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 .
This Office Action is in response to communications on 6/17/2026.
Claims 1, 2, 4, 5, 9-17 & 21-23 & 26-29 are pending and presented for examination.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 6/15/2026 was filed after the mailing date of the Non-final Rejection on 3/19/2026. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Response to Amendment
Claim 7 has been cancelled. Claims 3, 6, 8, 18-20, 24 & 25 were previously cancelled.
Claims 1 & 22 have been amended.
Claim 29 has been added and is presented for examination.
Rejections to claims 1, 2, 4, 5, 9-17 & 21-23 & 26-28 under 35 USC 103 made in the prior record Non-final Rejection dated 3/19/2026 have been withdrawn based on amendments to claims 1 & 22, but new grounds of rejections under 35 USC 103 have been made based on new references Yerramalli et al. (US 2019/0387418)(herein after “Yerramalli”) and Ladkat et al. (US 11853391)(herein after “Ladkat”) and previously presented reference Zilka et al. (US 11741191)(herein after “Zilka”).
Response to Arguments
Applicant's arguments filed 6/17/2026 have been fully considered but they are not persuasive.
Applicant argues that Narayanan fails to disclose the features “the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device” because Narayanan teaches of a WTRU that determines that training is complete based on its own criteria, not training that continues until the network device deactivates the training process, and Narayanan does not describe any iterative training that continues until such deactivation. Examiner respectfully disagrees, noting that [0126] of Narayanan teaches of deactivating a specific configuration set containing learning parameters associated with online training through MACE CE/DCI signaling. To someone having ordinary skill in the art, DCI is downlink control information transmitted from a network device to a user device such as a WTRU. Thus, Narayanan does not teach of a WTRU completing training based on its own criteria since a network device must send the DCI signaling to deactivate a specific configuration set for online training. Further, [0119] of Narayanan teaches of iterations of online training where learnable parameters are affected, and thus the online training deactivated through DCI signaling may be iterative online training.
Based on the above discussion, examiner maintains that Narayana teaches of “the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device”.
Applicant’s arguments, see “Remarks”, filed 6/17/2026, with respect to the rejections of claims 1, 2, 4, 5, 9-17 & 21-23 & 26-28 under 35 USC 103 have been fully considered and are persuasive. Therefore, these rejections have been withdrawn. However, upon further consideration, new grounds of rejections are made in view of new references Yerramalli et al. (US 2019/0387418)(herein after “Yerramalli”) and Ladkat et al. (US 11853391)(herein after “Ladkat”) and previously presented reference Zilka et al. (US 11741191)(herein after “Zilka”).
Regarding claim 1, applicant submits that amendments to this claim traverse the rejection of this claim under 35 USC 103 made in the Non-final Rejection dated 3/19/2026. Examiner agrees and withdraws rejection of claim 1 under 35 USC 103 made in the Non-final Rejection dated 3/19/2026. However, after further consideration, examiner introduces a new ground of rejection of claim 1 under 35 USC 103 based on new references Yerramalli and Ladkat and previously presented reference Zilka. Applicant’s arguments with respect to claim 1 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Regarding claim 22, applicant submits that this claim traverses the rejection of this claim under 35 USC 103 made in the Non-final Rejection dated 3/19/2026 due to similar amendments and arguments as made for claim 1. Examiner agrees and withdraws rejection of claim 22 under 35 USC 103 made in the Non-final Rejection dated 3/19/2026. However, for the same reasons as discussed above, examiner introduces a new ground of rejection of claim 22 under 35 USC 103 based on new references Yerramalli and Ladkat and previously presented reference Zilka.
Regarding claims 2, 4, 5, 9-17 & 21, 23 & 26-28, applicant submits that these claims traverse the rejections of these claims under 35 USC 103 made in the Non-final Rejection dated 3/19/2026 due to amendments and arguments made for claims 1 & 22 and due to their dependency on claims 1 or 22. Examiner agrees and withdraws rejections of claims 2, 4, 5, 9-17 & 21, 23 & 26-28 under 35 USC 103 made in the Non-final Rejection dated 3/19/2026. However, for the same reasons as discussed above, examiner introduces new grounds of rejections of claims 2, 4, 5, 9-17 & 21, 23 & 26-28 under 35 USC 103 based on new references Yerramalli and Ladkat and previously presented reference Zilka.
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 pre-AIA 35 U.S.C. 103(a) which forms the basis for all obviousness rejections set forth in this Office action:
(a) A patent may not be obtained though the invention is not identically disclosed or described as set forth in section 102, if the differences between the subject matter sought to be patented and the prior art are such that the subject matter as a whole would have been obvious at the time the invention was made to a person having ordinary skill in the art to which said subject matter 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 pre-AIA 35 U.S.C. 103(a) are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims under pre-AIA 35 U.S.C. 103(a), the examiner presumes that the subject matter of the various claims was commonly owned at the time any inventions covered therein were made absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and invention dates of each claim that was not commonly owned at the time a later invention was made in order for the examiner to consider the applicability of pre-AIA 35 U.S.C. 103(c) and potential pre-AIA 35 U.S.C. 102(e), (f) or (g) prior art under pre-AIA 35 U.S.C. 103(a).
Claims 1, 2, 4, 5, 9, 21-23 & 26-29 rejected under pre-AIA 35 U.S.C. 103(a) as being unpatentable over Shen et al. (WO 2021/035724)(herein after “Shen”) in view of Narayanan et al. (US 20230409963)(herein after “Narayanan”) and further in view of Anderson et al. (US 2022/0020142)(herein after “Anderson”) and Yerramalli et al. (US 2019/0387418)(herein after “Yerramalli”) and Zilka et al. (US 11741191)(herein after “Zilka”) and Ladkat et al. (US 11853391)(herein after “Ladkat”).
Regarding Claim 1, Shen discloses a communication method comprising: receiving, by a terminal device, first downlink control information (DCI) from a network device, wherein the first DCI is for a first training process, and wherein the first training process is for training a model corresponding to target information (Fig 2 & [0038] disclose wireless communication between a base station 110 (i.e. a network device) and a terminal 120 where the terminal receives a DCI from the base station. [0045] discloses a method wherein the DCI received by the terminal indicates a resource to be used for transmitting a training set. [0046] discloses the training set is used to train a machine learning model. [0051] discloses a target bit field included in the DCI to determine whether the resource indicated in the DCI is used to transmit the training set. [0063] discloses that when it is determined that the resource is used for transmitting a training set, the target bit field may also indicate use of the training set to train the machine learning model (i.e. an indication to activate a training process). [0047]-[0048] discloses the training set includes at least one known target configuration including scheduling parameters (i.e. target information).).
Shen fails to disclose wherein the first DCI is for activating a first training process; in response to receiving the first DCI, activating the first training process and iteratively performing model training on a first model in the first training process, the iteratively performing model training on the first model in the first training process comprises iteratively sending, to the network device using a semi persistent resource of a physical uplink shared channel, first parameter information of the first model obtained from performing iterative model training on the first model, and the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device; receiving, by the terminal device, second DCI from the network device, wherein the second DCI is for deactivating the first training process; and in response to receiving the second DCI, deactivating, by the terminal device, the first training process.
However, Narayanan teaches wherein the first DCI is for activating a first training process ([0126] discloses a DCI may be used to activate a specific configuration set, where a configuration set consists of learning parameters for online training (i.e. activate a specific training process).);
in response to receiving the first DCI, activating, by the terminal device, the first training process and iteratively performing model training on a first model in the first training process ([0126] discloses a WTRU receiving a DCI that may be used to activate a specific configuration set, where a configuration set consists of learning parameters for online training (i.e. activate a specific training process for a first model). A broadest reasonable interpretation is that a configuration set consisting of learning parameters for online training may be interpreted as a training process. Thus, activating a specific configuration set may be interpreted as activating a specific training process. The DCI that activates the specific configuration set may also trigger the WTRU to perform online training. [0119] & [0135] disclose that training may occur over a number of epochs, where each epoch is a full training pass over an entire training dataset, and during each iteration of training learnable parameters may be affected.),
the iteratively performing model training on the first model in the first training process comprises iteratively sending, to the network device using a semi persistent resource, first parameter information of the first model obtained from performing iterative model training on the first model ([0106]-[0107] discloses that a WTRU may be configured to report reconstruction loss (i.e. first parameter information) to the network semi-persistently. [0012] discloses that the reconstruction loss is part of an iterative training process where a first node (i.e. a WTRU) updates learnable parameters to reduce the reconstruction loss value. A broadest reasonable interpretation is that the WTRU iteratively and semi-persistently sends the reconstruction loss value each iteration of the training process.), and the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device ([0126] discloses that a MAC CE/DCI may be signaled (i.e. by a network device) to the WTRU to deactivate a training configuration set.);
receiving, by the terminal device, second DCI from the network device, wherein the second DCI is for deactivating the first training process ([0126] discloses a DCI may be signaled (i.e. received) by a WTRU to deactivate a specific configuration set, where a configuration set consists of learning parameters for online training (i.e. deactivate a specific training process for a first model). A broadest reasonable interpretation is that a configuration set consisting of learning parameters for online training may be interpreted as a training process. Thus, deactivating a specific configuration set may be interpreted as deactivating a specific training process.); and
in response to receiving the second DCI, deactivating, by the terminal device, the first training process ([0126] discloses a DCI may be signaled to deactivate, by the WTRU, a specific configuration set, where a configuration set consists of learning parameters for online training (i.e. deactivate a specific training process for a first model). A broadest reasonable interpretation is that a configuration set consisting of learning parameters for online training may be interpreted as a training process. Thus, a WTRU deactivating a specific configuration set in response to receiving a DCI indicating to deactivate a specific configuration set may be interpreted as the WTRU deactivating a specific training process in response to receiving a DCI indicating to deactivate a specific configuration set.).
Therefore, it would have been obvious to someone having ordinary skill in the art prior to the effective filing date of the claimed invention to have a communication method comprising: receiving, by a terminal device, first downlink control information (DCI) from a network device, wherein the first DCI is for a first training process, and wherein the first training process is for training a model corresponding to target information, as disclosed by Shen, wherein the first DCI is for activating a first training process; in response to receiving the first DCI, activating, by the terminal device, the first training process and iteratively performing model training on a first model in the first training process, the iteratively performing model training on the first model in the first training process comprises iteratively sending, to the network device using a semi persistent resource, first parameter information of the first model obtained from performing iterative model training on the first model, and the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device ; receiving, by the terminal device, second DCI from the network device, wherein the second DCI is for deactivating the first training process; and in response to receiving the second DCI, deactivating, by the terminal device, the first training process, as taught by Narayanan. The motivation to do so would be to have a communication method where a WTRU can receive: a first DCI providing information for a plurality of training sets (or training processes) for a training model that also indicates a specific training set is activated and training should be performed, in response to receiving the first DCI the WTRU iteratively performs training using the specific training set at each iteration sends a reconstruction loss value to a network using semi-persistent PUCCH resources so that the network can determine when the WTRU has accurately reconstructed the training model, so that the network can then send a second DCI indicating a specific training set is deactivated and in response the WTRU can deactivate the specific training set process, in order to improve performance of the WTRU while minimizing processing time in the WTRU through activating, using well defined standards based protocols (i.e. DCI), specific training sets for the training model only while the WTRU is iteratively performing training to reduce reconstruction loss to a point where the training model is accurate.
Shen fails to disclose but Yerramalli further teaches wherein the semi persistent resource is of a physical uplink shared channel ([0013] discloses that transmission parameters may be sent through semi-persistent scheduling (SPS) via a PUSCH (physical uplink shared channel).).
Therefore, it would have been obvious to someone having ordinary skill in the art prior to the effective filing date of the claimed invention to have a communication method comprising: iteratively sending, to a network device using a semi persistent resource, first parameter information of a first model obtained from performing iterative model training on the first model, as disclosed by Shen in view of Narayanan, wherein the semi persistent resource is of a physical uplink shared channel, as further taught by Yerramalli. The motivation to do so would be to have a communication method where a WTRU can send reconstruction loss values to a network, for iteratively training models of a training process, using semi-persistent PUSCH resources in order to allow multiple WTRUs to iteratively send reconstruction loss values to the network semi-persistently using different PUSCH shared resources to improve accuracy of the training models for each WTRU.
Shen fails to disclose wherein the network device performs model training on a second model in the first training process, and wherein the model corresponding to the target information comprises the first model and the second model;
However, Anderson further teaches wherein the network device performs model training on a second model in the first training process, and wherein the model corresponding to the target information comprises the first model and the second model ([0012] discloses applying a first trained model to input data to obtain a first output data and applying a second trained model to the input data to obtain a second output, wherein the first and second trained models are dependent on a hierarchical relationship between the first and second outputs. Both the first and second training models are part of the same training process with outputs with hierarchical dependency on their outputs corresponding the same clinical data target information.);
Therefore, it would have been obvious to someone having ordinary skill in the art prior to the effective filing date of the claimed invention to have a communication method comprising: receiving first downlink control information (DCI) from a network device, wherein the first DCI is for a first training process, and wherein the first training process is for training a model corresponding to target information, as disclosed by Shen, wherein the network device performs model training on a second model in the first training process, and wherein the model corresponding to the target information comprises the first model and the second model, as further taught by Anderson. The motivation to do so would be to have a communication method where a UE can receive a DCI providing information that activates a specific training set (or training process) of a plurality of training sets and initiates the UE to perform training using two training models with hierarchically dependency on their outputs in order to improve accuracy in training compared to using a single training model.
Shen fails to disclose but Zilka further teaches wherein the first parameter information comprises information of the first model, and wherein the network device adjusts a parameter of the second model based on the first parameter information and performs a further round of model training on the second model in response to the first parameter information (Fig 2 & col 8, lines 58-65 disclose a first machine learning model 208 of a local training system 202 in a user device. Fig 1 & col 5, lines 18-34 disclose a second machine learning model 108 of a global training system 102 (i.e. a network device). Fig 6 & col 13, lines 32-65 disclose that the global training system 102 receives parameter update data (i.e. first parameter information) of machine learning model 208 from one or more of the user devices and updates the current parameters of machine learning model 108 using the parameter update data from the user device. Fig 6 & col 14, lines 4-16 disclose that global training system 102 can determine to perform another training iteration on machine learning model 108 in response to a training termination criteria based on the parameter update received from the user device.).
Therefore, it would have been obvious to someone having ordinary skill in the art prior to the effective filing date of the claimed invention to have a communication method comprising: activating, by a terminal device, a first training process and iteratively performing model training on a first model in the first training process, wherein a network device performs model training on a second model in the first training process, the iteratively performing model training on the first model in the first training process comprises iteratively sending, to the network device using a semi persistent resource of a physical uplink shared channel, first parameter information of the first model obtained from performing iterative model training on the first model, and the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device, and wherein the model corresponding to the target information comprises the first model and the second model, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson, wherein the first parameter information comprises gradient information of the first model, and wherein the network device adjusts a parameter of the second model based on the first parameter information and performs a further round of model training on the second model in response to the first parameter information, as further taught by Zilka. The motivation to do so would be to have a communication method where a WTRU can iteratively perform training of a first machine learning (ML) model of a user device and a second ML model of a global network device, wherein parameter update data generated from the first ML model, based on a first iteration of training, is used to perform updates to the second ML model, so that the second ML model can determine to perform a second iteration of training of the second ML model based on the parameter update data from the first ML model, in order to optimize the first ML model and the second ML model to improve prediction accuracy of the first ML model and the second ML model.
Shen fails to disclose but Ladkat further teaches wherein the first parameter information comprises gradient information (Fig 6 & col 9, lines 11-31 disclose that updates to the weights of a machine learning model may be based on gradients (i.e. gradient information).).
Therefore, it would have been obvious to someone having ordinary skill in the art prior to the effective filing date of the claimed invention to have a communication method comprising: activating, by a terminal device, a first training process and iteratively performing model training on a first model in the first training process, wherein a network device performs model training on a second model in the first training process, the iteratively performing model training on the first model in the first training process comprises iteratively sending, to the network device using a semi persistent resource of a physical uplink shared channel, first parameter information of the first model obtained from performing iterative model training on the first model, and the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device, and wherein the model corresponding to the target information comprises the first model and the second model, wherein the first parameter information comprises gradient information of the first model, and wherein the network device adjusts a parameter of the second model based on the first parameter information and performs a further round of model training on the second model in response to the first parameter information, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka, wherein the first parameter information comprises gradient information, as further taught by Ladkat. The motivation to do so would be to have a communication method where a WTRU can iteratively perform training of a first machine learning (ML) model of a user device and a second ML model of a global network device, wherein parameter update data generated from the first ML model, based on a first iteration of training, is based on gradient information used to perform updates to the second ML model, so that the second ML model can determine to perform a second iteration of training of the second ML model based on the parameter update data from the first ML model, in order to optimize the first ML model and the second ML model to improve prediction accuracy of the first ML model and the second ML model based on direction and rate of change information provided by the gradient based parameter update data.
Regarding Claim 2. Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat discloses the method according to Claim 1.
Shen fails to disclose wherein the first training process further comprises: receiving first data from the network device; and wherein iteratively performing model training on the first model comprises: iteratively training the first model based on the first data in order to obtain the first parameter information of the first model.
However, Zilka teaches wherein the first training process further comprises: receiving first data from the network device (Col 2, lines 55-67 & col 3, lines 1-3 disclose receiving a first data from a communication network.); wherein iteratively performing model training on the first model comprises:
Iteratively training the first model based on the first data in order to obtain the first parameter information of the first model (Col 2, lines 55-67 & col 3, lines 1-6 and col 4, lines 58-67 & col 5, lines 1-5 disclose iteratively training a first machine learning model based on the first data received and obtaining an first parameter update for the first machine learning model.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of claim 1, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat, wherein the first training process further comprises: receiving first data from the network device; and wherein iteratively performing model training on the first model comprises: iteratively training the first model based on the first data in order to obtain the first parameter information of the first model, as further taught by Zilka. The motivation to do so would be to improve a machine learning network model corresponding to search results from search queries through iterative parameter feedback and model training based on a first training using first data in the machine learning model.
Regarding Claims 4 & 26, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat disclose the method according to claim 1 and the apparatus according to claim 22.
Shen fails to disclose wherein the first DCI is further for activating or indicating the semi-persistent resource.
However, Narayanan teaches wherein the first DCI is further for activating or indicating the semi-persistent resource ([0107] discloses that a PUCCH resource may be used to semi-persistently report reconstruction loss, and that a DCI may be used to request for (i.e. activate) the aperiodic reporting of reconstruction loss.).).
Therefore, it would have been obvious to someone having ordinary skill in the art prior to the effective filing date of the claimed invention to have the method according to claim 1 or the apparatus according to claim 22, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat, wherein the first DCI is further for activating or indicating the semi-persistent resource, as taught by Narayanan. The motivation to do so would be to have a WTRU, or a communication method for a WTRU, that can receive a DCI configuring semi-persistent PUCCH resources for the WTRU to iteratively report reconstruction loss for a training model to a network so that the network can determine when the WTRU has accurately reconstructed the training model so that the network can configure the WTRU to deactivate the training model to save processing power at the WTRU.
Regarding Claims 5 & 27, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat disclose the method according to claim 1 and the apparatus according to claim 22.
Shen discloses wherein the first DCI comprises a first indicator field ([0051] & [0055] discloses a first DCI comprising a first target field (i.e. first indicator field).).
Shen fails to disclose wherein the first indicator field indicates that the first DCI is for activating the first training process.
However, Narayanan teaches wherein the first indicator field indicates that the first DCI is for activating the first training process ([0126] discloses a DCI for activating a specific configuration for online training (i.e. a first training process).).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of Claim 1, or the apparatus of claim 22, wherein the first DCI comprises a first indicator field, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat, wherein the first indicator field indicates that the first DCI is for activating the first training process, as taught by Narayanan. The motivation to do so would be to have a WTRU, or a communication method for a WTRU, that can receive: a first DCI providing information for a plurality of training sets (or training processes) for a training model that also indicates a specific training set is activated, in order to simplify the process of activating and deactivating specific training sets for a training model at the UE by using well defined standards based protocols (i.e. DCI).
Regarding Claim 9, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat disclose the method according to claim 1.
Shen discloses wherein the second DCI comprises a second indicator field (Table 5 & [0085] disclose multiple DCI formats for a DCI. A first DCI format 1_2 is an indicator field scheduling a PDSCH carrying a downlink training set and a second DCI format 0_2 is an indicator field scheduling a PUSCH carrying an uplink training set.).
Shen fails to disclose wherein the second indicator field indicates that the second DCI is for deactivating the training process.
However, Narayanan teaches wherein the second indicator field indicates that the second DCI is for deactivating the first training process ([0126] discloses a first DCI to activate online training and a second DCI to deactivate online training.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of claim 1, wherein the second DCI comprises a second indicator field, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat, wherein the second indicator field indicates that the second DCI is for deactivating the first training process, as taught by Narayanan. The motivation to do so would be to provide a method for disabling training at a receiving module in response to an indication from a receiving module of reconstruction loss or unavailable status of an AI component.
Regarding claim 21, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat discloses the method according to Claim 1 wherein both the first DCI and the second DCI are associated with the first training process.
Shen discloses wherein both the first DCI and the second DCI are associated with a first radio network temporary identifier (RNTI) that is associated with the first training process (Table 5 & [0085] disclose multiple DCI formats. A first DCI format 1_2 identifier indicates the scheduling of a PDSCH carrying a downlink training set and a second DCI format 0_2 identifier indicates the scheduling of a PUSCH carrying an uplink training set. [0090] discloses that the resources indicated by the first DCI format 1_2 and the second DCI format 0_2 may be used to transmit a training set according to a target RNTI adopted by the first DCI format 1_2 and the second DCI format 0_2. Thus, disclosed are a first DCI identifier and a second DCI identifier associated with the an RNTI associated with a training process.).
Regarding claim 22, Shen discloses an apparatus (Fig 7 & [0195] disclose a terminal), comprising:
at least one processor (Fig 7 & [0195] disclose the terminal includes a processor 91); and
non-transitory computer readable storage storing executable instructions that are executed by at least one processor (Fig 7 & [0201] disclose a computer-readable storage medium storing at least one instruction that can be executed by the processor.), wherein execution of the instructions causes the apparatus to:
receive first downlink control information (DCI) from a network device, wherein the first DCI is for a first training process, and wherein the first training process is for training a model corresponding to target information (Fig 2 & [0038] disclose wireless communication between a base station 110 (i.e. a network device) and a terminal 120 where the terminal receives a DCI from the base station. [0045] discloses a method wherein the DCI received by the terminal indicates a resource to be used for transmitting a training set. [0046] discloses the training set is used to train a machine learning model. [0051] discloses a target bit field included in the DCI to determine whether the resource indicated in the DCI is used to transmit the training set. [0063] discloses that when it is determined that the resource is used for transmitting a training set, the target bit field may also indicate use of the training set to train the machine learning model (i.e. an indication to activate a training process). [0047]-[0048] discloses the training set includes at least one known target configuration including scheduling parameters (i.e. target information).);
Shen fails to disclose wherein the first DCI is for activating a first training process; in response to receiving the first DCI, activate the first training process and iteratively perform model training on a first model in the first training process, the iteratively performing model training on the first model in the first training process comprises iteratively sending, to the network device using a semi persistent resource, first parameter information of the first model obtained from iteratively performing model training on the first model in the first training process, and the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device; receive second DCI from the network device, wherein the second DCI is for deactivating the first training process; and in response to receiving the second DCI, deactivate the first training process.
However, Narayanan teaches wherein the first DCI is for activating a first training process ([0126] discloses a DCI may be used to activate a specific configuration set, where a configuration set consists of learning parameters for online training (i.e. activate a specific training process). A broadest reasonable interpretation is that a configuration set consisting of learning parameters for online training may be interpreted as a training process. Thus, activating a specific configuration set may be interpreted as activating a specific training process.);
in response to receiving the first DCI, activate the first training process and perform model training on a first model in the first training process ([0126] discloses a DCI may be used to activate a specific configuration set, where a configuration set consists of learning parameters for online training (i.e. activate a specific training process for a first model). A broadest reasonable interpretation is that a configuration set consisting of learning parameters for online training may be interpreted as a training process. Thus, activating a specific configuration set may be interpreted as activating a specific training process. The DCI that activates the specific configuration may also trigger a WTRU to perform online training.);
the iteratively performing model training on the first model in the first training process comprises iteratively sending, to the network device using a semi persistent resource, first parameter information of the first model obtained from iteratively performing model training on the first model in the first training process ([0106]-[0107] discloses that a WTRU may be configured to report reconstruction loss (i.e. first parameter information) to the network semi-persistently using PUCCH resources. [0012] discloses that the reconstruction loss is part of an iterative training process where a first node (i.e. a WTRU) updates learnable parameters to reduce the reconstruction loss value. A broadest reasonable interpretation is that the WTRU iteratively and semi-persistently sends the reconstruction loss value each iteration of the training process.), and the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device ([0126] discloses that a MAC CE/DCI may be signaled (i.e. by a network device) to the WTRU to deactivate a training configuration set.);
receive second DCI from the network device, wherein the second DCI is for deactivating the first training process ([0126] discloses a DCI may be used to deactivate a specific configuration set, where a configuration set consists of learning parameters for online training (i.e. deactivate a specific training process for a first model). A broadest reasonable interpretation is that a configuration set consisting of learning parameters for online training may be interpreted as a training process. Thus, deactivating a specific configuration set may be interpreted as deactivating a specific training process.); and
in response to receiving the second DCI, deactivate the first training process ([0126] discloses a DCI may be used to deactivate a specific configuration set, where a configuration set consists of learning parameters for online training (i.e. deactivate a specific training process for a first model). A broadest reasonable interpretation is that a configuration set consisting of learning parameters for online training may be interpreted as a training process. Thus, deactivating a specific configuration set may be interpreted as deactivating a specific training process.).
Therefore, it would have been obvious to someone having ordinary skill in the art prior to the effective filing date of the claimed invention to have an apparatus comprising: at least one processor; and non-transitory computer readable storage storing executable instructions that are executed by at least one processor, wherein execution of the instructions causes the apparatus to: receive first downlink control information (DCI) from a network device, wherein the first DCI is for a first training process, and wherein the first training process is for training a model corresponding to target information, as disclosed by Shen, wherein the first DCI is for activating a first training process; in response to receiving the first DCI, activate the first training process and perform model training on a first model in the first training process, the iteratively performing model training on the first model in the first training process comprises iteratively sending, to the network device using a semi persistent resource, first parameter information of the first model obtained from iteratively performing model training on the first model in the first training process, and the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device; receive second DCI from the network device, wherein the second DCI is for deactivating the first training process; and in response to receiving the second DCI, deactivate the first training process, as taught by Narayanan. The motivation to do so would be to have a WTRU, comprising a processor and memory with instructions that can be executed by the processor, that can receive: a first DCI providing information for a plurality of training sets (or training processes) for a training model that also indicates a specific training set is activated and training should be performed, in response to receiving the first DCI the WTRU iteratively performs training using the specific training set at each iteration sends a reconstruction loss value to a network using semi-persistent PUCCH resources so that the network can determine when the WTRU has accurately reconstructed the training model, so that the network can then send a second DCI indicating a specific training set is deactivated and in response the WTRU can deactivate the specific training set process, in order to improve performance of the WTRU while minimizing processing time in the WTRU through activating, using well defined standards based protocols (i.e. DCI), specific training sets for the training model only while the WTRU is iteratively performing training to reduce reconstruction loss to a point where the training model is accurate.
Shen fails to disclose but Yerramalli further teaches wherein the semi persistent resource is of a physical uplink shared channel ([0013] discloses that transmission parameters may be sent through semi-persistent scheduling (SPS) via a PUSCH (physical uplink shared channel).).
Therefore, it would have been obvious to someone having ordinary skill in the art prior to the effective filing date of the claimed invention to have an apparatus comprising: at least one processor; and non-transitory computer readable storage storing executable instructions that are executed by at least one processor, wherein execution of the instructions causes the apparatus to: iteratively send, to a network device using a semi persistent resource, first parameter information of a first model obtained from perform iterative model training on the first model, as disclosed by Shen in view of Narayanan, wherein the semi persistent resource is of a physical uplink shared channel, as further taught by Yerramalli. The motivation to do so would be to have a WTRU that can send reconstruction loss values to a network, for iteratively training models of a training process, using semi-persistent PUSCH resources in order to allow multiple WTRUs to iteratively send reconstruction loss values to the network semi-persistently using different PUSCH shared resources to improve accuracy of the training models for each WTRU.
Shen fails to disclose wherein the network device performs model training on a second model in the first training process, and wherein the model corresponding to the target information comprises the first model and the second model;
However, Anderson further teaches wherein the network device performs model training on a second model in the first training process, and wherein the model corresponding to the target information comprises the first model and the second model ([0012] discloses applying a first trained model to input data to obtain a first output data and applying a second trained model to the input data to obtain a second output, wherein the first and second trained models are dependent on a hierarchical relationship between the first and second outputs. Both the first and second training models are part of the same training process with outputs with hierarchical dependency on their outputs corresponding the same clinical data target information.);
Therefore, it would have been obvious to someone having ordinary skill in the art prior to the effective filing date of the claimed invention to have an apparatus that: receives first downlink control information (DCI) from a network device, wherein the first DCI is for a first training process, and wherein the first training process is for training a model corresponding to target information, as disclosed by Shen, wherein the network device performs model training on a second model in the first training process, and wherein the model corresponding to the target information comprises the first model and the second model, as further taught by Anderson. The motivation to do so would be to have a UE that can receive a DCI providing information that activates a specific training set (or training process) of a plurality of training sets and initiates the UE to perform training using two training models with hierarchically dependency on their outputs in order to improve accuracy in training compared to using a single training model.
Shen fails to disclose but Zilka further teaches wherein the first parameter information comprises information of the first model, and wherein the network device adjusts a parameter of the second model based on the first parameter information and performs a further round of model training on the second model in response to the first parameter information (Fig 2 & col 8, lines 58-65 disclose a first machine learning model 208 of a local training system 202 in a user device. Fig 1 & col 5, lines 18-34 disclose a second machine learning model 108 of a global training system 102 (i.e. a network device). Fig 6 & col 13, lines 32-65 disclose that the global training system 102 receives parameter update data (i.e. first parameter information) of machine learning model 208 from one or more of the user devices and updates the current parameters of machine learning model 108 using the parameter update data from the user device. Fig 6 & col 14, lines 4-16 disclose that global training system 102 can determine to perform another training iteration on machine learning model 108 in response to a training termination criteria based on the parameter update received from the user device.).
Therefore, it would have been obvious to someone having ordinary skill in the art prior to the effective filing date of the claimed invention to have an apparatus comprising at least one processor; and non-transitory computer readable storage storing executable instructions that are executed by at least one processor, wherein execution of the instructions causes the apparatus to: activate, by a terminal device, a first training process and iteratively perform model training on a first model in the first training process, wherein a network device performs model training on a second model in the first training process, the iteratively performing model training on the first model in the first training process comprises iteratively sending, to the network device using a semi persistent resource of a physical uplink shared channel, first parameter information of the first model obtained from performing iterative model training on the first model, and the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device, and wherein the model corresponding to the target information comprises the first model and the second model, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson, wherein the first parameter information comprises information of the first model, and wherein the network device adjusts a parameter of the second model based on the first parameter information and performs a further round of model training on the second model in response to the first parameter information, as further taught by Zilka. The motivation to do so would have been to have a WTRU that can iteratively perform training of a first machine learning (ML) model of a user device and a second ML model of a global network device, wherein parameter update data generated from the first ML model, based on a first iteration of training, is used to perform updates to the second ML model, so that the second ML model can determine to perform a second iteration of training of the second ML model based on the parameter update data from the first ML model, in order to optimize the first ML model and the second ML model to improve prediction accuracy of the first ML model and the second ML model.
Shen fails to disclose but Ladkat further teaches wherein the first parameter information comprises gradient information (Fig 6 & col 9, lines 11-31 disclose that updates to the weights of a machine learning model may be based on gradients (i.e. gradient information).).
Therefore, it would have been obvious to someone having ordinary skill in the art prior to the effective filing date of the claimed invention to have an apparatus comprising at least one processor; and non-transitory computer readable storage storing executable instructions that are executed by at least one processor, wherein execution of the instructions causes the apparatus to: activate, by a terminal device, a first training process and iteratively perform model training on a first model in the first training process, wherein a network device performs model training on a second model in the first training process, the iteratively performing model training on the first model in the first training process comprises iteratively sending, to the network device using a semi persistent resource of a physical uplink shared channel, first parameter information of the first model obtained from performing iterative model training on the first model, and the iteratively performing model training on the first model in the first training process continues until the first training process is deactivated by the network device, and wherein the model corresponding to the target information comprises the first model and the second model, wherein the first parameter information comprises information of the first model, and wherein the network device adjusts a parameter of the second model based on the first parameter information and performs a further round of model training on the second model in response to the first parameter information, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka, wherein the first parameter information comprises gradient information, as further taught by Ladkat. The motivation to do so would have been to have a communication method where a WTRU can iteratively perform training of a first machine learning (ML) model of a user device and a second ML model of a global network device, wherein parameter update data generated from the first ML model, based on a first iteration of training, is based on gradient information used to perform updates to the second ML model, so that the second ML model can determine to perform a second iteration of training of the second ML model based on the parameter update data from the first ML model, in order to optimize the first ML model and the second ML model to improve prediction accuracy of the first ML model and the second ML model based on direction and rate of change information provided by the gradient based parameter update data.
Regarding claim 23, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat disclose the apparatus according to claim 22.
Shen fails to disclose wherein the first training process further comprises: receiving first data from the network device; wherein performing model training on the first model comprises: training the first model based on the first data in order to obtain first parameter information of the first model; and wherein the first training process further comprises: sending the first parameter information to the network device.
However, Zilka further teaches wherein the first training process further comprises: receiving first data from the network device (Col 2, lines 55-67 & col 3, lines 1-3 disclose receiving a first data from a communication network.); wherein performing model training on the first model comprises:
training the first model based on the first data in order to obtain first parameter information of the first model (Col 2, lines 55-67 & col 3, lines 1-3 disclose training a first machine learning model based on the first data received and obtaining an first parameter update for the first machine learning model.); and
wherein the first training process further comprises: sending the first parameter information to the network device. (Col 2, lines 55-67 & col 3, lines 1-3 disclose transmitting the first parameter update to the communication network.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the apparatus of claim 22, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat, wherein the first training process further comprises: receiving first data from the network device; wherein performing model training on the first model comprises: training the first model based on the first data in order to obtain first parameter information of the first model; and wherein the first training process further comprises: sending the first parameter information to the network device, as further taught by Zilka. The motivation to do so would be to have a UE assist in improving a machine learning network model corresponding to search results from search queries through parameter feedback based on a first training using first data in the machine learning model.
Regarding Claim 28, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat disclose the apparatus according to claim 22.
Shen discloses wherein both the first DCI and the second DCI indicate an identifier of the first training process (Table 5 & [0085] disclose multiple DCI formats. A first DCI format 1_2 identifier indicates the scheduling of a PDSCH carrying a downlink training set and a second DCI format 0_2 identifier indicates the scheduling of a PUSCH carrying an uplink training set.).
Regarding claim 29, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat disclose the method according to claim 1.
Shen fails to disclose but Zilka further teaches wherein the iteratively performing model training on the first model in the first training process further comprises: sending, to the network device, second data obtained by processing training data using the first model (Fig 2 & col 8, lines 58-65 disclose a first machine learning (ML) model 208 of a local training system 202 in a user device. Fig 1 & col 5, lines 18-34 disclose a second ML model 108 of a global training system 102 (i.e. a network device). Fig 6 & col 13, lines 32-65 disclose that the global training system 102 receives parameter update data (i.e. second data) based on ML model 208 training using data defining current parameter values from global training system 102.);
receiving, from the network device, second parameter information, wherein the second parameter information is information of the second model trained by the network device based on the second data (Fig 6, col 13, lines 40-67 & col 14, lines 1-16 disclose that the user device receives from the global training system data defining the current parameter value of machine learning model 108 at each iteration of training. Thus, after a first iteration of training of the first ML model of the user device, based on first update data received from the global training system, and after the user device sends the global training system parameter update data (i.e. based on second data), and after the global training system updates the parameters of ML learning model 108 (i.e. second model trained by the network device), the global training system sends second data defining the current parameter values of ML model 108 (i.e. second parameter information) to the user device.); and
adjusting a parameter of the first model based on the second parameter information (Fig 6, col 13, lines 40-62 discloses that the user device receives the current parameter values (i.e. second parameter information) from the global training system and updates ML model 208 based on the received current parameter values.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method according to claim 1, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat, wherein the iteratively performing model training on the first model in the first training process further comprises: sending, to the network device, second data obtained by processing training data using the first model; receiving, from the network device, second parameter information, wherein the second parameter information is information of the second model trained by the network device based on the second data; and adjusting a parameter of the first model based on the second parameter information, as further taught by Zilka. The motivation to do so would have been to have a WTRU that can iteratively perform training of a first machine learning (ML) model of a user device and a second ML model of a global network device, wherein parameter update data generated from the first ML model, based on a first iteration of training, is used to perform updates to the second ML model, so that the global network device can perform training of the second ML model to determine new parameter update values to send to the user device to adjust the first ML model for performing a second iteration of training of the first ML model based on the new parameter update values from the global network device, in order to optimize the first ML model and the second ML model to improve prediction accuracy of the first ML model and the second ML model.
Shen fails to disclose but Ladkat further teaches wherein the second parameter information comprises gradient information (Fig 6 & col 9, lines 11-31 disclose that updates to the weights of a machine learning model may be based on gradients (i.e. gradient information).).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method according to claim 1, wherein the iteratively performing model training on the first model in the first training process further comprises: sending, to the network device, second data obtained by processing training data using the first model; receiving, from the network device, second parameter information, wherein the second parameter information is information of the second model trained by the network device based on the second data; and adjusting a parameter of the first model based on the second parameter information, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat, wherein the second parameter information comprises gradient information as further taught by Ladkat. The motivation to do so would have been to have a WTRU that can iteratively perform training of a first machine learning (ML) model of a user device and a second ML model of a global network device, wherein parameter update data generated from the first ML model, based on a first iteration of training, is used to perform updates to the second ML model, so that the global network device can perform training of the second ML model to determine new gradient based parameter update values to send to the user device to adjust the first ML model for performing a second iteration of training of the first ML model based on the new parameter update values from the global network device, in order to optimize the first ML model and the second ML model to improve prediction accuracy of the first ML model and the second ML model based on direction and rate of change information provided by the gradient based the new gradient based parameter update values.
Claims 10-17 rejected under pre-AIA 35 U.S.C. 103(a) as being unpatentable over Shen et al. (WO 2021/035724) in view of Narayanan et al. (US 20230409963)(herein after “Narayanan”) and Anderson et al. (US 2022/0020142)(herein after “Anderson”) and Yerramalli et al. (US 2019/0387418)(herein after “Yerramalli”) and Zilka et al. (US 11741191)(herein after “Zilka”) and Ladkat et al. (US 11853391)(herein after “Ladkat”), as applied to claim 1, and further in view of Bai et al. (WO 2022000365)(herein after “Bai”).
Regarding Claim 10, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat discloses the method according to Claim 1.
Shen discloses receiving a second DCI from a base station (Fig 2 & [0038] disclose wireless communication between a base station 110 (i.e. a network device) and a terminal 120 where the terminal receives a DCI from the base station. Table 5 & [0085] further disclose multiple DCI formats for the DCI received from the base station. A first DCI format 1_2 is received when scheduling a PDSCH carrying a downlink training set and a second DCI format 0_2 is received when scheduling a PUSCH carrying an uplink training set. Thus, further disclosed is the receiving of a second DCI from a base station.).
Shen fails to disclose further comprising receiving third DCI from the network device.
However, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have a third DCI received from a network device, since it has been held that mere duplication of parts that does not produce new or unexpected results involves only routine skill in the art and thus has no patentable significance (see MPEP Section 2144.04, subsection VI.B). The motivation to do so would be to define a plurality of DCI formats for sending multiple DCIs to indicate various uses of resources, activation of processes or other control messages from a base station.
Shen fails to disclose wherein the third DCI is for activating a prediction process, and the prediction process comprises a process of predicting the target information by using the model corresponding to the target information.
However, Bai further teaches wherein the third DCI is for activating a prediction process, and the prediction process comprises a process of predicting the target information by using the model corresponding to the target information ([00104]-[00105] discloses delivering and activating a machine learning model structure through a DCI. [0059] discloses that a SOC in a UE may receive the machine learning model from a base station through a DCI, including code to predict a future downlink data channel (i.e. target information) based on using the machine learning model corresponding to estimating and predicting the downlink data channel.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method according to claim 1 further comprising receiving third DCI from the network device, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat, wherein the third DCI is for activating a prediction process, and the prediction process comprises a process of predicting the target information by using the model corresponding to the target information, as further taught by Bai. The motivation to do so would be to provide a method for a base station to indicate to a UE, through a DCI, to activate future downlink data prediction based on a machine learning model if the base station is detecting poor downlink data channel performance at the UE without use of the machine learning model to predict future downlink data channel state.
Regarding Claim 11, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai disclose the method according to Claim 10.
Shen discloses a second DCI comprising a second indicator field (Table 5 & [0085] disclose multiple DCI formats for a DCI. A first DCI format 1_2 is an indicator field scheduling a PDSCH carrying a downlink training set and a second DCI format 0_2 is an indicator field scheduling a PUSCH carrying an uplink training set. Thus, disclosed is a second DCI comprising a second indicator field.).
Shen fails to disclose wherein the third DCI comprises a third indicator field.
However, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have a third indicator field, since it has been held that mere duplication of parts that does not produce new or unexpected results involves only routine skill in the art and thus has no patentable significance (see MPEP Section 2144.04, subsection VI.B). The motivation to do so would be to define a plurality of DCI indicator fields for sending multiple DCIs to indicate various uses of resources, activation of processes or other control messages from a base station.
Shen fails to disclose wherein the third indicator field indicates that the third DCI is for activating the prediction process.
However, Bai further teaches wherein the third indicator field indicates that the third DCI is for activating the prediction process ([00104]-[00105] discloses delivering and activating a machine learning model structure through a DCI (that would include an indicator field). [0059] discloses that a SOC in a UE may receive the machine learning model from a base station through a DCI (that would include an indicator field), including code to predict a future downlink data channel (i.e. target information) based on using the machine learning model corresponding to estimating and predicting the downlink data channel.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method according to claim 10 wherein the third DCI comprises a third indicator field, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai, and wherein the third indicator field indicates that the third DCI is for activating the prediction process, as further taught by Bai. The motivation to do so would be to provide a method for a base station to indicate to a UE, through an indicator field in a DCI, to activate future downlink data prediction based on a machine learning model if the base station is detecting poor downlink data channel performance at the UE without use of the machine learning model to predict future downlink data channel state.
Regarding Claim 12, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai disclose the method according to Claim 10.
Shen discloses receiving a second DCI from a base station (Fig 2 & [0038] disclose wireless communication between a base station 110 (i.e. a network device) and a terminal 120 where the terminal receives a DCI from the base station. Table 5 & [0085] further disclose multiple DCI formats for the DCI received from the base station. A first DCI format 1_2 is received when scheduling a PDSCH carrying a downlink training set and a second DCI format 0_2 is received when scheduling a PUSCH carrying an uplink training set. Thus, further disclosed is the receiving of a second DCI from a base station.).
Shen fails to disclose further comprising receiving fourth DCI from the network device.
However, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have a fourth DCI received from a network device, since it has been held that mere duplication of parts that does not produce new or unexpected results involves only routine skill in the art and thus has no patentable significance (see MPEP Section 2144.04, subsection VI.B). The motivation to do so would be to define a plurality of DCI formats for sending multiple DCIs to indicate various uses of resources, activation of processes or other control messages from a base station.
Shen fails to disclose wherein the fourth DCI is for deactivating a process.
However, Narayanan teaches wherein the fourth DCI is for deactivating a process ([0126] discloses a first DCI to activate online training and a second DCI to deactivate online training.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of claim 10 further comprising receiving fourth DCI from the network device, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai, wherein the fourth DCI is for deactivating a process, as taught by Narayanan. The motivation to do so would be to provide a method for base station to disable a process such as a training process at a terminal in response to an indication from the terminal of reconstruction loss or unavailable status of an AI component.
Shen fails to disclose wherein the process is a prediction process.
However, Bai further teaches wherein the process is a prediction process ([0059] discloses that a SOC in a UE may receive a machine learning model including code to predict a future downlink data channel (i.e. target information) based on using the machine learning model corresponding to estimating and predicting the downlink data channel).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of claim 10 wherein the fourth DCI is for deactivating a process, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai, and wherein the process is a prediction process, as further taught by Bai. The motivation to do so would be to provide a method for a base station to indicate to a UE, through an indicator field in a DCI, to deactivate future downlink data prediction based on a machine learning model if the base station is detecting superior downlink data channel performance at the UE to save battery power at the UE.
Regarding Claim 13 Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai disclose the method according to Claim 12.
Shen discloses a first DCI identifier and a second DCI identifier associated with an RNTI (Table 5 & [0085] disclose multiple DCI formats. A first DCI format 1_2 identifier indicates the scheduling of a PDSCH carrying a downlink training set and a second DCI format 0_2 identifier indicates the scheduling of a PUSCH carrying an uplink training set. [0090] discloses that the resources indicated by the first DCI format 1_2 and the second DCI format 0_2 may be used to transmit a training set according to a target RNTI adopted by the first DCI format 1_2 and the second DCI format 0_2. Thus, disclosed are a first DCI identifier and a second DCI identifier associated with an RNTI.).
Shen fails to disclose wherein both the third DCI and the fourth DCI indicate an identifier of the process, and/or both the third DCI and the fourth DCI are associated with a second radio network temporary identifier (RNTI).
However, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have a third DCI and a fourth DCI indicate an identifier of a process, and/or both the third DCI and the fourth DCI are associated with a second RNTI, since it has been held that mere duplication of parts that does not produce new or unexpected results involves only routine skill in the art and thus has no patentable significance (see MPEP Section 2144.04, subsection VI.B). The motivation to do so would be to define a plurality of DCI formats for sending multiple DCIs to indicate various uses of resources, activation of processes or other control messages from a base station.
Shen fails to disclose wherein the process is a prediction process.
However, Bai further teaches a wherein the process is a prediction process ([0059] discloses that a SOC in a UE may receive a machine learning model including code to predict a future downlink data channel (i.e. target information) based on using the machine learning model corresponding to estimating and predicting the downlink data channel).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of claim 12, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai, wherein both the third DCI and the fourth DCI indicate an identifier of a process, and/or both the third DCI and the fourth DCI are associated with a second RNTI, and wherein the process is a prediction process, as further taught by Bai. The motivation to do so would be to define multiple RNTIs for scrambling and masking CRC bits of a multiple DCI messages, allowing multiple individuals or multiple groups of devices to distinguish the multiple DCIs intended to provide multiple different training set information for multiple different machine learning processing.
Regarding Claim 14, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai disclose the method according to claim 13.
Shen discloses wherein the second RNTI is one of the following RNTIs: an artificial intelligence RNTI, a prediction process RNTI, a cell RNTI, a prediction RNTI, or a semi-persistent scheduling RNTI ([0090]-[0092] disclose an RNTI may be a new training process RNTI, for example ML-RNTI, for scheduling resources used to transmit a training set, or can be a C-RNTI.).
Regarding Claim 15, Shen in view Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai disclose the method according to Claim 12.
Shen discloses a second DCI comprising a second indicator field (Table 5 & [0085] disclose multiple DCI formats for a DCI. A first DCI format 1_2 is an indicator field scheduling a PDSCH carrying a downlink training set and a second DCI format 0_2 is an indicator field scheduling a PUSCH carrying an uplink training set. Thus, disclosed is a second DCI comprising a second indicator field.).
Shen fails to disclose wherein the fourth DCI comprises a fourth indicator field.
However, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have a fourth indicator field, since it has been held that mere duplication of parts that does not produce new or unexpected results involves only routine skill in the art and thus has no patentable significance (see MPEP Section 2144.04, subsection VI.B). The motivation to do so would be to define a plurality of DCI indicator fields for sending multiple DCIs to indicate various uses of resources, activation of processes or other control messages from a base station.
Shen fails to disclose wherein the fourth indicator field indicates that the fourth DCI is for deactivating a process.
However, Narayanan teaches wherein the fourth indicator field indicates that the fourth DCI is for deactivating a process ([0126] discloses a DCI to activate online training and a separate DCI to deactivate online training.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of claim 12, as disclosed by Shen in view Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai, wherein the fourth indicator field indicates that the fourth DCI is for deactivating a process, as taught by Narayanan. The motivation to do so would be to provide a method for disabling training at a receiving module in response to an indication from a receiving module of reconstruction loss or unavailable status of an AI component.
Shen fails to disclose wherein the process is a prediction process.
However, Bai further teaches wherein the process is a prediction process ([0059] discloses that a SOC in a UE may receive a machine learning model including code to predict a future downlink data channel (i.e. target information) based on using the machine learning model corresponding to estimating and predicting the downlink data channel).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of claim 12, as disclosed by Shen in view Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai, wherein the process is a prediction process, as further taught by Bai. The motivation to do so would be to provide a method for a base station to indicate to a UE, through an indicator field in a DCI, to deactivate future downlink data prediction based on a machine learning model if the base station is detecting superior downlink data channel performance at the UE to save battery power at the UE.
Regarding Claim 16, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat disclose the method according to Claim 1.
Shen fails to disclose wherein the second DCI is for deactivating the first training process.
However, Narayanan teaches wherein the second DCI is for deactivating the first training process ([0126] discloses a first DCI to activate online training and a second DCI to deactivate online training.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of claim 1, as disclosed by Shen in view Narayanan and Yerramalli and Anderson and Zilka and Ladkat, wherein the second DCI is for deactivating the first training process, as taught by Narayanan. The motivation to do so would be to provide a method for base station to disable training at a terminal in response to an indication from the terminal of reconstruction loss or unavailable status of an AI component.
Shen fails to disclose wherein the second DCI is for activating a prediction process, and wherein the prediction process comprises a process of predicting the target information by using the model corresponding to the target information
However, Bai further teaches wherein the second DCI is for activating a prediction process, and the prediction process comprises a process of predicting the target information by using the model corresponding to the target information ([00104]-[00105] discloses delivering and activating a machine learning model structure through a DCI. [0059] discloses that a SOC in a UE may receive the machine learning model from a base station through a DCI, including code to predict a future downlink data channel (i.e. target information) based on using the machine learning model corresponding to estimating and predicting the downlink data channel.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of claim 1, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat, wherein the second DCI is for activating a prediction process, and the prediction process comprises a process of predicting the target information by using the model corresponding to the target information, as further taught by Bai. The motivation to do so would reduce signaling overhead by have a single DCI indicate both disabling of training and activation of a prediction process at a terminal in response to an indication from the terminal of reconstruction loss or unavailable status of an AI component.
Regarding Claim 17, Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai disclose the method according to Claim 16.
Shen discloses a second DCI comprising a second indicator field (Table 5 & [0085] disclose multiple DCI formats for a DCI. A first DCI format 1_2 is an indicator field scheduling a PDSCH carrying a downlink training set and a second DCI format 0_2 is an indicator field scheduling a PUSCH carrying an uplink training set. Thus, disclosed is a second DCI comprising a second indicator field.).
Shen fails to disclose wherein the second DCI comprises a fifth indicator field.
However, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have a fifth indicator field, since it has been held that mere duplication of parts that does not produce new or unexpected results involves only routine skill in the art and thus has no patentable significance (see MPEP Section 2144.04, subsection VI.B). The motivation to do so would be to define a plurality of DCI indicator fields for sending multiple DCIs to indicate various uses of resources, activation of processes or other control messages from a base station.
Shen fails to disclose wherein the fifth indicator field indicates that the fifth DCI is for deactivating the first training process.
However, Narayanan teaches wherein the fifth indicator field indicates that the fifth DCI is for deactivating the first training process and activating a process ([0126] discloses a DCI to activate online training and a separate DCI to deactivate online training.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of claim 16, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai, wherein the fifth indicator field indicates that the fifth DCI is for deactivating the first training process and activating a process, as taught by Narayanan. The motivation to do so would be to provide a method for disabling training at a receiving module in response to an indication from a receiving module of reconstruction loss or unavailable status of an AI component.
Shen fails to disclose wherein the process is a prediction process and wherein the fifth indicator field indicates that the second DCI is for activating the prediction process.
However, Bai further teaches wherein the process is a prediction process and wherein the fifth indicator field indicates that the second DCI is for activating the prediction process ([0059] discloses that a SOC in a UE may receive a machine learning model including code to predict a future downlink data channel (i.e. target information) based on using the machine learning model corresponding to estimating and predicting the downlink data channel. [00104]-[00105] discloses delivering and activating a machine learning model structure through a DCI.).
Therefore, it would have been obvious to someone of ordinary skill in the art prior to the effective filing date of the claimed invention to have the method of Claim 16, as disclosed by Shen in view of Narayanan and Yerramalli and Anderson and Zilka and Ladkat and Bai, wherein the process is a prediction process and wherein the fifth indicator field indicates that the second DCI is for activating the prediction process, as further taught by Bai. The motivation to do so would be to provide a method for a base station to indicate to a UE, through an indicator field in a DCI, to deactivate future downlink data prediction based on a machine learning model if the base station is detecting superior downlink data channel performance at the UE to save battery power at the UE and activate a prediction process at a terminal in response to an indication from the terminal of reconstruction loss or unavailable status of an AI component, in order to reduce signaling overhead by having a single indicator field indicate a DCI is for both disabling a training process and enabling a prediction process.
Conclusion
The following prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Liu et al. (US 2017/0024849) discloses Learning Convolution Neural Networks on Heterogeneous CPU-GPU Platform.
Kumar et al. (US 11574637) discloses Spoken Language Understanding Models.
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 JAMES P SEYMOUR whose telephone number is (571)272-7654. The examiner can normally be reached M-F 8-5 EST.
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, Nishant Divecha can be reached at 571-270-3125. 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.
/JAMES P SEYMOUR/Examiner, Art Unit 2419
/Nishant Divecha/Supervisory Patent Examiner, Art Unit 2419
Prosecution Insights