DETAILED ACTION
This nonfinal office action is responsive to claims filed on March 7, 2024. Claims 1-24 are pending. Claims 1 and 13 are independent.
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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on February 9, 2022 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim(s) 1-5 and 13-18 is/are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Pei (US20210042889), hereinafter Pei.
Regarding claim 1, Pei teaches the communication method:
determining a first submodel and a second submodel, (Pei, paragraph 0046: “In the disclosure, two sub-models, namely, a first sub-model and a second sub-model, are generated according to an AI processing model. The first sub-model is deployed in a mobile device. The second sub-model is deployed in a server. A computing task is processed by using the AI processing model in a manner of using the first sub-model in combination with the second sub-model, so as to reduce a computing workload and an internal memory overhead of a model of the AI processing model deployed in the mobile device, so that the AI processing model may be deployed in the mobile device.”) wherein the first submodel and the second submodel are used in a matching manner; and (Pei, paragraph 0046: “A computing task is processed by using the AI processing model in a manner of using the first sub-model in combination with the second sub-model, so as to reduce a computing workload and an internal memory overhead of a model of the AI processing model deployed in the mobile device, so that the AI processing model may be deployed in the mobile device.” – The first sub-model being used in combination with the second sub-model is analogous to the first submodel and second submodel being used in a matching manner.)
sending first information, wherein the first information indicates one or more of input data of the first submodel or output data of the first submodel. (Pei, paragraph 0024: “After obtaining data to be processed, the mobile device 102 performs first processing on the data by using the first sub-model to obtain an intermediate processing result, and transmits the intermediate processing result to the server 106 by using the network 104.” – The intermediate results is analogous to the first information as it is the output of the first submodel.)
Regarding claim 2, Pei teaches the communication method of claim 1, as cited above.
Pei further teaches:
an output of the first submodel is used to determine an input of the second submodel; or an output of the second submodel is used to determine an input of the first submodel. (Pei, paragraph 0063: “When the neural network layers included in the second computing segment are trained, the neural network layers included in the second computing segment may be trained based on a model training algorithm by using the output of the first sub-model as the input of the second computing segment and using a training data processing result of the AI processing model (on which model compression is not performed), to obtain a network recognition layer (the second sub-model) not sacrificing a final accuracy.”)
Regarding claim 3, Pei teaches the communication method of claim 1, as cited above.
Pei further teaches:
the first submodel is used to send information at a transmit end, and the second submodel is used to receive the information at a receive end; or the second submodel is used to send information at the transmit end, and the first submodel is used to receive the information at the receive end. (Pei, paragraphs 0030-0031: “Operation S204: Transmit the intermediate processing result to a first server, the first server being configured to perform second processing on the intermediate processing result by using the second sub-model, to obtain a target processing result, target processing including the first processing and the second processing. Operation S206: Receive the target processing result returned by the first server.” – The first sub-model is deployed on a mobile device, see paragraph 0023, and is therefore sending information at a transmit end and the second sub-model is deployed in the server and is therefore used to receive information at a receive end.)
Regarding claim 4, Pei teaches the communication method of claim 1, as cited above.
Pei further teaches:
wherein the first submodel and the second submodel belong to a bilateral model. (Pei, paragraph 0046: “A computing task is processed by using the AI processing model in a manner of using the first sub-model in combination with the second sub-model, so as to reduce a computing workload and an internal memory overhead of a model of the AI processing model deployed in the mobile device, so that the AI processing model may be deployed in the mobile device.” – The first sub-model used in combination with the second sub-model, wherein the output of the first sub-model is used as input into the second sub-model is analogous to the first submodel and second submodel belonging to a bilateral model.)
Regarding claim 5, Pei teaches the communication method of claim 1, as cited above.
Pei further teaches:
the first information is used to train a third submodel, and a function of the third submodel is identical to a function of the first submodel; and/or an input type of the third submodel is identical to an input type of the first submodel, and an output type of the third submodel is identical to an output type of the first submodel; and/or a dimension of input data of the third submodel is identical to as a dimension of the input data of the first submodel, and a dimension of output data of the third submodel is identical to a dimension of the output data of the first submodel; and/or an input of the third submodel is identical to the input of the first submodel, and a difference between an output of the third submodel and the output of the first submodel is less than a first threshold. (Pei, paragraph 0104-0105: “When the AI processing model is trained, for an uncompressed AI processing model, an original model may be trained, to obtain a high-accuracy AI processing model. For a partially compressed AI processing model, based on the foregoing training, the first n (for example, 3) neural network layers of the AI processing model may be trained and compressed by using a first algorithm (for example, a distillation method) for compressing a model and using obtained intermediate data of the high-accuracy AI processing model as a label, to obtain a mobile computing part. Other neural network layers other than the first n (for example, 3) layers of the AI processing model may be trained by using a second algorithm (for example, transfer learning), using the intermediate data and/or final data of the high-accuracy AI processing model as a label, and using an output result of the mobile computing part as an input, to obtain a cloud computing part. The partial compression manner of the AI processing model may be applied to deployment of all AI processing models at the mobile sides.” – The partial compressed model being applied to deployment of all AI processing models that the mobile sides indicates that there is more than one mobile device which the first sub-model is deployed to and therefore the training using the intermediate data is analogous to training the third submodel using the first information. The first sub-model being deployed to a second mobile device would be analogous to the third sub-model which is identical to the first submodel. Therefore, the function of the third submodel is identical to the function of the first submodel and the input type of the third submodel is identical to the input type of the first submodel, and the dimension of input data and output data of the third submodel is identical to the input data and output data of the first submodel.)
Regarding claim 13, Pei teaches the communication method:
obtaining first information, wherein the first information indicates one or more of input data of a first submodel or output data of the first submodel; and (Pei, paragraph 0024: “After obtaining data to be processed, the mobile device 102 performs first processing on the data by using the first sub-model to obtain an intermediate processing result, and transmits the intermediate processing result to the server 106 by using the network 104.” – The server obtains the intermediate processing result of the first sub-model from the mobile device which is analogous to obtaining first information that indicates one or more output data of the first submodel as the intermediate result is the output of the first sub-model.)
training a third submodel based on the first information. (Pei, paragraph 0104: “For a partially compressed AI processing model, based on the foregoing training, the first n (for example, 3) neural network layers of the AI processing model may be trained and compressed by using a first algorithm (for example, a distillation method) for compressing a model and using obtained intermediate data of the high-accuracy AI processing model as a label, to obtain a mobile computing part.” – The partial compressed model being applied to deployment of all AI processing models that the mobile sides indicates that there is more than one mobile device which the first sub-model is deployed to and therefore the training using the intermediate data is analogous to training the third submodel using the first information.)
Regarding claim 14, Pei teaches the communication method of claim 13, as cited above.
Pei further teaches:
a function of the third submodel is identical to a function of the first submodel; and/or an input type of the third submodel is identical to an input type of the first submodel, and an output type of the third submodel is identical to an output type of the first submodel; and/or a dimension of input data of the third submodel is identical to a dimension of the input data of the first submodel, and a dimension of output data of the third submodel is identical to a dimension of the output data of the first submodel; and/or an input of the third submodel is identical to an input of the first submodel, and a difference between an output of the third submodel and an output of the first submodel is less than a first threshold. (Pei, paragraph 0104-0105: “When the AI processing model is trained, for an uncompressed AI processing model, an original model may be trained, to obtain a high-accuracy AI processing model. For a partially compressed AI processing model, based on the foregoing training, the first n (for example, 3) neural network layers of the AI processing model may be trained and compressed by using a first algorithm (for example, a distillation method) for compressing a model and using obtained intermediate data of the high-accuracy AI processing model as a label, to obtain a mobile computing part. Other neural network layers other than the first n (for example, 3) layers of the AI processing model may be trained by using a second algorithm (for example, transfer learning), using the intermediate data and/or final data of the high-accuracy AI processing model as a label, and using an output result of the mobile computing part as an input, to obtain a cloud computing part. The partial compression manner of the AI processing model may be applied to deployment of all AI processing models at the mobile sides.” – The first sub-model being deployed to a second mobile device would be analogous to the third sub-model which is identical to the first submodel. Therefore, the function of the third submodel is identical to the function of the first submodel and the input type of the third submodel is identical to the input type of the first submodel, and the dimension of input data and output data of the third submodel is identical to the input data and output data of the first submodel.)
Regarding claim 15, Pei teaches the communication method of claim 13, as cited above.
Pei further teaches:
wherein the first submodel and a second submodel are used in a matching manner. (Pei, paragraph 0046: “A computing task is processed by using the AI processing model in a manner of using the first sub-model in combination with the second sub-model, so as to reduce a computing workload and an internal memory overhead of a model of the AI processing model deployed in the mobile device, so that the AI processing model may be deployed in the mobile device.” – The first sub-model being used in combination with the second sub-model is analogous to the first submodel and second submodel being used in a matching manner.)
Regarding claim 16, Pei teaches the communications method of claim 15, as cited above.
Pei further teaches:
wherein the output of the first submodel is used to determine an input of the second submodel; or an output of the second submodel is used to determine the input of the first submodel. (Pei, paragraph 0063: “When the neural network layers included in the second computing segment are trained, the neural network layers included in the second computing segment may be trained based on a model training algorithm by using the output of the first sub-model as the input of the second computing segment and using a training data processing result of the AI processing model (on which model compression is not performed), to obtain a network recognition layer (the second sub-model) not sacrificing a final accuracy.”)
Regarding claim 17, Pei teaches the communication method of claim 15, as cited above.
Pei further teaches:
the first submodel is used to send information at a transmit end, and the second submodel is used to receive the information at a receive end; or the second submodel is used to send information at the transmit end, and the first submodel is used to receive the information at the receive end. (Pei, paragraphs 0030-0031: “Operation S204: Transmit the intermediate processing result to a first server, the first server being configured to perform second processing on the intermediate processing result by using the second sub-model, to obtain a target processing result, target processing including the first processing and the second processing. Operation S206: Receive the target processing result returned by the first server.” – The first sub-model is deployed on a mobile device, see paragraph 0023, and is therefore sending information at a transmit end and the second sub-model is deployed in the server and is therefore used to receive information at a receive end.)
Regarding claim 18, Pei teaches the communication method of claim 15, as cited above.
Pei further teaches:
wherein the first submodel and the second submodel belong to a bilateral model. (Pei, paragraph 0046: “A computing task is processed by using the AI processing model in a manner of using the first sub-model in combination with the second sub-model, so as to reduce a computing workload and an internal memory overhead of a model of the AI processing model deployed in the mobile device, so that the AI processing model may be deployed in the mobile device.” – The first sub-model used in combination with the second sub-model, wherein the output of the first sub-model is used as input into the second sub-model is analogous to the first submodel and second submodel belonging to a bilateral model.)
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 6-12 and 19-24 are rejected under 35 U.S.C. 103 as being unpatentable over Pei in view of Liu (US20240259072), hereinafter Liu.
Regarding claim 6, Pei teaches the communication method of claim 1, as cited above.
Pei does not explicitly teach:
the determining the first submodel and the second submodel comprises: determining the first submodel and the second submodel based on training data, wherein the training data comprises N pieces of channel information, N is a positive integer, and the channel information comprises a downlink channel feature or a downlink channel.
However, Liu teaches:
the determining the first submodel and the second submodel comprises: determining the first submodel and the second submodel based on training data, wherein the training data comprises N pieces of channel information, N is a positive integer, and the channel information comprises a downlink channel feature or a downlink channel. (Liu, paragraph 0015: “The terminal device sends the N bitstreams and the N pieces of codebook output information to a network device; the N pieces of codebook output information are used to obtain N first feedback information tags, and the N bitstreams and the N first feedback information tags are used by the network device to update a decoder in the channel feedback model, the decoder is configured to perform decoding based on a to-be-decoded bitstream of the terminal device to obtain corresponding feedback information.” And paragraph 0017-0018: “The first training module is configured to update a channel feedback model based on N first feedback information tags respectively corresponding to N pieces of channel information. The N first feedback information tags include N pieces of codebook output information obtained by respectively quantizing, based on a codebook, the N pieces of channel information; the channel feedback model is used to obtain corresponding feedback information based on channel information which needs to be feedback, and N is an integer greater than or equal to 1.” – The N feedback information tags corresponding to N pieces of channel information that is used to train the encoder and decoder, e.g., first and second sub-models, is analogous to the N pieces of channel information used as training data. And paragraph 0166: “Optionally, when update of the model is required, a party triggering the update sends indication information to notify another party to update the model.” And paragraph 0170: “As an example, various indication information in the embodiment of the disclosure may be Downlink Control Information (DCI) or Uplink Control Information (UCI), or may be indication information dedicated to AI functions. For example, the first indication information may be DCI. The second indication information may be UCI.” – The indication information is used to update the models and is therefore equivalent to the training using the channel information, the indication information including the downlink control information is analogous to the downlink channel information.)
Liu is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning and network communication. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Pei, which already teaches determining a first submodel and second submodel but does not explicitly teach that the training data comprises N pieces of channel information which includes a downlink channel feature or a downlink channel, to include the teachings of Liu which does teach that the training data comprises N pieces of channel information which includes a downlink channel feature or a downlink channel in order to improve accuracy, generalization, and stability. (Liu, paragraph 0061)
Regarding claim 7, Pei teaches the communication method of claim 1, as cited above.
Pei does not explicitly teach:
the input data of the first submodel comprises M pieces of channel information, and M is a positive integer.
However, Liu teaches:
the input data of the first submodel comprises M pieces of channel information, and M is a positive integer. (Liu, paragraph 0010: “A network device receives, from a terminal device, N pieces of codebook output information and N bitstreams respectively corresponding to N pieces of channel information, the N bitstreams are obtained by the terminal device respectively processing the N pieces of channel information by using an encoder in a channel feedback model, the N pieces of codebook output information are obtained by the terminal device respectively quantizing, based on a codebook, the N pieces of channel information.” – The N pieces of channel information are used in the encoder, e.g., the first submodel and is therefore analogous to the input data of the first submodel comprising M pieces of channel information. Paragraph 0018 notes that N is an integer greater than or equal to 1.)
Liu is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning and network communication. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Pei, which already teaches the communication method wherein the first submodel has input data but does not explicitly teach that the input data of the first submodel comprises M pieces of channel information with M being a positive integer, to include the teachings of Liu which does teach that the input data of the first submodel comprises M pieces of channel information with M being a positive integer in order to improve accuracy, generalization, and stability. (Liu, paragraph 0061)
Regarding claim 8, Pei and Liu teach the communication method of claim 7, as cited above.
Pei does not explicitly teach:
the output data of the first submodel comprises feature bits corresponding to the M pieces of channel information.
However, Liu further teaches:
the output data of the first submodel comprises feature bits corresponding to the M pieces of channel information. (Liu, paragraph 0020: “The second communication module is configured to receive, from a terminal device, N pieces of codebook output information and N bitstreams respectively corresponding to N pieces of channel information, the N bitstreams are obtained by the terminal device respectively processing the N pieces of channel information by using an encoder in a channel feedback model, the N pieces of codebook output information are obtained by the terminal device respectively quantizing, based on a codebook, the N pieces of channel information.” – The N bitstreams are output from the encoder, e.g., the first submodel, and are therefore analogous to the feature bits corresponding to the M pieces of channel information.)
Regarding claim 9, Pei teaches the communication method of claim 5, as cited above.
Pei does not explicitly teach:
obtaining information indicating a first feature bit, wherein the output of the third submodel comprises the first feature bit; and
obtaining first channel information based on the second submodel and the first feature bit, wherein the input of the second submodel comprises the first feature bit, and the output of the second submodel comprises the first channel information.
However, Liu teaches:
obtaining information indicating a first feature bit, wherein the output of the third submodel comprises the first feature bit; and (Liu, paragraph 0020: “The second communication module is configured to receive, from a terminal device, N pieces of codebook output information and N bitstreams respectively corresponding to N pieces of channel information, the N bitstreams are obtained by the terminal device respectively processing the N pieces of channel information by using an encoder in a channel feedback model, the N pieces of codebook output information are obtained by the terminal device respectively quantizing, based on a codebook, the N pieces of channel information.” – The third sub-model is identical to the first sub-model, as taught by Pei above. Thus, the encoder on the terminal device is analogous to the third sub-model and the second communication module receiving the bitstream is analogous to obtaining a first feature bit that is output of the third submodel.)
obtaining first channel information based on the second submodel and the first feature bit, wherein the input of the second submodel comprises the first feature bit, and the output of the second submodel comprises the first channel information. (Liu, paragraphs 0021-0022: “The first processing module is configured to obtain N first feedback information tags based on the N pieces of codebook output information, and update a decoder in the channel feedback model based on the N bitstreams and the N first feedback information tags. The decoder is configured to perform decoding based on a to-be-decoded bitstream received from the terminal device to obtain corresponding feedback information.” – The decoder is analogous to the second submodel and therefore the decoder performing decoding based on the bitstream to obtain feedback information is analogous to the second submodel using the feature bit as input and outputting the first channel information.)
Liu is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning and network communication. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Pei, which already teaches a first submodel, a second submodel, and a third submodel being generated, where the third submodel is identical to the first submodel but does not explicitly teach the output of the third submodel comprises the first feature bit which is input into the second submodel to obtain first channel information, to include the teachings of Liu which does teach the output of the third submodel comprises the first feature bit which is input into the second submodel to obtain first channel information in order to improve accuracy, generalization, and stability. (Liu, paragraph 0061)
Regarding claim 10, Pei teaches the communication method of claim 1, as cited above.
Pei does not explicitly teach:
the input data of the first submodel comprises M feature bits, and M is a positive integer.
However, Liu teaches:
the input data of the first submodel comprises M feature bits, and M is a positive integer. (Liu, paragraph 0021: “The first processing module is configured to obtain N first feedback information tags based on the N pieces of codebook output information, and update a decoder in the channel feedback model based on the N bitstreams and the N first feedback information tags.” And paragraph 0008: “The channel feedback model is used to obtain corresponding feedback information based on channel information which needs to be feedback, the N first feedback information tags include N pieces of codebook output information obtained by respectively quantizing, based on a codebook, the N pieces of channel information; and N is an integer greater than or equal to 1.” – Claim 1 does not provide any specifics on the first submodel or the second submodel, other than specifying them as “first” and “second.” Therefore, the decoder of Liu is analogous to the first submodel and the input to the decoder is the N bitstreams which is analogous to the M feature bits.)
Liu is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning and network communication. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Pei, which already teaches a first submodel with input data but does not explicitly teach that the input data of the first submodel comprises M feature bits and M being a positive integer, to include the teachings of Liu which does teach that the input data of the first submodel comprises M feature bits and M being a positive integer in order to improve accuracy, generalization, and stability. (Liu, paragraph 0061)
Regarding claim 11, Pei and Liu teach the communication method of claim 10, as cited above.
Pei does not explicitly teach:
the output data of the first submodel comprises channel information corresponding to the M feature bits.
However, Liu further teaches:
the output data of the first submodel comprises channel information corresponding to the M feature bits. (Liu, paragraph 0022: “The decoder is configured to perform decoding based on a to-be-decoded bitstream received from the terminal device to obtain corresponding feedback information.” And paragraph 0061: “Since a tag information-based deep learning training mode is used, the channel feedback model may accurately obtain corresponding feedback information based on to-be-predicted channel information.” – The output of the decoder, i.e. the first submodel, is the feedback information which is channel feedback and is analogous to the channel information corresponding to the M feature bits.)
Regarding claim 12, Pei and Liu teach the communication method of claim 10, as cited above.
Pei does not explicitly teach:
determining a second feature bit based on second channel information and the second submodel, wherein the input of the second submodel comprises the second channel information, and the output of the second submodel comprises the second feature bit; and
sending information indicating the second feature bit.
However, Liu further teaches:
determining a second feature bit based on second channel information and the second submodel, wherein the input of the second submodel comprises the second channel information, and the output of the second submodel comprises the second feature bit; and (Liu, paragraph 0099: “That is, the channel feedback model includes an encoder and a decoder, the encoder is used by the terminal device to perform encoding based on the channel information which needs to be feedback to obtain a corresponding bitstream, and the decoder is used by the network device to perform decoding based on the bitstream to obtain corresponding feedback information.” – The decoder is analogous to the first submodel, as noted in claim 10 above. Therefore, the encoder is analogous to the second submodel. The encoder takes the channel information as an input, which is analogous to the second channel information, and outputs a corresponding bitstream, which is analogous to the second feature bit.)
sending information indicating the second feature bit. (Liu, paragraph 0155: “The network device receives, from a terminal device, N pieces of codebook output information and N bitstreams respectively corresponding to N pieces of channel information, the N bitstreams are obtained by the terminal device respectively processing the N pieces of channel information by using an encoder in the channel feedback model.” – The encoder is on the terminal device, therefore the network device receiving the bitstreams from the terminal device is analogous to sending information indicating the second feature bit.)
Regarding claim 19, Pei teaches the communication method of claim 13, as cited above.
Pei does not explicitly teach:
the input data of the first submodel comprises M pieces of channel information, and M is a positive integer.
However, Liu teaches:
the input data of the first submodel comprises M pieces of channel information, and M is a positive integer. (Liu, paragraph 0010: “A network device receives, from a terminal device, N pieces of codebook output information and N bitstreams respectively corresponding to N pieces of channel information, the N bitstreams are obtained by the terminal device respectively processing the N pieces of channel information by using an encoder in a channel feedback model, the N pieces of codebook output information are obtained by the terminal device respectively quantizing, based on a codebook, the N pieces of channel information.” – The N pieces of channel information are used in the encoder, e.g., the first submodel and is therefore analogous to the input data of the first submodel comprising M pieces of channel information. Paragraph 0018 notes that N is an integer greater than or equal to 1.)
Liu is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning and network communication. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Pei, which already teaches the communication method wherein the first submodel has input data but does not explicitly teach that the input data of the first submodel comprises M pieces of channel information with M being a positive integer, to include the teachings of Liu which does teach that the input data of the first submodel comprises M pieces of channel information with M being a positive integer in order to improve accuracy, generalization, and stability. (Liu, paragraph 0061)
Regarding claim 20, Pei and Liu teach the communication method of claim 19, as cited above.
Pei does not explicitly teach:
the output data of the first submodel comprises feature bits corresponding to the M pieces of channel information.
However, Liu further teaches:
the output data of the first submodel comprises feature bits corresponding to the M pieces of channel information. (Liu, paragraph 0022: “The decoder is configured to perform decoding based on a to-be-decoded bitstream received from the terminal device to obtain corresponding feedback information.” And paragraph 0061: “Since a tag information-based deep learning training mode is used, the channel feedback model may accurately obtain corresponding feedback information based on to-be-predicted channel information.” – The N bitstreams are output from the encoder, e.g., the first submodel, and are therefore analogous to the feature bits corresponding to the M pieces of channel information.)
Regarding claim 21, Pei teaches the communication method of claim 13, as cited above.
Pei does not explicitly teach:
determining a first feature bit based on third channel information and the third submodel, wherein the input of the third submodel comprises the third channel information, and the output of the third submodel comprises the first feature bit; and
sending information indicating the first feature bit.
However, Liu teaches:
determining a first feature bit based on third channel information and the third submodel, wherein the input of the third submodel comprises the third channel information, and the output of the third submodel comprises the first feature bit; and (Liu, paragraph 0099: “That is, the channel feedback model includes an encoder and a decoder, the encoder is used by the terminal device to perform encoding based on the channel information which needs to be feedback to obtain a corresponding bitstream, and the decoder is used by the network device to perform decoding based on the bitstream to obtain corresponding feedback information.” – The third submodel is identical to the first submodel, as noted above. Therefore the channel information being input into the encoder to obtain a corresponding bitstream is analogous to the third channel information being input to the third submodel and output of the third submodel comprising the first feature bit.)
sending information indicating the first feature bit. (Liu, paragraph 0155: “The network device receives, from a terminal device, N pieces of codebook output information and N bitstreams respectively corresponding to N pieces of channel information, the N bitstreams are obtained by the terminal device respectively processing the N pieces of channel information by using an encoder in the channel feedback model.”- The network device receiving the bitstreams from the terminal device is analogous to the terminal device sending the information indicating the first feature bit.)
Liu is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning and network communication. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Pei, which already teaches generating a third submodel but does not explicitly teach using the third submodel to determine and send a first feature bit, to include the teachings of Liu which does teach using the third submodel to determine and send a first feature bit in order to improve accuracy, generalization, and stability. (Liu, paragraph 0061)
Regarding claim 22, Pei teaches the communication method of claim 13, as cited above.
Pei does not explicitly teach:
an input parameter of the first submodel comprises M feature bits, and M is a positive integer.
However, Liu teaches:
an input parameter of the first submodel comprises M feature bits, and M is a positive integer. (Liu, paragraph 0022: “The decoder is configured to perform decoding based on a to-be-decoded bitstream received from the terminal device to obtain corresponding feedback information.” And paragraph 0061: “Since a tag information-based deep learning training mode is used, the channel feedback model may accurately obtain corresponding feedback information based on to-be-predicted channel information.” – Claim 13 does not provide any specifics on the first submodel or the third submodel, other than specifying them as “first” and “third.” Therefore, the decoder of Liu is analogous to the first submodel and the input to the decoder is the N bitstreams which is analogous to the M feature bits.)
Liu is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning and network communication. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Pei, which already teaches a first submodel with input data but does not explicitly teach that the input data of the first submodel comprises M feature bits and M being a positive integer, to include the teachings of Liu which does teach that the input data of the first submodel comprises M feature bits and M being a positive integer in order to improve accuracy, generalization, and stability. (Liu, paragraph 0061)
Regarding claim 23, Pei and Liu teach the communication method of claim 22, as cited above.
Pei does not explicitly teach:
an output parameter of the first submodel comprises channel information corresponding to the M feature bits.
However, Liu further teaches:
an output parameter of the first submodel comprises channel information corresponding to the M feature bits. (Liu, paragraph 0022: “The decoder is configured to perform decoding based on a to-be-decoded bitstream received from the terminal device to obtain corresponding feedback information.” And paragraph 0061: “Since a tag information-based deep learning training mode is used, the channel feedback model may accurately obtain corresponding feedback information based on to-be-predicted channel information.” – The output of the decoder, i.e. the first submodel, is the feedback information which is channel feedback and is analogous to the channel information corresponding to the M feature bits.)
Regarding claim 24, Pei and Liu teach the communication method of claim 22, as cited above.
Pei does not explicitly teach:
obtaining information indicating a second feature bit; and
obtaining fourth channel information based on the third submodel and the second feature bit, wherein the input of the third submodel comprises the second feature bit, and the output of the third submodel comprises the fourth channel information.
However, Liu further teaches:
obtaining information indicating a second feature bit; and (Liu, paragraph 0020: “The second communication module is configured to receive, from a terminal device, N pieces of codebook output information and N bitstreams respectively corresponding to N pieces of channel information, the N bitstreams are obtained by the terminal device respectively processing the N pieces of channel information by using an encoder in a channel feedback model, the N pieces of codebook output information are obtained by the terminal device respectively quantizing, based on a codebook, the N pieces of channel information.” – Obtaining N bitstreams corresponding to N pieces of channel information indicates that the method is capable of obtaining a second bitstream which is analogous to obtaining information indicating a second feature bit.)
obtaining fourth channel information based on the third submodel and the second feature bit, wherein the input of the third submodel comprises the second feature bit, and the output of the third submodel comprises the fourth channel information. (Liu, paragraphs 0021-0022: “The first processing module is configured to obtain N first feedback information tags based on the N pieces of codebook output information, and update a decoder in the channel feedback model based on the N bitstreams and the N first feedback information tags. The decoder is configured to perform decoding based on a to-be-decoded bitstream received from the terminal device to obtain corresponding feedback information.” – The third sub-model is identical to the first sub-model, as taught by Pei and Liu above. Thus, the decoder on the terminal device is analogous to the third sub-model and therefore the decoder performing decoding based on the bitstream to obtain feedback information is analogous to the second submodel using the feature bit as input and outputting the fourth channel information.)
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Suo and Yang (US20240320547)
Tao et al. (US20220215298)
Dai et al. (CiNet: Redesigning Deep Neural Networks for Efficient Mobile-Cloud Collaborative Inference)
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JACQUELINE MEYER whose telephone number is (703)756-5676. The examiner can normally be reached M-F 8:00 am - 4:30 pm 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, Tamara Kyle can be reached at 571-272-4241. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/J.C.M./Examiner, Art Unit 2144
/SHOURJO DASGUPTA/Primary Examiner, Art Unit 2144
Prosecution Insights