DETAILED ACTION
Applicant’s response filed on 12/01/2025 has been entered and made of record.
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 .
Claim Status
Claims 1-20, 23 and 25-30 are amended.
No new claim is/are added.
Claims 1-30 are pending for examination.
Applicant Argument
Applicant’s arguments (remark pages 11-16), filed on 12/01/2025, with respect to claims 1-30 have been considered but are moot in view of the new ground of rejection below which better address the claimed invention as amended.
This Office Action is made Final.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1-4, 6-7, 10-21, 23-24 and 26-30 are rejected under 35 U.S.C. 103 as being unpatentable over Timo et al. (US 20220149904 A1), hereinafter “Timo”, in view of Song et al. (US 20230040739 A1), hereinafter “Song”.
Per claim 1, 11, 18 and 26:
Regarding claim 1, Timo teaches ‘An apparatus for wireless communication at a user equipment (UE)’ (Timo: [FIG.5]: “Terminal device”; [0125]: “The terminal device may for example be a user equipment (UE)”); ‘comprising: one or more memories’ (Timo: [0082]: “memory”); ‘one or more processors’ (Timo: [0342]: “processor”; [0082]: “processing circuitry”); ‘coupled with the one or more memories’ (this is implied), ‘configured to cause the UE to’ (Timo: [0082]: “The at least one memory may for example contain instructions executable by the processing circuitry whereby the terminal device is operable to perform the method”);
‘receive a first message that indicates’ (Timo: [FIG.5]: step 503: “Network node”-> “Terminal node”; [0144]: “the network node 501 transmits 403 the first set of parameters 503, and the terminal device 502 receives it”; [0083]: “The first set of parameters indicates a compression function for compressing downlink channel estimates at a terminal device”; [0206]-[0208]: “channel state compression (CSC) framework … a configurable CS compressor”);
‘a first configuration associated with a codebook-based channel state information compression scheme’ (Timo: [0049]: “CSI Feedback: 3GPP Implicit Type I and Type II”; [0054]: “The Type I precoder codebooks”; [0057]: “type II feedback consists of explicit feedback and/or codebook-based feedback with higher spatial resolution”; [0255]-[0259]: “To enable the proposed CSC framework, the CSI ReportConfig can be modified to include: an advanced CSI Type II with “compressed explicit channel matrix feedback … If configured for Type-I or Type-II CSI feedback, the ReportConfig contains a CodebookConfig that specifies configuration parameters for Type-1 and Type-II CSI feedback”; codebook-based compression function);
‘a second configuration associated with a neural network-based channel state information compression scheme’ (Timo: [0067]: “CSI Feedback: Compressive-Sensing Dimensionality Reduction”; [0146]: “the second function of the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks”; [FIG.15]; [0268]: “a compression function 1501 and a decompression function 1502 provided in the form of a neural network”, neural network-based compression function);
‘transmit, based at least in part on reception of the first message, a second message that comprises channel information estimated by the UE and an indication that indicates one or more compression schemes, from the codebook-based channel state information compression scheme and the neural network-based channel state information compression scheme, selected by the UE to compress the channel information indicated via the second message’ (Timo: [FIG.5]: step 504: “Terminal device”-> “Network node”; [0144]: “The terminal device 502 transmits 305 the compressed downlink channel estimates 504, and the network node 501 received them”; [FIG.10]: step 1009: “Transmit updated parameter value”; [FIG.11]: step 1102: “Terminal device”-> “Network node”, step 1104: “Terminal device”-> “Network Node”; [0195]: “one or more parameters 1103 indicates an objective function for evaluating performance of the compression function … the updated value for at least one parameter from the first set of parameters is determined at step 1002 using the objective function”; [0198]: “the terminal device 502 indicates via the updated value that the network node 502 should use a different decompression function … the updated value 1104 for at least one parameter from the second set of parameters may be transmitted 1009 from the terminal device 502 to the network node 501”, updated value indicating the decompression function corresponding to a more suitable compressed function selected by terminal device; [0131]: “the compression function is formed based on the first set of parameters, these parameters may be employed to control which compression function to be used at the terminal device”; [0191]: “The terminal device may determine of compute updated values for one of more of the first set of parameters, so as to obtain a more suitable compression function”; terminal device indicating to network node that the compressed downlink channel estimates are compressed by codebook-based compression function or neural network-based compression function (to decompress by codebook-based decompression function or neural network-based decompression function)). However, Timo fails to expressly teach a compressed CSI and an indication in one message.
However, Song in the same field of endeavor teaches terminal device sends an indication of the codebook (compressed function) together with the compressed CSI (Song: [0090]: “the terminal device 120 may transmit, to the network device 110, the indication of the codebook together with the compressed CSI … thus a reduced complexity can be achieved at the terminal device”; [0110]: “an indication of a codebook for compressing the channel state information”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Song’s teaching with that of Timo in order to acquire CSI more accurately with reduced complexity (Song: [Abstract]: “CSI can be acquired more accurately with reduced complexity and overhead”).
Regarding claim 11, Timo teaches ‘An apparatus for wireless communication at a network entity’ (Timo: [FIG.5]: “Network node”; [0133]: “The network node may for example be referred to as a base station”); ‘comprising: one or more memories’ (Timo: [0085]: “memory”); ‘one or more processors’ (Timo: [0342]: “processor”; [0085]: “processing circuitry”); ‘coupled with the one or more memories’ (this is implied); ‘configured to cause the network entity’ (Timo: [0085]: “The at least one memory may for example contain instructions executable by the processing circuitry whereby the network node is operable to perform the method”);
‘transmit a first message that indicates’ (Timo: [FIG.5]: step 503: “Network node”-> “Terminal node”; [0144]: “the network node 501 transmits 403 the first set of parameters 503, and the terminal device 502 receives it”; [0083]: “The first set of parameters indicates a compression function for compressing downlink channel estimates at a terminal device”; [0206]-[0208]: “channel state compression (CSC) framework … a configurable CS compressor”);
‘a first configuration associated with a codebook-based channel state information compression scheme’ (Timo: [0049]: “CSI Feedback: 3GPP Implicit Type I and Type II”; [0054]: “The Type I precoder codebooks”; [0057]: “type II feedback consists of explicit feedback and/or codebook-based feedback with higher spatial resolution”; [0255]-[0259]: “To enable the proposed CSC framework, the CSI ReportConfig can be modified to include: an advanced CSI Type II with “compressed explicit channel matrix feedback … If configured for Type-I or Type-II CSI feedback, the ReportConfig contains a CodebookConfig that specifies configuration parameters for Type-1 and Type-II CSI feedback”; codebook-based compression function);
‘a second configuration associated with a neural network-based channel state information compression scheme’ (Timo: [0067]: “CSI Feedback: Compressive-Sensing Dimensionality Reduction”; [0146]: “the second function of the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks”; [FIG.15]; [0268]: “a compression function 1501 and a decompression function 1502 provided in the form of a neural network”, neural network-based compression function);
‘receive, based at least in part on transmission of the first message, a second message that comprises channel information estimated by a user equipment (UE) and an indication that indicates one or more compression schemes, from the codebook- based channel state information compression scheme and the neural network-based channel state information compression scheme, selected by the UE to compress the channel information indicated via the second message’ (Timo: [FIG.5]: step 504: “Terminal device”-> “Network node”; [0144]: “The terminal device 502 transmits 305 the compressed downlink channel estimates 504, and the network node 501 received them”; [FIG.10]: step 1009: “Transmit updated parameter value”; [FIG.11]: step 1102: “Terminal device”-> “Network node”, step 1104: “Terminal device”-> “Network Node”; [0195]: “one or more parameters 1103 indicates an objective function for evaluating performance of the compression function … the updated value for at least one parameter from the first set of parameters is determined at step 1002 using the objective function”; [0198]: “the terminal device 502 indicates via the updated value that the network node 502 should use a different decompression function … the updated value 1104 for at least one parameter from the second set of parameters may be transmitted 1009 from the terminal device 502 to the network node 501”, updated value indicating the decompression function corresponding to a more suitable compressed function selected by terminal device; [0131]: “the compression function is formed based on the first set of parameters, these parameters may be employed to control which compression function to be used at the terminal device”; [0191]: “The terminal device may determine of compute updated values for one of more of the first set of parameters, so as to obtain a more suitable compression function”; terminal device indicating to network node that the compressed downlink channel estimates are compressed by codebook-based compression function or neural network-based compression function (to decompress by codebook-based decompression function or neural network-based decompression function)). However, Timo fails to expressly teach a compressed CSI and an indication in one message.
However, Song teaches terminal device sends an indication of the codebook (compressed function) together with the compressed CSI (Song: [0090]: “the terminal device 120 may transmit, to the network device 110, the indication of the codebook together with the compressed CSI … thus a reduced complexity can be achieved at the terminal device”; [0110]: “an indication of a codebook for compressing the channel state information”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Song’s teaching with that of Timo in order to acquire CSI more accurately with reduced complexity (Song: [Abstract]: “CSI can be acquired more accurately with reduced complexity and overhead”).
Regarding claim 18, claim 18 recites the method implemented by the apparatus of claim 1 (see rejection of claim 1 above).
Regarding claim 26, claim 26 recites the method implemented by the apparatus of claim 11 (see rejection of claim 11 above).
Per claim 2 and 19:
Regarding claim 2, combination of Timo and Song teaches the apparatus of claim 1 (discussed above).
Timo teaches ‘determine, using a neural network associated with the neural network-based channel state information compression scheme’ (Timo: [FIG.15]; [0268]: “a compression function 1501 and a decompression function 1502 provided in the form of a neural network”, neural network-based compression function);
‘based at least in part on the channel information’ (Timo: [0040]: “The UEs estimate their channels (or important parameters thereof) using the downlink reference signals”; [0067]-[0068]: “CSI Feedback: Compressive-Sensing Dimensionality Reduction … If the channels are sparse in some basis, but the exact sparse basis is not known, then it is possible to use compressive-sensing techniques”; [0049]-[0056]: “CSI Feedback: 3GPP Implicit Type I and Type II … Type-I feedback is mostly useful for SU-MIMO … Type-II CSI will enable the BS to perform more advanced MU-MIMO”);
‘whether to compress the channel information in accordance with the codebook-based channel state information compression scheme, or in accordance with the neural network-based channel state information compression scheme, or both’ (Timo: [0206]: “channel state compression (CSC) framework”; [0243]-[0250]: “The CSC parameters (a*,b*) … The parameters (a*,b*) can be tailored to match the rank and expected codebook-based precoding mode … The parameters (a*,b*) can also be learned (or, periodically updated) by the BS using a supervised-learning or reinforcement-learning process”; accordance with codebook-based compression function or accordance with neural-network-based compression function; [0289]-[0309]: “The encoder/decoder weights and encoder/decoder biases can be updated … choosing the closed U and V in the codebook to the trained solution … with autoencoder replaced by a much more sophisticated network (for example, a deep convolutions neural network”; [0146]: “the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks. An activation function may provide a “threshold” that allows things to be “tuned on” (i.e. activated) or “turned off” (i.e. deactivated)”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0232]: “The instantaneous performance of (fa, gb) on
PNG
media_image1.png
24
121
media_image1.png
Greyscale
can be quantified by the compression rate …
PNG
media_image2.png
35
93
media_image2.png
Greyscale
”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0272]: “The non-linear activation function 1509 can turn on/off different elements in the output from linear part 1508 to achieve better compression performance”).
Regarding claim 19, claim 19 recites the method implemented by the apparatus of claim 2 (see rejection of claim 2 above).
Per claim 3 and 20:
Regarding claim 3, combination of Timo and Song teaches the apparatus of claim 2 (discussed above).
Timo teaches ‘wherein the first message comprises a configuration of the neural network for determination of’ (Timo: [0126]: “The first set of parameters may for example be configured/selected by a network node and communicated to the terminal device using a CSI report configuration”; [0131]: “the compression function is formed based on the first set of parameters, these parameters may be employed to control which compression function to be used at the terminal device”; [FIG.15]; [0268]: “a compression function 1501 and a decompression function 1502 provided in the form of a neural network”; [0173]: “In the autoencoder example described below with reference to FIG. 15, the compression function 1501 only has a single part 1508 defined via the first set of parameters”);
‘whether to compress the channel information in accordance with the codebook-based channel state information compression scheme, or in accordance with the neural network-based channel state information compression scheme, or both’ (Timo: [0206]: “channel state compression (CSC) framework”; [0243]-[0250]: “The CSC parameters (a*,b*) … The parameters (a*,b*) can be tailored to match the rank and expected codebook-based precoding mode … The parameters (a*,b*) can also be learned (or, periodically updated) by the BS using a supervised-learning or reinforcement-learning process”; accordance with codebook-based compression function or accordance with neural-network-based compression function; [0146]: “the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks. An activation function may provide a “threshold” that allows things to be “tuned on” (i.e. activated) or “turned off” (i.e. deactivated)”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0232]: “The instantaneous performance of (fa, gb) on
PNG
media_image1.png
24
121
media_image1.png
Greyscale
can be quantified by the compression rate …
PNG
media_image2.png
35
93
media_image2.png
Greyscale
”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0272]: “The non-linear activation function 1509 can turn on/off different elements in the output from linear part 1508 to achieve better compression performance”).
Regarding claim 20, claim 20 recites the method implemented by the apparatus of claim 3 (see rejection of claim 3 above).
Per claim 4 and 21:
Regarding claim 4, combination of Timo and Song teaches the apparatus of claim 1 (discussed above).
Timo teaches ‘determine a first mean squared error associated with a compression in accordance with the codebook-based channel state information compression scheme’ (Timo: [0231]-[0238]: “compressor-decompressor-cost function … cost functions are the average square error … codebook of precoders”, mean (average) squared error associated with a codebook based compression function; [0195]: “one or more parameters 1103 indicates an objective function for evaluating performance of the compression function”; [0131]: “the compression function is formed based on the first set of parameters, these parameters may be employed to control which compression function to be used at the terminal device”);
‘determine a second mean squared error associated with a compression in accordance with the neural network-based channel state information compression scheme’ (Timo: [0231]-[0234]: “compressor-decompressor-cost function … cost functions are the average square error”; [0268]: “a compression function 1501 and a decompression function 1502 provided in the form of a neural network”, mean (average) squared error associated with neural network-based compression function; [0195]: “one or more parameters 1103 indicates an objective function for evaluating performance of the compression function”; [0131]: “the compression function is formed based on the first set of parameters, these parameters may be employed to control which compression function to be used at the terminal device”);
‘determine, based at least in part on a comparison between the first mean squared error and the second mean squared error’ (Timo: [0215]: “A cost function (also referred to as an objective function)”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”, based on the cost (compare the MSEs) and choose a compression function with better accuracy (more similar to the original CSI));
‘whether to compress the channel information in accordance with the codebook-based channel state information compression scheme, or in accordance with the neural network-based channel state information compression scheme, or both’ (Timo: [0206]: “channel state compression (CSC) framework”; [0243]-[0250]: “The CSC parameters (a*,b*) … The parameters (a*,b*) can be tailored to match the rank and expected codebook-based precoding mode … The parameters (a*,b*) can also be learned (or, periodically updated) by the BS using a supervised-learning or reinforcement-learning process”, accordance with codebook-based compression function or accordance with neural-network-based compression function; [0289]-[0309]: “The encoder/decoder weights and encoder/decoder biases can be updated … choosing the closed U and V in the codebook to the trained solution … with autoencoder replaced by a much more sophisticated network (for example, a deep convolutions neural network”; [0146]: “the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks. An activation function may provide a “threshold” that allows things to be “tuned on” (i.e. activated) or “turned off” (i.e. deactivated)”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0232]: “The instantaneous performance of (fa, gb) on
PNG
media_image1.png
24
121
media_image1.png
Greyscale
can be quantified by the compression rate …
PNG
media_image2.png
35
93
media_image2.png
Greyscale
”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0272]: “The non-linear activation function 1509 can turn on/off different elements in the output from linear part 1508 to achieve better compression performance”).
Regarding claim 21, claim 21 recites the method implemented by the apparatus of claim 4 (see rejection of claim 4 above).
Per claim 6 and 23:
Regarding claim 6, combination of Timo and Song teaches the apparatus of claim 1 (discussed above).
Timo teaches ‘wherein the first message indicates a criteria for the UE to determine’ (Timo: [0195]: “The third set of one or more parameters 1103 indicates an objective function for evaluating performance of the compression function. In this scenario, the method 1000 comprises receiving 1007 the third set of one or more parameters 1103. In the present embodiment, the updated value for at least one parameter from the first set of parameters is determined at step 1002 using the objective function”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; signaling indicating a criteria (an objective function to evaluate the accuracy of estimates));
‘whether to compress the channel information in accordance with the codebook-based channel state information compression scheme, or in accordance with the neural network-based channel state information compression scheme, or both’ (Timo: [0206]: “channel state compression (CSC) framework”; [0243]-[0250]: “The CSC parameters (a*,b*) … The parameters (a*,b*) can be tailored to match the rank and expected codebook-based precoding mode … The parameters (a*,b*) can also be learned (or, periodically updated) by the BS using a supervised-learning or reinforcement-learning process”, accordance with codebook-based compression function or accordance with neural-network-based compression function; [0289]-[0309]: “The encoder/decoder weights and encoder/decoder biases can be updated … choosing the closed U and V in the codebook to the trained solution … with autoencoder replaced by a much more sophisticated network (for example, a deep convolutions neural network”; [0146]: “the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks. An activation function may provide a “threshold” that allows things to be “tuned on” (i.e. activated) or “turned off” (i.e. deactivated)”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0232]: “The instantaneous performance of (fa, gb) on
PNG
media_image1.png
24
121
media_image1.png
Greyscale
can be quantified by the compression rate …
PNG
media_image2.png
35
93
media_image2.png
Greyscale
”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0272]: “The non-linear activation function 1509 can turn on/off different elements in the output from linear part 1508 to achieve better compression performance”);
Regarding claim 23, claim 23 recites the method implemented by the apparatus of claim 6 (see rejection of claim 6 above).
Per claim 7 and 24:
Regarding claim 7, combination of Timo and Song teaches the apparatus of claim 1 (discussed above).
Timo teaches ‘train an encoder based at least in part on a first decoder indicated by the first configuration, or on a second decoder indicated by the second configuration, or both’ (Timo: [0211]-[0219]: “the CSC framework can be used … While connected to the BS, the BS and/or UE can evaluate the compression performance and, if needed, train and update the configuration of the CS compressor and decompressor”; [0220]-[0267]: “The 3GPP standard can be modified to define a class of channel state compression (CSC) encoders … Updating CSC Parameters”);
‘compress the channel information based at least in part on the trained encoder’ (Timo: [0267]-[0268]: “CS Compressors and Decompressors Based on an Autoencoder Framework .. compression and decompression is based on an autoencoder framework”).
Regarding claim 24, claim 24 recites the method implemented by the apparatus of claim 7 (see rejection of claim 7 above).
Per claim 10 and 17:
Regarding claim 10, combination of Timo and Song teaches the apparatus of claim 1 (discussed above).
Timo teaches ‘one or more antennas’ (Timo: [0002]: “Antenna”); ‘receive the first message’ (Timo: [FIG.5]: step 503: “Network node”-> “Terminal node”; [0144]: “the network node 501 transmits 403 the first set of parameters 503, and the terminal device 502 receives it”), ‘transmit the second message, or both’ (these are optional).
Regarding claim 17, combination of Timo and Song teaches the apparatus of claim 11 (discussed above).
Timo teaches ‘one or more antennas’ (Timo: [0002]: “Antenna”); ‘receive the first message’ (Timo: [FIG.5]: step 503: “Network node”-> “Terminal node”; [0144]: “the network node 501 transmits 403 the first set of parameters 503, and the terminal device 502 receives it”), ‘transmit the second message, or both’ (these are optional).
Per claim 12 and 27:
Regarding claim 12, combination of Timo and Song teaches the apparatus of claim 11 (discussed above).
Timo teaches ‘wherein the first message indicates a criteria for a determination of determine whether to compress the channel information in accordance with the codebook-based channel state information compression scheme, or in accordance with the neural network-based channel state information compression scheme, or both’ (Timo: [0206]: “channel state compression (CSC) framework”; [0243]-[0250]: “The CSC parameters (a*,b*) … The parameters (a*,b*) can be tailored to match the rank and expected codebook-based precoding mode … The parameters (a*,b*) can also be learned (or, periodically updated) by the BS using a supervised-learning or reinforcement-learning process”; accordance with codebook-based compression function or accordance with neural-network-based compression function; [0289]-[0309]: “The encoder/decoder weights and encoder/decoder biases can be updated … choosing the closed U and V in the codebook to the trained solution … with autoencoder replaced by a much more sophisticated network (for example, a deep convolutions neural network”; [0146]: “the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks. An activation function may provide a “threshold” that allows things to be “tuned on” (i.e. activated) or “turned off” (i.e. deactivated)”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0232]: “The instantaneous performance of (fa, gb) on
PNG
media_image1.png
24
121
media_image1.png
Greyscale
can be quantified by the compression rate …
PNG
media_image2.png
35
93
media_image2.png
Greyscale
”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0272]: “The non-linear activation function 1509 can turn on/off different elements in the output from linear part 1508 to achieve better compression performance”).
Regarding claim 27, claim 27 recites the method implemented by the apparatus of claim 12 (see rejection of claim 12 above).
Per claim 13 and 28:
Regarding claim 13, combination of Timo and Song teaches the apparatus of claim 11 (discussed above).
Timo teaches ‘wherein the first message comprises a configuration of a neural network at the UE for determination of’ (Timo: [0126]: “The first set of parameters may for example be configured/selected by a network node and communicated to the terminal device using a CSI report configuration”; [0131]: “the compression function is formed based on the first set of parameters, these parameters may be employed to control which compression function to be used at the terminal device”; [FIG.15]; [0268]: “a compression function 1501 and a decompression function 1502 provided in the form of a neural network”; [0173]: “In the autoencoder example described below with reference to FIG. 15, the compression function 1501 only has a single part 1508 defined via the first set of parameters”);
‘whether to compress the channel information in accordance with the codebook-based channel state information compression scheme, or in accordance with the neural network-based channel state information compression scheme, or both’ (Timo: [0206]: “channel state compression (CSC) framework”; [0243]-[0250]: “The CSC parameters (a*,b*) … The parameters (a*,b*) can be tailored to match the rank and expected codebook-based precoding mode … The parameters (a*,b*) can also be learned (or, periodically updated) by the BS using a supervised-learning or reinforcement-learning process”; accordance with codebook-based compression function or accordance with neural-network-based compression function; [0289]-[0309]: “The encoder/decoder weights and encoder/decoder biases can be updated … choosing the closed U and V in the codebook to the trained solution … with autoencoder replaced by a much more sophisticated network (for example, a deep convolutions neural network”; [0146]: “the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks. An activation function may provide a “threshold” that allows things to be “tuned on” (i.e. activated) or “turned off” (i.e. deactivated)”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0232]: “The instantaneous performance of (fa, gb) on
PNG
media_image1.png
24
121
media_image1.png
Greyscale
can be quantified by the compression rate …
PNG
media_image2.png
35
93
media_image2.png
Greyscale
”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0272]: “The non-linear activation function 1509 can turn on/off different elements in the output from linear part 1508 to achieve better compression performance”).
Regarding claim 28, claim 28 recites the method implemented by the apparatus of claim 13 (see rejection of claim 13 above).
Regarding claim 14, combination of Timo and Song teaches the apparatus of claim 11 (discussed above).
Timo teaches ‘wherein the neural network-based channel state information compression scheme associated with a smaller payload than the codebook-based channel state information compression scheme’ (this is optional);
‘a greater degree of compression than the codebook-based channel state information compression scheme’ (Timo: [0146]: “the second function of the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks”; [0148]: “the first function of the compression function is configured to output a plurality of numbers, and the second function of the compression function is configured to apply a scalar non-linear function to each of the plurality of numbers”; [0150]: “the first function of the compression function is a linear function”; [0078]: “The second function is a non-linear function”; [0272]: “The non-linear activation function 1509 can turn on/off different elements in the output from linear part 1508 to achieve better compression performance”; [0289]-[0309]: “The encoder/decoder weights and encoder/decoder biases can be updated … choosing the closed U and V in the codebook to the trained solution … with autoencoder replaced by a much more sophisticated network (for example, a deep convolutions neural network”; more sophisticated neural-network-based compression would have greater degree of compression with additional non-linear function than codebook-based compression); ‘or both’ (this is optional).
Per claim 15 and 29:
Regarding claim 15, combination of Timo and Song teaches the apparatus of claim 11 (discussed above).
Timo teaches ‘train an encoder and a decoder associated with the neural network-based channel state information compression scheme’ (Timo: [0211]-[0219]: “the CSC framework can be used … While connected to the BS, the BS and/or UE can evaluate the compression performance and, if needed, train and update the configuration of the CS compressor and decompressor”; [0220]-[0267]: “The 3GPP standard can be modified to define a class of channel state compression (CSC) encoders … Updating CSC Parameters”);
‘wherein the first message indicates a configuration of the trained encoder and a configuration of the trained decoder’ (Timo: [0142]: “different compression functions may be suitable for compression of the downlink channel estimates, for example depending on factors … the network node may determine which compression function to be employed at a terminal device, and may signal this to the terminal device via the first set of parameters”; [0131]: “the compression function is formed based on the first set of parameters, these parameters may be employed to control which compression function to be used at the terminal device”).
Regarding claim 29, claim 29 recites the method implemented by the apparatus of claim 15 (see rejection of claim 15 above).
Per claim 16 and 30:
Regarding claim 16, combination of Timo and Song teaches the apparatus of claim 11 (discussed above).
Timo teaches ‘train a first decoder associated with the codebook-based channel state information compression scheme and a second decoder associated with the neural network-based channel state information compression scheme’ (Timo: [FIG.15]; [0268]: “a compression function 1501 and a decompression function 1502 provided in the form of a neural network … compression and decompression is based on an autoencoder framework”; [0289]-[0309]: “The encoder/decoder weights and encoder/decoder biases can be updated … choosing the closed U and V in the codebook to the trained solution … with autoencoder replaced by a much more sophisticated network (for example, a deep convolutions neural network”, train a decoder associated with a codebook-based compression and a decoder associated with a neural network based compression);
‘wherein the first message indicates a configuration of the trained first decoder and a configuration of the trained second decoder’ (Timo: [0142]: “different compression functions may be suitable for compression of the downlink channel estimates, for example depending on factors … the network node may determine which compression function to be employed at a terminal device, and may signal this to the terminal device via the first set of parameters”; [0131]: “the compression function is formed based on the first set of parameters, these parameters may be employed to control which compression function to be used at the terminal device”).
Regarding claim 30, claim 30 recites the method implemented by the apparatus of claim 16 (see rejection of claim 16 above).
Claims 5 and 22 are rejected under 35 U.S.C. 103 as being unpatentable over combination of Timo and Song, in view of Joung et al. (“Channel Correlation Modeling and its Application to Massive MIMO Channel Feedback Reduction”), hereinafter “Joung”.
Per claim 5 and 22:
Regarding claim 5, combination of Timo and Song teaches the apparatus of claim 1 (discussed above).
Timo teaches ‘determine a first compression of the channel information based at least in part on the codebook-based channel state information compression scheme’ (Timo: [0195]: “one or more parameters 1103 indicates an objective function for evaluating performance of the compression function”; [0131]: “the compression function is formed based on the first set of parameters, these parameters may be employed to control which compression function to be used at the terminal device”; 0231]-[0238]: “compressor-decompressor-cost function … cost functions are the average square error … codebook of precoders”; mean (average) squared error associated with a codebook based compression function; [0215]: “A cost function (also referred to as an objective function)”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0243]-[0246]: “The CSC parameters (a*,b*) … The parameters (a*,b*) can be tailored to match the rank and expected codebook-based precoding mode”; determine a compression function based on codebook-based CSI);
‘determine a second compression of the channel information based at least in part on the neural network-based channel state information compression scheme’ (Timo: [0195]: “one or more parameters 1103 indicates an objective function for evaluating performance of the compression function”; [0131]: “the compression function is formed based on the first set of parameters, these parameters may be employed to control which compression function to be used at the terminal device”; [0231]-[0234]: “compressor-decompressor-cost function … cost functions are the average square error”; [0268]: “a compression function 1501 and a decompression function 1502 provided in the form of a neural network”, mean (average) squared error associated with neural network-based compression function; [0215]: “A cost function (also referred to as an objective function)”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0243]-[0250]: “The CSC parameters (a*,b*) … The parameters (a*,b*) can also be learned (or, periodically updated) by the BS using a supervised-learning or reinforcement-learning process”; determine a compression function based on neural-network-based CSI);
‘determine, based at least in part on a cross-correlation between the first compression and the second compression, whether to compress the channel information in accordance with the codebook-based channel state information compression scheme, or in accordance with the neural network-based channel state information compression scheme, or both’ (Timo: [0206]: “channel state compression (CSC) framework”; [0243]-[0250]: “The CSC parameters (a*,b*) … The parameters (a*,b*) can be tailored to match the rank and expected codebook-based precoding mode … The parameters (a*,b*) can also be learned (or, periodically updated) by the BS using a supervised-learning or reinforcement-learning process”, accordance with codebook-based compression function or accordance with neural-network-based compression function; [0289]-[0309]: “The encoder/decoder weights and encoder/decoder biases can be updated … choosing the closed U and V in the codebook to the trained solution … with autoencoder replaced by a much more sophisticated network (for example, a deep convolutions neural network”; [0146]: “the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks. An activation function may provide a “threshold” that allows things to be “tuned on” (i.e. activated) or “turned off” (i.e. deactivated)”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0221]: “The CSC encoders can be used by the UEs to compress their raw “explicit” channel matrices”; [0223]: “The cost functions can be used by the BS and/or UEs to measure the performance of a given compressor-decompressor pair”; [0042]-[0043]: “CSI Feedback: Raw Channel Measurements … It is not possible for the UEs to include “raw” downlink channel estimates in its CSI reports—the resulting overhead would simply suffocate the uplink”; [0232]: “The instantaneous performance of (fa, gb) on
PNG
media_image1.png
24
121
media_image1.png
Greyscale
can be quantified by the compression rate …
PNG
media_image2.png
35
93
media_image2.png
Greyscale
”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0272]: “The non-linear activation function 1509 can turn on/off different elements in the output from linear part 1508 to achieve better compression performance”; based on the cost and choose a compression function with better performance on compression rate and accuracy (more similar to the uncompressed CSI), in another word, compare compressed CSI with codebook-based CSI compression function to uncompressed CSI and compare compressed CSI with neural-based CSI compression function to uncompressed CSI, choose the one with better performance on size and similarity to the uncompressed CSI, which implicitly teaches based at least in part on a cross-correlation between the codebook-based compression and the neural network-based compression).
Moreover, Joung in the same field of endeavor teaches exploit the correlation information between CSIs to reduce the feedback amount (Joung: [Page 1, Col 2]: “exploits the correlation information between CSIs to reduce the feedback amount”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Joung’s teaching with that of combination of Timo and Song to choose compression function based at least in part on a cross-correlation between the codebook-based compression and the neural network-based compression in order to reduce the feedback amount (see reference quote in element above).
Regarding claim 22, claim 22 recites the method implemented by the apparatus of claim 5 (see rejection of claim 5 above).
Claims 8 and 25 are rejected under 35 U.S.C. 103 as being unpatentable over combination of Timo and Song, in view of Lee et al. (“Deep CSI Compression and Coordinated Precoding for Multicell Downlink Systems”), hereinafter “Lee”.
Per claim 8 and 25:
Regarding claim 8, combination of Timo and Song teaches the apparatus of claim 1 (discussed above).
Timo teaches ‘determine, based at least in part on a first power consumption associated with the codebook-based channel state information compression scheme’ (Timo: [0049]: “CSI Feedback: 3GPP Implicit Type I and Type II”; [0054]: “The Type I precoder codebooks”; [0057]: “type II feedback consists of explicit feedback and/or codebook-based feedback with higher spatial resolution”; [0255]-[0259]: “To enable the proposed CSC framework, the CSI ReportConfig can be modified to include: an advanced CSI Type II with “compressed explicit channel matrix feedback … If configured for Type-I or Type-II CSI feedback, the ReportConfig contains a CodebookConfig that specifies configuration parameters for Type-1 and Type-II CSI feedback”; codebook-based compression function). However, combination of Timo and Song fails to expressly teach based at least in part on a power consumption;
‘or a second power consumption associated with the neural network-based channel state information compression scheme, or both’ (these are optional);
‘whether to compress the channel information in accordance with the codebook-based channel state information compression scheme, or in accordance with the neural network-based channel state information compression scheme, or both’ (Timo: [0206]: “channel state compression (CSC) framework”; [0243]-[0250]: “The CSC parameters (a*,b*) … The parameters (a*,b*) can be tailored to match the rank and expected codebook-based precoding mode … The parameters (a*,b*) can also be learned (or, periodically updated) by the BS using a supervised-learning or reinforcement-learning process”; accordance with codebook-based compression function or accordance with neural-network-based compression function; [0289]-[0309]: “The encoder/decoder weights and encoder/decoder biases can be updated … choosing the closed U and V in the codebook to the trained solution … with autoencoder replaced by a much more sophisticated network (for example, a deep convolutions neural network”; [0146]: “the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks. An activation function may provide a “threshold” that allows things to be “tuned on” (i.e. activated) or “turned off” (i.e. deactivated)”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0232]: “The instantaneous performance of (fa, gb) on
PNG
media_image1.png
24
121
media_image1.png
Greyscale
can be quantified by the compression rate …
PNG
media_image2.png
35
93
media_image2.png
Greyscale
”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0272]: “The non-linear activation function 1509 can turn on/off different elements in the output from linear part 1508 to achieve better compression performance”).
However, Lee in the same field of endeavor teaches CSI compression with consideration of power consumption (Lee: [Page 1, Col 1-2]: “Two CSI compression techniques … designs were proposed to maximize the energy efficiency, that is, the ratio of the weighted sum rate to the total power consumption”; [Page 4, Col 1]: “Pb to obtain an estimate of the transmit power
PNG
media_image3.png
44
256
media_image3.png
Greyscale
”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Lee’s teaching with that of combination of Timo and Song in order to maximize the energy efficiency (see reference quote in element above).
Regarding claim 25, claim 25 recites the method implemented by the apparatus of claim 8 (see rejection of claim 8 above).
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over combination of Timo and Song, in view of Nagata et al. (US 20150264694 A1), hereinafter “Nagata”.
Regarding claim 9, combination of Timo and Song teaches the apparatus of claim 1 (discussed above).
Timo teaches ‘determine, based at least in part on a first processing load associated with the codebook-based channel state information compression scheme’ (Timo: [0049]: “CSI Feedback: 3GPP Implicit Type I and Type II”; [0054]: “The Type I precoder codebooks”; [0057]: “type II feedback consists of explicit feedback and/or codebook-based feedback with higher spatial resolution”; [0255]-[0259]: “To enable the proposed CSC framework, the CSI ReportConfig can be modified to include: an advanced CSI Type II with “compressed explicit channel matrix feedback … If configured for Type-I or Type-II CSI feedback, the ReportConfig contains a CodebookConfig that specifies configuration parameters for Type-1 and Type-II CSI feedback”; codebook-based compression function). However, combination of Timo and Song fails to expressly teach based at least in part on a processing load;
‘or a second processing load associated with the neural network-based channel state information compression scheme, or both’ (these are optional);
‘whether to compress the channel information in accordance with the codebook- based channel state information compression scheme, or in accordance with the neural network-based channel state information compression scheme, or both’ (Timo: [0206]: “channel state compression (CSC) framework”; [0243]-[0250]: “The CSC parameters (a*,b*) … The parameters (a*,b*) can be tailored to match the rank and expected codebook-based precoding mode … The parameters (a*,b*) can also be learned (or, periodically updated) by the BS using a supervised-learning or reinforcement-learning process”; accordance with codebook-based compression function or accordance with neural-network-based compression function; [0289]-[0309]: “The encoder/decoder weights and encoder/decoder biases can be updated … choosing the closed U and V in the codebook to the trained solution … with autoencoder replaced by a much more sophisticated network (for example, a deep convolutions neural network”; [0146]: “the compression function and/or the second function of the decompression function may comprise a non-linear activation function. Activation functions are often employed in machine learning, such as neural networks. An activation function may provide a “threshold” that allows things to be “tuned on” (i.e. activated) or “turned off” (i.e. deactivated)”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0131]: “The second function introduces nonlinearities into the compression function that allow the downlink channel estimates to be better approximated”; [0232]: “The instantaneous performance of (fa, gb) on
PNG
media_image1.png
24
121
media_image1.png
Greyscale
can be quantified by the compression rate …
PNG
media_image2.png
35
93
media_image2.png
Greyscale
”; [0196]: “The objective function may for example be a cost function or a loss function for evaluating whether downlink channel estimates after compression and decompression are similar to the original downlink channel estimates”; [0272]: “The non-linear activation function 1509 can turn on/off different elements in the output from linear part 1508 to achieve better compression performance”).
However, Nagata in the same field of endeavor teaches calculate CSI with consideration of processing load (Nagata: [0037]: “to lighten the processing load for calculating a plurality of kinds of CSI”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Nagata’s teaching with that of combination of Timo and Song in order to lighten the processing load for calculating CSI (see reference quotes in element above).
Conclusion
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 GUOXING FAN whose telephone number is (703)756-1310. The examiner can normally be reached Monday - Friday 8:30am - 5:30pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Yemane Mesfin can be reached at (571)272-3927. 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.
/G.F./Examiner, Art Unit 2462
/YEMANE MESFIN/Supervisory Patent Examiner, Art Unit 2462