claim 2, the combination of Wang, Vitthaladevuni, and Timo teaches the one or more processors of claim 1, as set forth above. Vitthaladevuni further teaches “wherein the CSI information to be used by the UE includes the one or more CSI (CSI) resources.” [As noted in the rejection of the parent independent claim, the “channel state reference signal identifier” and the “type of reference signal” are examples of “one or more CSI resources” to be used by the UE. This information is also part of “CSI information to be used by the UE,” since the term “CSI information” (channel state information information, i.e., information pertaining to channel state information, including the information sent to the UE for operations related to CSI information) generically includes the aforementioned resources.] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far, including the teachings of Vitthaladevuni discussed above for the instant dependent claim, so as to also arrive at the further limitations of the instant dependent claim. The motivation for doing so is covered by the motivation given for Vitthaladevuni in the rejection of the parent independent claim. As to claim 3, the combination of Wang, Vitthaladevuni, and Timo teaches the one or more processors of claim 1, as set forth above. Timo further teaches “wherein the one or more formats to report the CSI information includes a channel state information (CSI) report format.” [[0242]: “When a new UE joins a cell, the serving BS may configure the UE CSI reporting settings as part of the RRC (see Section 5.2 in 3GPP TS 38.214 V15.4.0).” [0254]: “Each Reporting Setting ReportConfig is associated with a single downlink BWP (higher layer parameter bandwidthPartId) and contains the reported parameter(s) for one CSI reporting band: CSI Type (I or II) if reported, codebook configuration including codebook subset restriction, time-domain behavior, frequency granularity for CQI and PMI, measurement restriction configurations, the strongest layer indicator (SLI), the reported L1-RSRP parameter(s), CRI, and SSBRI (SSB Resource Indicator). Each ReportConfig contains a ReportConfigID to identify the ReportConfig, a ReportConfigType to specify the time domain behavior of the report (either aperiodic, semi-persistent, or periodic), a ReportQuantity to indicate the CSI-related or L1-RSRP-related quantities to report, a ReportFreqConfiguration to indicate the reporting granularity in the frequency domain.”] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far, including the teachings of Timo noted above, so as to also arrive at the claimed invention of the instant dependent claim. The motivation for doing so is the same as the one given for Timo in the rejection of the parent independent claim. As to claim 4, the combination of Wang, Vitthaladevuni, and Timo teaches the one or more processors of claim 1, wherein the circuitry is to cause the UE to train [Wang, [0029]: “The base station 120 indicates, to the UE federated learning manager 220, to perform a training procedure and/or to transmit updated ML information”] at least a portion of the autoencoder. [The “autoencoder” is taught by Vitthaladevuni and, in the rejection of the parent claim, is implemented as the neural network being trained at the UE in the combined teachings of the references. Therefore, the combination of the references and the associated rationale discussed in the rejection of the parent independent claim covers the instant limitation.] As to claim 5, the combination of Wang, Vitthaladevuni, and Timo teaches the processor of claim 1, wherein the one or more capabilities of the UE to train one or more neural networks to generate information about one or more 5G signals include […] training capabilities of the UE […] [Wang, [0068]: “…As one example, the UE capability information message includes ML capabilities (e.g., any one or more of: supported ML architectures, supported number of layers, available processing power, memory/storage capabilities, available power budget, fixed-point processing vs. floating-point processing, maximum kernel size capability, computation capability), and the base station 120 selects the initial ML configuration based on a common ML capability supported by the UEs 111, 112, and 113.”]. Vitthaladevuni further teaches “one or more channel state information (CSI) autoencoder” training and “the circuitry is to cause the UE to train an encoder of a CSI autoencoder.” [As discussed in the rejection of the parent independent claim, Vitthaladevuni teaches the training of an autoencoder. The training of the encoder specifically is taught because the encoder is on the UE (see [0136]), and the network entity also directs the UE to train specific parts of the encoder, as described in [0168]-[0169]: “Referring again to FIG. 4A, the CSI sequence encoder 406 may capture time-dependency (for example, the Doppler component), and the CSI instance encoder 404 may capture Largely time-invariant aspects of the channels (for example, number of reflector, scatterers, or angle that the beam reaches the antenna of the UE). … For example, in case the UE is stationary in a relatively stable environment, then the change to the channel due to the Doppler effect may be mild. In such a case, the UE may not need to train every layer of the neural network. Accordingly, the UE may train the time dependency capturing layers.”] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far, including the teachings of Vitthaladevuni discussed above for the instant dependent claim, so as to also arrive at the further limitations of the instant dependent claim. The motivation for doing so is covered by the motivation given for Vitthaladevuni in the rejection of the parent independent claim. As to claim 6, the combination of Wang, Vitthaladevuni, and Timo teaches the one or more processors of claim 1, wherein the circuitry is to cause a wireless radio network base station to train [Wang, [0080]: “Accordingly, at 640, the base station 120 receives updated ML information from at least some of the UEs, where the updated ML information can indicate any combination of ML parameters, ML architectures, and/or ML gradients.” Wang, [0082]: “In determining the common ML configuration, the base station 120 applies federated learning techniques that aggregate the updated ML information received from multiple UEs…the base station 120 performs averaging that aggregates ML parameters, gradients, and so forth.” Wang, [0084]: “At 655, the base station 120 directs the subset of UEs to update the DNN formed at 615 and at 616 using the common ML configuration determined at 650.” Note that this process of aggregating information constitutes “training” in the absence of further limitations defining what specific training operations are performed. The limitation of “wireless radio” is disclosed in e.g., [0021]: “The base stations 120 communicate with the user equipment 110 using the wireless links 130, which may be implemented as any suitable type of wireless link… The wireless links 130 may include one or more wireless links (e.g., radio links)”] “at least a portion of the autoencoder” [As discussed in the rejection of the parent independent claim, Vitthaladevuni teaches an autoencoder, which, in the combined teachings of the references as set forth in the rejection of the parent claim, is implemented as the neural network being trained at the UE. Therefore, the combination of the references and the associated rationale discussed in the rejection of the parent independent claim covers the instant limitation.] As to claim 8, the combination of Wang, Vitthaladevuni, and Timo teaches the one or more of claim 1, as set forth above. Vitthaladevuni further teaches “wherein the CSI autoencoder training configuration further comprises an identifier of a type of training to be performed by the UE based, at least in part on the one or more CSI autoencoder training capabilities” [[0173]: “For another example, the message may indicate the layer IDs to be trained and implicitly indicate that the remaining layers may be frozen. For another example, the message may indicate the layer IDs to be frozen and implicitly indicate that the remaining layers to be trained. For yet another example, the message may include a bit string of length corresponding to the number of layers, each bit indicating whether each layer should be frozen or trained.” [0168]-[0169]: “Referring again to FIG. 4A, the CSI sequence encoder 406 may capture time-dependency (for example, the Doppler component), and the CSI instance encoder 404 may capture Largely time-invariant aspects of the channels (for example, number of reflector, scatterers, or angle that the beam reaches the antenna of the UE). … For example, in case the UE is stationary in a relatively stable environment, then the change to the channel due to the Doppler effect may be mild. In such a case, the UE may not need to train every layer of the neural network. Accordingly, the UE may train the time dependency capturing layers.” That is, training different parts of the network constitutes different types of training to be performed by the UE. Moreover, the limitation of “based, at least in part on the one or more CSI autoencoder training capabilities” is already taught by Wang.] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far, including the teachings of Vitthaladevuni discussed above for the instant dependent claim, so as to also arrive at the further limitations of the instant dependent claim. The motivation for doing so is covered by the motivation given for Vitthaladevuni in the rejection of the parent independent claim. As to claim 9, this claim is directed to a system that performs operations that are the same or substantially the same as those of claim 1. Therefore, the rejection made to claim 1 is applied to claim 9. Additionally, Wang teaches the following limitations of claim 9 which differ from those of claim 1: “a system, comprising: one or more processors” [[0033]: “The base station 120 also includes processor(s) 260 and computer-readable storage media 262.” [0042]: “The core network server 302…includes processor(s) 304 and computer-readable storage media 306.”] “one or more memories to store at least a portion of the one or more neural networks. [[0033]: “The base station 120 also includes processor(s) 260 and computer-readable storage media 262.” [0037]: “The CRM 262 includes a training module 272 and a neural network table 274. In implementations, the base station 120 manages and deploys NN formation configurations to UE 110.” Additionally, [0080]: “Accordingly, at 640, the base station 120 receives updated ML information from at least some of the UEs, where the updated ML information can indicate any combination of ML parameters, ML architectures, and/or ML gradients.” That is, the model updates, which are a portion of the one or more neural networks, is also stored in the memory of the base station.] As to claim 10, the combination of Wang, Vitthaladevuni, and Timo teaches the system of claim 9, as set forth above. Vitthaladevuni further teaches “wherein the one or more resources to be used by the UE […] includes the one or more channel state information (CSI) resources” [As noted in the rejection of the parent independent claim, the “channel state reference signal identifier” and the “type of reference signal” are examples of “one or more CSI resources” to be used by the UE. This information is also part of “CSI information to be used by the UE,” since the term “CSI information” (channel state information information, i.e., information pertaining to channel state information, including the information sent to the UE for operations related to CSI information) generically includes the aforementioned resources.] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far, including the teachings of Vitthaladevuni discussed above for the instant dependent claim, so as to also arrive at the further limitations of the instant dependent claim. The motivation for doing so is covered by the motivation given for Vitthaladevuni in the rejection of the parent independent claim. Timo further teaches “the…more formats to report the generated information about the one or more 5G signals includes a CSI report format” [[0242]: “When a new UE joins a cell, the serving BS may configure the UE CSI reporting settings as part of the RRC (see Section 5.2 in 3GPP TS 38.214 V15.4.0).” [0254]: “Each Reporting Setting ReportConfig is associated with a single downlink BWP (higher layer parameter bandwidthPartId) and contains the reported parameter(s) for one CSI reporting band: CSI Type (I or II) if reported, codebook configuration including codebook subset restriction, time-domain behavior, frequency granularity for CQI and PMI, measurement restriction configurations, the strongest layer indicator (SLI), the reported L1-RSRP parameter(s), CRI, and SSBRI (SSB Resource Indicator). Each ReportConfig contains a ReportConfigID to identify the ReportConfig, a ReportConfigType to specify the time domain behavior of the report (either aperiodic, semi-persistent, or periodic), a ReportQuantity to indicate the CSI-related or L1-RSRP-related quantities to report, a ReportFreqConfiguration to indicate the reporting granularity in the frequency domain.”] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far, including the teachings of Timo noted above, so as to also arrive at the claimed invention of the instant dependent claim. The motivation for doing so is the same as the one given for Timo in the rejection of the parent independent claim. As to claim 11, the combination of Wang, Vitthaladevuni, and Timo teaches the system of claim 9, wherein the one or more processors are to cause the autoencoder to be trained based, at least in part, on an event trigger. [Wang, [0029]: “To illustrate, the base station 120 indicates, to the UE federated learning manager 220, to perform a training procedure and/or to transmit updated ML information in response to identifying a trigger event (e.g., changing ML parameters, changing ML architectures, changing signal or link quality parameters, changing UE-location).” Wang, [0036]: “The BS federated learning manager 270 indicates, to the UE 110, one or more update conditions (e.g., a trigger event, a periodicity) that specify when to perform a training procedure and/or when to report updated ML information to the BS federated learning manager 270.” The “autoencoder” is taught by Vitthaladevuni and, in the rejection of the parent claim, is implemented as the neural network being trained at the UE in the modification of Wang by the teachings of the other references. Therefore, the combination of the references and the associated rationale discussed in the rejection of the parent independent claim covers the instant limitation.] As to claim 12, the combination of Wang, Vitthaladevuni, and Timo teaches the system of claim 9, wherein the one or more processors are to cause the autoencoder to be trained by a user equipment (UE) device [As noted in the rejection of claim 1, the user equipment (UEs) and base station collectively train the model. See, e.g., Wang, [0078]: “At 630, at 631, and at 632, the UEs 111, 112, and 113 optionally perform a training procedure to generate the updated ML information…”; Wang, [0030]: “The UE training module 222 teaches and trains DNNs using known input data and/or by providing feedback to the ML algorithm.” The “autoencoder” is taught by Vitthaladevuni and, in the rejection of the parent claim, is implemented as the neural network being trained at the UE in the modification of Wang by the teachings of the other references. Therefore, the combination of the references and the associated rationale discussed in the rejection of the parent independent claim covers the instant limitation.] and a wireless radio network base station. [Wang, [0022]: “The base stations 120 are collectively a Radio Access Network 140 (e.g., RAN, Evolved Universal Terrestrial Radio Access Network, E-UTRAN, 5G NR RAN or NR RAN).” That is, the base stations and the UEs are part of a wireless radio network. Wang, [0080]: “Accordingly, at 640, the base station 120 receives updated ML information from at least some of the UEs, where the updated ML information can indicate any combination of ML parameters, ML architectures, and/or ML gradients.” Wang, [0082]: “In determining the common ML configuration, the base station 120 applies federated learning techniques that aggregate the updated ML information received from multiple UEs…the base station 120 performs averaging that aggregates ML parameters, gradients, and so forth.” Wang, [0084]: “At 655, the base station 120 directs the subset of UEs to update the DNN formed at 615 and at 616 using the common ML configuration determined at 650.”] As to claim 14, the combination of Wang, Vitthaladevuni, and Timo teaches the system of claim 9, wherein the one or more processors are to cause the autoencoder to be trained by the UE. [As noted in the rejection of claim 1, the user equipment (UEs) and base station collectively train the model. See, e.g., [0078]: “At 630, at 631, and at 632, the UEs 111, 112, and 113 optionally perform a training procedure to generate the updated ML information…”; [0030]: “The UE training module 222 teaches and trains DNNs using known input data and/or by providing feedback to the ML algorithm.” The “autoencoder” is taught by Vitthaladevuni and, in the rejection of the parent claim, is implemented as the neural network being trained at the UE in the modification of Wang by the teachings of the other references. Therefore, the combination of the references and the associated rationale discussed in the rejection of the parent independent claim covers the instant limitation.] As to claim 15, this claim is directed to a machine-readable medium for performing operations that are the same or substantially the same as those of claim 1. Therefore, the rejection made to claim 1 is applied to claim 15. Additionally, Wang teaches a machine-readable medium having stored thereon a set of instructions, which if performed by one or more processors, cause the one or more processors to at least: [[0104]: “Some operations of the example methods may be described in the general context of executable instructions stored on computer-readable storage memory that is local and/or remote to a computer processing system, and implementations can include software applications, programs, functions, and the like.” [0033]: “The base station 120 also includes processor(s) 260 and computer-readable storage media 262.” [0042]: “The core network server 302…includes processor(s) 304 and computer-readable storage media 306.” Any one or combination of the processors described above may correspond to the instant limitation.] As to claim 16, the combination of Wang, Vitthaladevuni, and Timo teaches the non-transitory machine-readable medium of claim 15, as set forth above. Timo further teaches “wherein the one or more formats to report the generated information about the one or more 5G signals includes a channel state information (CSI) codebook format.” [[0242]: “When a new UE joins a cell, the serving BS may configure the UE CSI reporting settings as part of the RRC (see Section 5.2 in 3GPP TS 38.214 V15.4.0).” [0254]: “Each Reporting Setting ReportConfig is associated with a single downlink BWP (higher layer parameter bandwidthPartId) and contains the reported parameter(s) for one CSI reporting band: CSI Type (I or II) if reported, codebook configuration including codebook subset restriction, time-domain behavior, frequency granularity for CQI and PMI, measurement restriction configurations, the strongest layer indicator (SLI), the reported L1-RSRP parameter(s), CRI, and SSBRI (SSB Resource Indicator).” [0259]: “If configured for Type-I or Type-II CSI feedback, the ReportConfig contains a CodebookConfig that specifies configuration parameters for Type-1 and Type-II CSI feedback.” See also [0057]-[0058]; [0172]; [0303] which contain further examples of codebook format.] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far, including the teachings of Timo noted above, so as to also arrive at the claimed invention of the instant dependent claim. The motivation for doing so is the same as the one given for Timo in the rejection of the parent independent claim. As to claim 17, the combination of Wang, Vitthaladevuni, and Timo teaches the non-transitory machine-readable medium of claim 15, wherein the set of instructions, which if performed by the one or more processors, cause the one or more processors to at least cause one or more of the UE, a wireless radio network base station, and a wireless radio network operations, administration, and maintenance (OAM) node to train [As noted in the rejection of claim 1, the user equipment (UEs) and base station collectively train the model. See, e.g., Wang, [0078]: “At 630, at 631, and at 632, the UEs 111, 112, and 113 optionally perform a training procedure to generate the updated ML information…”; Wang, [0030]: “The UE training module 222 teaches and trains DNNs using known input data and/or by providing feedback to the ML algorithm.” The Examiner notes that the phrase “one or more of” denotes an alternate expression, which is met with at least respect to the alternative of “UE.”] the autoencoder. [The “autoencoder” is taught by Vitthaladevuni and, in the rejection of the parent claim, is implemented as the neural network being trained at the UE in the modification of Wang by the teachings of the other references. Therefore, the combination of the references and the associated rationale discussed in the rejection of the parent independent claim covers the instant limitation.] As to claim 18, the combination of Wang, Vitthaladevuni, and Timo teaches the machine-readable medium of claim 15, wherein the set of instructions, which if performed by the one or more processors, cause the one or more processors to cause the CSA autoencoder training configuration to be sent to the UE [Wang, [0067]: “As illustrated, at 605, the base station 120 selects an initial ML configuration for a DNN that processes wireless network communications.” From FIG. 6, it is understood that the initial ML configuration is sent to the UEs. See also Wang, [0071]: “At 610, the base station 120 directs each UE in a set of UEs (e.g., UE 111, UE 112, UE 113) to form a (respective) DNN using the initial ML configuration. To illustrate, the base station 120 transmits an indication of an index value of a neural network table to the UEs 111, 112, and 113.” That is, that the content of the initial ML configuration is transmitted via indices of the neural network table.] to perform one or more of split training and federated training with one or more of a wireless radio network base station and a wireless radio network operations, administration, and maintenance (OAM) node. [Wang, [0029]: “The UE federated learning manager 220 identifies requests from the base station 120 that indicate one or more conditions that specify when to train a DNN and/or when to report the updated ML information … To illustrate, the base station 120 indicates, to the UE federated learning manager 220, to perform a training procedure and/or to transmit updated ML information.” Wang, [0082]: “In determining the common ML configuration, the base station 120 applies federated learning techniques that aggregate the updated ML information received from multiple UEs…the base station 120 performs averaging that aggregates ML parameters, gradients, and so forth.” The limitation of “wireless radio” is disclosed in e.g., Wang, [0021]: “The base stations 120 communicate with the user equipment 110 using the wireless links 130, which may be implemented as any suitable type of wireless link… The wireless links 130 may include one or more wireless links (e.g., radio links).” The phrase “one or more of” denotes an alternate expression, which is met here at least with respect to the alternative of “base station” such that the alternative of “OAM node” does not need to be met.]. As to claim 19, the combination of Wang, Vitthaladevuni, and Timo teaches the non-transitory machine-readable medium of claim 15, wherein the one or more capabilities of the UE to train one or more neural networks to generate information about one or more 5G signals include one or more of a computational capability of the UE to perform training and a memory storage of the UE to perform training. [Wang, [0068]: “…As one example, the UE capability information message includes ML capabilities (e.g., any one or more of: supported ML architectures, supported number of layers, available processing power, memory/storage capabilities, available power budget, fixed-point processing vs. floating-point processing, maximum kernel size capability, computation capability), and the base station 120 selects the initial ML configuration based on a common ML capability supported by the UEs 111, 112, and 113.”] As to claim 21, this claim is directed to a method comprising operations that are the same or substantially the same as those of claim 1. Therefore, the rejection made to claim 1 is applied to claim 21. As to claim 22, the combination of Wang, Vitthaladevuni, and Timo teaches the method of claim 21, as set forth above. Vitthaladevuni further teaches “wherein the CSI information includes one or more channel state information (CSI) resources to be used by the UE to perform CSI measurements to be used by the UE to train the autoencoder[As noted in the rejection of the parent independent claim, the “channel state reference signal identifier” and the “type of reference signal” are examples of “one or more CSI resources” to be used by the UE. This information is also part of “CSI information to be used by the UE,” since the term “CSI information” (channel state information information, i.e., information pertaining to channel state information, including the information sent to the UE for operations related to CSI information) generically includes the aforementioned resources. Furthermore, the CSI measurements are used by the UE to train the autoencoder. See, e.g., [0168]: “The UE may train the neural network to perform one or more of a wireless channel compression at the UE, a wireless channel measurement at the UE, a wireless interference measurement at the UE, UE positioning, wireless waveform determination at the UE, etc.”; [0106]: “an encoding device (e.g., a UE) may train one or more neural networks to learn dependence of measured qualities on individual parameters, isolate the measured qualities through various layers of the one or more neural networks (also referred to as “operations”), and compress measurements in a way that limits compression loss.”] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far, including the teachings of Vitthaladevuni discussed above for the instant dependent claim, so as to also arrive at the further limitations of the instant dependent claim. The motivation for doing so is covered by the motivation given for Vitthaladevuni in the rejection of the parent independent claim. As to claim 23, the combination of Wang, Vitthaladevuni, and Timo teaches the method of claim 21, wherein the one or more capabilities of the UE to train one or more neural networks to generate information about one or more 5G signals include one or more capabilities of the UE to train [Wang, [0068]: “…As one example, the UE capability information message includes ML capabilities (e.g., any one or more of: supported ML architectures, supported number of layers, available processing power, memory/storage capabilities, available power budget, fixed-point processing vs. floating-point processing, maximum kernel size capability, computation capability), and the base station 120 selects the initial ML configuration based on a common ML capability supported by the UEs 111, 112, and 113.”] “at least a portion of the autoencoder that includes the one or more neural networks” [The “autoencoder” is taught by Vitthaladevuni and, in the rejection of the parent claim, is implemented as the neural network being trained at the UE in the combined teachings of the references. Therefore, the combination of the references and the associated rationale discussed in the rejection of the parent independent claim covers the instant limitation.] As to claim 24, the combination of Wang, Vitthaladevuni, and Timo teaches the method of claim 21, further comprisingsending one or more training configurations to one or more devices based, at least in part, on the one or more capabilities of the UE to train one or more neural networks to generate information about one or more 5G signals. [Wang, [0067]: “As illustrated, at 605, the base station 120 selects an initial ML configuration for a DNN that processes wireless network communications.” From FIG. 6, it is understood that the initial ML configuration is sent to the UEs. See also Wang, [0071]: “At 610, the base station 120 directs each UE in a set of UEs (e.g., UE 111, UE 112, UE 113) to form a (respective) DNN using the initial ML configuration. To illustrate, the base station 120 transmits an indication of an index value of a neural network table to the UEs 111, 112, and 113.” That is, that the content of the initial ML configuration is transmitted via indices of the neural network table. Furthermore, as discussed in the rejection of claim 1, the ML configuration is based on the capabilities. See Wang, [0068]: “In aspects, the base station receives a UE capability information message (not illustrated) from each UE and selects the initial ML configuration based on a common UE capability between the UEs 111, 112, and 113.”] As to claim 26, the combination of Wang, Vitthaladevuni, and Timo teaches the method of claim 21, as set forth above, Vitthaladevuni further teaches “wherein the autoencoder is a first autoencoder, and the method further includes causing the UE to train the first autoencoder and another UE to train a second autoencoder” [As noted in the rejection of claim 1, Vitthaladevuni teaches training an autoencoder. The autoencoder or an instance of an autoencoder on one device can be regarded as a first autoencoder. Furthermore, Vitthaladevuni teaches a training command, as noted in the rejection of parent claim. See [0182]: “The network entity 1104 may instruct the UE 1102 to train the neural network to perform one or more of a wireless channel compression at the UE 1102, a wireless channel measurement at the UE 1102.” The limitation of “and another UE to train a second autoencoder” naturally flows from the combined teachings of the references in which the model of Vitthaladevuni is applied to the federated learning context of Wang, since in the combination of the references, each local device would be training a model that is an autoencoder.] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far, including the teachings of Vitthaladevuni discussed above for the instant dependent claim, so as to also arrive at the further limitations of the instant dependent claim. The motivation for doing so is covered by the motivation given for Vitthaladevuni in the rejection of the parent independent claim. As to claim 27, this claim is directed to a wireless radio network base station that performs operations that are the same or substantially the same as those of claim 1. Therefore, the rejection made to claim 1 is applied to claim 27. Furthermore, Wang teaches a wireless radio network base station, [[0080]: “Accordingly, at 640, the base station 120 receives updated ML information from at least some of the UEs, where the updated ML information can indicate any combination of ML parameters, ML architectures, and/or ML gradients.” The limitation of “wireless radio” is disclosed in e.g., [0021]: “The base stations 120 communicate with the user equipment 110 using the wireless links 130, which may be implemented as any suitable type of wireless link… The wireless links 130 may include one or more wireless links (e.g., radio links).”]. As to claim 28, the further limitations of this claims are substantially the same as those of claim 22. Therefore, the rejection made to claim 22 is applied to this claim. As to claim 29, the combination of Wang, Vitthaladevuni, and Timo teaches the wireless radio network base station of claim 27, wherein the circuitry is to cause the CSI autoencoder training configuration to be sent to the UE based, at least in part, on the one or more capabilities of the UE to update one or more parameters of one or more neural networks to generate information about one or more 5G signals. [Wang, [0067]: “As illustrated, at 605, the base station 120 selects an initial ML configuration for a DNN that processes wireless network communications.” From FIG. 6, it is understood that the initial ML configuration is sent to the UEs. See also Wang, [0071]: “At 610, the base station 120 directs each UE in a set of UEs (e.g., UE 111, UE 112, UE 113) to form a (respective) DNN using the initial ML configuration. To illustrate, the base station 120 transmits an indication of an index value of a neural network table to the UEs 111, 112, and 113.” That is, that the content of the initial ML configuration is transmitted via indices of the neural network table.] As to claim 31, the combination of Wang, Vitthaladevuni, and Timo teaches the wireless radio network base station of claim 27, wherein the circuitry is to aggregate locally trained autoencoder models from a plurality of UE devices that includes the UE. [Wang, [0036]: “The BS federated learning manager 270 also receives updated ML information from a set of UEs and aggregates the updated ML information to determine a common ML configuration usable by a subset of UEs to form DNNs that process wireless communications.” Wang, [0082]: “In determining the common ML configuration, the base station 120 applies federated learning techniques that aggregate the updated ML information received from multiple UEs (e.g., updated ML information transmitted at 635, 636, and/or 637) without potentially exposing private data used at the UE to generate the updated ML information.” The limitation of “autoencoder” is taught by Vitthaladevuni as applied in the existing combination of references.] As to claim 32, the combination of Wang, Vitthaladevuni, and Timo teaches the wireless radio network base station of claim 27, wherein the circuitry is to cause […] training configurations to be sent to the UE device based, at least in part, on the one or more capabilities of the UE to update one or more parameters of one or more neural networks to generate information about one or more 5G signals. [Wang, [0045]: “the base station 120 obtains the various criteria and/or link quality indications (e.g., any one or more of: RSSI, power information, SINR, RSRP, CQI, CSI, Doppler feedback, BLER, HARQ, timing measurements, error metrics, etc.) during the communications with the UE and forwards the criteria and/or link quality indications to the core network neural network manager 312.” Wang, [0068]: “The neural network table 318 stores multiple different NN formation configuration elements generated using the training module 316. In some implementations, the neural network table includes input characteristics for each NN formation configuration element and/or NN formation configuration, where the input characteristics describe properties about the training data used to generate the NN formation configuration. For instance, the input characteristics can include a power information, SINR information, CQI, CSI, Doppler feedback, RSS, error metrics, etc.” Note that the neural network table is a training configuration (see, e.g., Wang, [0039]: “the neural network table includes input characteristics for each NN formation configuration element and/or NN formation configuration, where the input characteristics describe properties about the training data used to generate the NN formation configuration element and/or NN formation configuration.”). Thus, the neural network table, which includes CSI input characteristics, constitutes a CSI training configuration. Regarding the limitation of “to be sent,” the configuration/neural network table is sent to the UEs as disclosed in [0071]: “At 610, the base station 120 directs each UE in a set of UEs (e.g., UE 111, UE 112, UE 113) to form a (respective) DNN using the initial ML configuration. To illustrate, the base station 120 transmits an indication of an index value of a neural network table to the UEs 111, 112, and 113.” That is, that the content of the initial ML configuration is transmitted vias indices of the neural network table.] Vitthaladevuni further teaches the limitation of “CSI autoencoder” the training configurations. While this limitation is not explicitly the same as the CSI autoencoder training configuration in the parent independent claim, this limitation is rendered obvious and covered by the combined teachings of the references for the same reasons given for that corresponding limitation in the rejection for the parent independent claim. 2. Claims 7 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Wang in view of Vitthaladevuni and Timo, and further in view of Kumar et al. (US 2022/0377844 A1) (“Kumar”). As to claim 7, the combination of Wang, Vitthaladevuni, and Timo teaches the processor of claim 1, wherein the circuitry is to cause […] of a wireless radio network to train [[0022]: “The base stations 120 are collectively a Radio Access Network 140 (e.g., RAN, Evolved Universal Terrestrial Radio Access Network, E-UTRAN, 5G NR RAN or NR RAN).” That is, the base stations and the UEs are part of a wireless radio network. Furthermore, the base stations and network perform the training process. See [0080]: “Accordingly, at 640, the base station 120 receives updated ML information from at least some of the UEs, where the updated ML information can indicate any combination of ML parameters, ML architectures, and/or ML gradients.” [0082]: “In determining the common ML configuration, the base station 120 applies federated learning techniques that aggregate the updated ML information received from multiple UEs…the base station 120 performs averaging that aggregates ML parameters, gradients, and so forth.” [0084]: “At 655, the base station 120 directs the subset of UEs to update the DNN formed at 615 and at 616 using the common ML configuration determined at 650.”] at least a portion of the autoencoder. [The “autoencoder” is taught by Vitthaladevuni and, in the rejection of the parent claim, is implemented as the neural network being trained at the UE in the combined teachings of the references. Therefore, the combination of the references and the associated rationale discussed in the rejection of the parent independent claim covers the instant limitation.] The combination of references thus far does not explicitly teach the element of “an operations, administration, and maintenance (OAM) node” as being caused to perform the training. Kumar, which generally pertains to machine learning model training (see title) in a federated learning context (see [0083]), teaches the above limitations of “an operations, administration, and maintenance (OAM) node” [[0086]: “The training may be initiated via a network entity 506 correspond to any of an OAM, a base station (e.g., including a RAN-based ML controller), a NWDAF, etc., that transmits, at 510, a model provisioning request to a model repository 508.” See also [0082] (“The training request may be based on protocols of an operations, administration, and maintenance (OAM) entity, a base station (e.g., including a RAN-based ML controller), and/or a network data analytics function (NWDAF)”); [0088] (“An ML function manager (e.g., OAM, RAN-based ML controller, etc.) may initiate the training procedure…).” [0087]: “The network entity 506 may or may not forward, at 530 b, the model training results to the model repository 508 for data aggregation, at either 532 a or 532 b. That is, the model may then be aggregated, at 532 a, via the network entity 506 and/or aggregated, at 532 b, via the model repository 508.” As shown in FIG. 5, the network entity 506 can be an OAM or base station.] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far with the teachings of Kumar by implementing “an operations, administration, and maintenance (OAM) node” as being caused to perform the training. Doing so would have been obvious as a combination of prior art elements according to known methods to yield predictable results, supported by the factors in MPEP § 2143(I)(A). Here, Wang as modified thus far teaches a processor which differed from the claimed invention by the lack of the OAM component. However, OAMs are known in Kumar are known in the art as being parts of a network for performing distributed learning and for being suitable for performing functions in distributed learning that can also be performed by a base station. Therefore, one of ordinary skill in the art could have combined the elements as claimed by known methods, and that in combination, each element merely performs the same function as it does separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable, namely the use of another component suitable for implementing federated learning. As to claim 13, the combination of Wang, Vitthaladevuni, and Timo teaches the system of claim 9, wherein the one or more processors are to cause the autoencoder to be trained by a user equipment (UE) device [As noted in the rejection of claim 1, the user equipment (UEs) and base station collectively train the model. See, e.g., Wang, [0078]: “At 630, at 631, and at 632, the UEs 111, 112, and 113 optionally perform a training procedure to generate the updated ML information…”; Wang, [0030]: “The UE training module 222 teaches and trains DNNs using known input data and/or by providing feedback to the ML algorithm.”] […] of a wireless radio network. [Wang, [0022]: “The base stations 120 are collectively a Radio Access Network 140 (e.g., RAN, Evolved Universal Terrestrial Radio Access Network, E-UTRAN, 5G NR RAN or NR RAN).” The “autoencoder” is taught by Vitthaladevuni and, in the rejection of the parent claim, is implemented as the neural network being trained at the UE in the combined teachings of the references. Therefore, the combination of the references and the associated rationale discussed in the rejection of the parent independent claim covers the instant limitation.], but does not teach the element of “an operations, administration, and maintenance (OAM) node.” Kumar, which generally pertains to machine learning model training (see title) in a federated learning context (see [0083]), teaches the above limitations of “an operations, administration, and maintenance (OAM) node” [[0086]: “The training may be initiated via a network entity 506 correspond to any of an OAM, a base station (e.g., including a RAN-based ML controller), a NWDAF, etc., that transmits, at 510, a model provisioning request to a model repository 508.” See also [0082] (“The training request may be based on protocols of an operations, administration, and maintenance (OAM) entity, a base station (e.g., including a RAN-based ML controller), and/or a network data analytics function (NWDAF)”); [0088] (“An ML function manager (e.g., OAM, RAN-based ML controller, etc.) may initiate the training procedure…).” [0087]: “The network entity 506 may or may not forward, at 530 b, the model training results to the model repository 508 for data aggregation, at either 532 a or 532 b. That is, the model may then be aggregated, at 532 a, via the network entity 506 and/or aggregated, at 532 b, via the model repository 508.” As shown in FIG. 5, the network entity 506 can be an OAM or base station.] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far with the teachings of Kumar by implementing “an operations, administration, and maintenance (OAM) node” as being caused to perform the training. Doing so would have been obvious as a combination of prior art elements according to known methods to yield predictable results, supported by the factors in MPEP § 2143(I)(A). Here, Wang as modified thus far teaches a processor which differed from the claimed invention by the lack of the OAM component. However, OAMs are known in Kumar are known in the art as being parts of a network for performing distributed learning and for being suitable for performing functions in distributed learning that can also be performed by a base station. Therefore, one of ordinary skill in the art could have combined the elements as claimed by known methods, and that in combination, each element merely performs the same function as it does separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable, namely the use of another component suitable for implementing federated learning. 3. Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over Wang in view of Vitthaladevuni and Timo, and further in view of Choi et al. (US 2022/0245527 A1) (“Choi”). As to claim 20, the combination of Wang, Vitthaladevuni, and Timo teaches the machine-readable medium of claim 15, but does not explicitly teach the further limitations of the instant dependent claim. Choi, which pertains to federated learning systems (see title), teaches “wherein the one or more capabilities are one or more of one or more types of inputs supported by a user equipment (UE) device to perform training and one or more quantization types supported by the UE device.” [[0077]: “In some examples, the UE 115 may transmit a capability message to the server (via the base station 105) indicating a set of quantization levels supported by the UE 115.” [0085]: “In some examples, each worker 215 may transmit a capability message 235 (e.g., capability messages 235-a, 235-b, and 235-c) to the server 205 that indicates the set of quantization levels supported by the worker 215. Here, the server 205, for each worker 215, may select a quantization level from the set of quantization levels supported by the worker 215.” Note that the instant wherein clause recites an alternate expression denoted by the phrase “one or more of….and…” Therefore, when the alternative of quantization types is met by the prior art, the entire alternate expression is satisfied and the other alternative of “types of inputs” do not need to be met.] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of Wang with the teachings of choi by implementing the one or more capabilities to be one or more quantization types supported by the UE device. The motivation would be to enable to the server or central controller to select a quantization level that is supported, in order to implement adaptive quantization level selection for data compression. See Choi, abstract (“To support adaptive quantization level selection in federated learning, a server may cause a base station to transmit an indication of a quantization level for a user equipment (UE) to use to compress gradient data output by a machine learning model.”) 3. Claims 25 and 30 are rejected under 35 U.S.C. 103 as being unpatentable over Wang in view of Vitthaladevuni and Timo, and further in view of Namgoong et al. (US 2023/0261909 A1) (“Namgoong”). As to claim 25, the combination of Wang, Vitthaladevuni, and Timo teaches the method of claim 21, as set forth above. Vitthaladevuni further teaches “further comprising causing the UE to train an encoder of the autoencoder” [As discussed in the rejection of the parent independent claim, Vitthaladevuni teaches the training of an autoencoder. The training of the encoder specifically is taught because the encoder is on the UE (see [0136]), and the network entity also directs the UE to train specific parts of the encoder, as described in [0168]-[0169]: “Referring again to FIG. 4A, the CSI sequence encoder 406 may capture time-dependency (for example, the Doppler component), and the CSI instance encoder 404 may capture Largely time-invariant aspects of the channels (for example, number of reflector, scatterers, or angle that the beam reaches the antenna of the UE). … For example, in case the UE is stationary in a relatively stable environment, then the change to the channel due to the Doppler effect may be mild. In such a case, the UE may not need to train every layer of the neural network. Accordingly, the UE may train the time dependency capturing layers.”] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far, including the teachings of Vitthaladevuni discussed above for the instant dependent claim, so as to also arrive at the further limitations of the instant dependent claim. The motivation for doing so is covered by the motivation given for Vitthaladevuni in the rejection of the parent independent claim. The combination of references thus far does not explicitly teach “and causing another device to train a decoder of the autoencoder.” Namgoong teaches “and causing another device to train a decoder of the autoencoder” [The limitation of “another device” is met by the fact that there are other UEs (clients). See, e.g., [0033]: “Federated learning is a machine learning technique that enables multiple clients to collaboratively learn neural network models”; [0062]: “At base station 110, the uplink signals from UE 120 and other UEs may be received by antennas 234.” In general, UEs, including another UE, also takes part in training their local autoencoders, which include a decoder portion as shown in FIG. 3. See [0102]: “B [note that “B” this is a misprint of “θ”] represents the parameters for the decoders”; [0103]: “pθ(x|h, z) is parameterized by the decoder of the autoencoder.” As noted above, the client determines the updated set of parameters (see [0104]: “In some aspects, it may be desired to find the neural network parameters θ and ϕ that maximizes the ELBO…”), which includes the parameters θ for the decoder.] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far with the teachings of Namgoong by implementing training of autoencoders in other devices in a federated learning system, so as to arrive at the claimed invention. The motivation would have been to use train autoencoders collaboratively using multiple UEs, to facilitate better physical layer link performance. Namgoong, [0033] (“During the training, the autoencoders and the classifier are collaboratively learned using the federated learning techniques. In this way, aspects may facilitate better physical layer link performance.”). As to claim 30, the combination of Wang, Vitthaladevuni, and Timo teaches the wireless radio network base station of claim 27, as set forth above. Vitthaladevuni further teaches “wherein the circuitry is to cause the UE to update parameters of an encoder” [As discussed in the rejection of the parent independent claim, Vitthaladevuni teaches the training of an autoencoder. The training of the encoder specifically is taught because the encoder is on the UE (see [0136]), and the network entity also directs the UE to train specific parts of the encoder, as described in [0168]-[0169]: “Referring again to FIG. 4A, the CSI sequence encoder 406 may capture time-dependency (for example, the Doppler component), and the CSI instance encoder 404 may capture Largely time-invariant aspects of the channels (for example, number of reflector, scatterers, or angle that the beam reaches the antenna of the UE). … For example, in case the UE is stationary in a relatively stable environment, then the change to the channel due to the Doppler effect may be mild. In such a case, the UE may not need to train every layer of the neural network. Accordingly, the UE may train the time dependency capturing layers.”] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far, including the teachings of Vitthaladevuni discussed above for the instant dependent claim, so as to also arrive at the further limitations of the instant dependent claim. The motivation for doing so is covered by the motivation given for Vitthaladevuni in the rejection of the parent independent claim. The combination of references thus far does not teach updating “a decoder of the autoencoder.” Namgoong teaches updating “a decoder of the autoencoder.” [As shown in FIG. 3, each autoencoder include an “encoder.” See [0080]: “As shown, the client autoencoder 310 may include an encoder 314 configured to receive an observed wireless communication vector, x, as input and to provide a latent vector, h, as output.” The encoder parameters are also being trained in the training process. See [0102] (“qϕ(h|x,z) is parameterized by the encoder of the autoencoder… KL (qϕ(h|x, z)∥pθ(h|z)) is the regularization term for the autoencoder”), where the loss function in this paragraph is used in accordance with the loss function in [0104] that is “used as a loss function in the training” in order to optimize the neural network. See [0104] (“to find the neural network parameters θ and ϕ that maximizes the ELBO…” in regards to “decoder” specifically in general, the local autoencoders also include a decoder portion as shown in FIG. 3. See [0102]: “B [note that “B” this is a misprint of “θ”] represents the parameters for the decoders”; [0103]: “pθ(x|h, z) is parameterized by the decoder of the autoencoder.” As noted above, the client determines the updated set of parameters (see [0104]: “In some aspects, it may be desired to find the neural network parameters θ and ϕ that maximizes the ELBO…”), which includes the parameters θ for the decoder.] It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far with the teachings of Namgoong by implementing training of autoencoders in other devices in a federated learning system, so as to arrive at the claimed invention. The motivation would have been to use train autoencoders collaboratively using multiple UEs, to facilitate better physical layer link performance. Namgoong, [0033] (“During the training, the autoencoders and the classifier are collaboratively learned using the federated learning techniques. In this way, aspects may facilitate better physical layer link performance.”). Conclusion The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. The following references depict the state of the art. Jeon et al. (US 20250008449 A1) teaches training an autoencoder in a federated learning system for telecom use. Guo et al., “User-Centric Online Gossip Training for Autoencoder-Based CSI Feedback,” in IEEE Journal of Selected Topics in Signal Processing, vol. 16, no. 3, pp. 559-572, April 2022 teaches training a CSI autoencoder with a distributed learning system Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a). A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action. Any inquiry concerning this communication or earlier communications from the examiner should be directed to YAO DAVID HUANG whose telephone number is (571)270-1764. The examiner can normally be reached Monday - Friday 9:00 am - 5:30 pm. Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Miranda Huang can be reached at (571) 270-7092. 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. /Y.D.H./Examiner, Art Unit 2124 /MIRANDA M HUANG/Supervisory Patent Examiner, Art Unit 2124