DETAILED ACTION
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 of this title, 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 1, 4, 5, 7, and 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over Arfaoui et al. (US 2026/0089515, “Arfaoui”) in view of Sangdeh et al. (US 2025/0038811, “Sangdeh”).
Regarding claim 1, Arfaoui discloses an apparatus for wireless communication, comprising:
- at least one memory; and at least one processor coupled with the at least one memory and configured to cause the apparatus (See Fig.1B and ¶.34, a memory and a processor), comprising
- an encoder to receive input data and transmit encoded output data from the encoder (See Fig.2, AI/ML encoder in WTRU Tx; See 303, 304, & 305 Fig.3A, UE receives inputs from NW) to a decoder comprised in a second apparatus to (See Fig.2, ‘AI/ML decoder in gNB Rx; See ¶.5, the WTRU may receive a CSI reference signal (RS). The WTRU may generate (e.g., using an AI or ML encoder) and send a CSI feedback report to the network entity based on the CSI RS. The NW may construct a downlink CSI estimate based on the CSI feedback report (e.g., using an AI or ML decoder)):
- provide the input data to the encoder to generate the encoded output data (See 305 Fig.3A, WTRU receives CSI-RS, generates the CSI feedback report and transmits the CSI feedback report to the NW),
- wherein a size of the encoded output data is dependent on the input data provided to the encoder (See ¶.79, SF CSI compression includes the operation of compressing the CSI estimates in the spatial/frequency (SF) domain by the WTRU to a quantized-binary representation with a predefined feedback size and transmitting it to the NW. The NW, in turn, may reconstruct the SF CSI estimates by decompressing the received CSI feedback from the WTRU; See ¶.83, TSF CSI compression refers to generating a quantized-binary representation of the CSI with a predefined feedback size based on the current and the prior SF CSI estimates, and then transmitting it to the NW. The NW, in turn, can reconstruct the current SF CSI estimate by jointly incorporating the reconstructed prior SF CSI estimates and the received CSI feedback from the WTRU; See ¶.85, joint CSI compression and prediction may comprise generating a quantized-binary representation with a predefined feedback size based on the current and the prior SF CSI estimates, and then transmitting it to the NW; See 303 Fig.3A and ¶.103, at 303, the WTRU may receive an indication that a training session of a CSI feedback function is starting); and
- transmit the encoded output data to the decoder (See 305 Fig.3A, transmits the CSI feedback report to the NW).
Arfaoui describes the method of providing and transmitting and the examiner provides a secondary prior art by Sangdeh that shows the whole system comprising an encoder, a latent representation, and a decoder (Sangdeh, See Fig.7 and ¶.61,
PNG
media_image1.png
497
771
media_image1.png
Greyscale
).
Regarding claim 4, Arfaoui discloses “the encoder comprises at least one neural network model and wherein the at least one processor is configured to cause the apparatus to receive information associated with the at least one neural network model (See ¶.9, the WTRU and/or the NW may update one or more AI or ML models for CSI feedback functions (e.g., an encoder and/or a decoder) based on the computed/received error metric or statistic determined by the WTRU. In examples, the WTRU and/or NW may perform backpropagation and/or a learning algorithm, such as a gradient descent algorithm, when updating the AI or ML models; See further ¶.88-89), the information comprising a structure or weights of the at least one neural network model, information regarding a set of neural network models from a plurality of sets of neural network models, a set of neural network models based on an indication or configuration, or a combination thereof (See Fig.4 and ¶.151, FIG. 4 presents an example of the CsiNet model architecture. CsiNet is a well-known AI/ML model structure comprising an encoder (the WTRU AI/ML model) and a decoder (the NW AI/ML model). Each of the encoder and the decoder comprise convolutional layers, batch normalization layers, leaky ReLU layers, Sigmoid activation layers, and residual connections; See ¶.73 for Deep Neural Networks (DNNs); See further ¶.87-88 for AI/ML models; See ¶.98-99 for AI/ML models parameters).”
Regarding claim 5, Arfaoui discloses “the input data is based on channel data and comprises a representation of a measured channel matrix of a radio channel, a precoder for the radio channel, or a combination thereof (See ¶.76, measurements of channel; See ¶.74, latent representation vector; See ¶.87, a precoder).”
Regarding claim 7, Arfaoui discloses “the encoder comprises a neural network-based latent-generator (See ¶.74, latent vector using a deep neural networks (DNN)).”
Regarding claim 18, it is a processor of an apparatus claim corresponding to the claim 1 and is therefore rejected for the similar reasons set forth in the rejection of the claim.
Regarding claim 19, it is a method claim corresponding to the claim 1 and is therefore rejected for the similar reasons set forth in the rejection of the claim.
Regarding claim 20, it is an apparatus claim at performed at a network corresponding to the claim 1 at performed at a UE and is therefore rejected for the similar reasons set forth in the rejection of the claim.
Claims 2, 3, and 13-15 are rejected under 35 U.S.C. 103 as being unpatentable over Arfaoui in view of Sangdeh and further in view of Jeon (US 2024/0097764, “Jeon”).
Regarding claim 2, Arfaoui discloses “cause the apparatus to transmit decoder assistance information to the decoder, the decoder assistance information indicating the size of the encoded output data (Arfaoui, See ¶.79, SF CSI compression includes the operation of compressing the CSI estimates in the spatial/frequency (SF) domain by the WTRU to a quantized-binary representation with a predefined feedback size and transmitting it to the NW. The NW, in turn, may reconstruct the SF CSI estimates by decompressing the received CSI feedback from the WTRU).”
Arfaoui, further, discloses that “WTRU transmits AI/ML-based CSI feedback capabilities indication to the network” (Arfaoui, See 302 Fig.3A), but Arfaoui and Sangdeh do not explicitly disclose what Jeon discloses “the size of the encoded output data (Jeon, See ¶.108-109 and ¶.114, UE capability information includes an indication such as CSI encoded output vector size).”
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to apply the method of “the size of the encoded output data” as taught by Jeon into the system of Arfaoui and Sangdeh, so that it provides a way of generating different output vector sizes (Jeon, See ¶.114).
Regarding claim 3, Arfaoui and Jeon disclose “cause the apparatus to transmit the encoded output data in a first message (Arfaoui, See ¶.4, the WTRU transmits an indication that the WTRU supports end-to-end learning of AI or ML CSI feedback functions; See ¶.5, the WTRU may generate (e.g., using an AI or ML encoder) and send a CSI feedback report to the network entity based on the CSI RS) and the decoder assistance information in a second message (Arfaoui, See 302 Fig.3A, WTRU transmits CSI feedback capabilities and Jeon discloses that the capability information includes the encoder output size as rejected in claim 2). Therefore, this claim is rejected with the similar reasons and motivation set forth in the rejection of claim 2.
Regarding claim 13, Arfaoui discloses “the encoded output data is further based on a scaler quantizer, a vector quantizer, or a combination thereof (Arfaoui, See ¶.74, a lower dimensional latent vector; See ¶.79, a quantized-binary representation),” but Arfaoui and Sangdeh do not explicitly disclose what Jeon discloses the limitation “a scaler quantizer, a vector quantizer (Jeon, See ¶.108, the UE capability information includes an indication on the supported quantization method for the AI/ML-based CSI encoder output such as vector, scalar, uniform, and/or non-uniform quantization).” Therefore, this claim is rejected with the similar reasons and motivation set forth in the rejection of claim 2.
Regarding claim 14, Arfaoui and Sangdeh do not explicitly disclose what Jeon discloses “the at least one processor is configured to cause the apparatus to receive quantizer information defining parameters of the scaler quantizer, the vector quantizer, or a combination thereof (Jeon, See ¶.109, the UE is provided, from the network, an AI/ML model for CSI feedback, e.g., via model ID, model transfer, or model description including neural network structure, parameter values, and feedback format such as the CSI encoder output vector size, quantization method, and the number of bits for quantization; See ¶.114, a UE can be provided by the network the AI/ML-based CSI encoder output vector size, quantization method, and the quantization resolution, e.g., resulting total number of bits after quantization, quantization bits per element in the CSI encoder output vector, or the ratio of output vector size to the total resulting quantized number of bits).” Therefore, this claim is rejected with the similar reasons and motivation set forth in the rejection of claim 2.
Regarding claim 15, Arfaoui and Sangdeh do not explicitly disclose what Jeon discloses “the at least one processor is configured to cause the apparatus to determine the size of the encoded output data by adjusting the parameters of the scaler quantizer or by selection of a subset of a codebook or codewords of a vector quantization scheme (Jeon, See ¶.102, the UE may have a trained CSI encoder model for one channel environment and retrain the model for different channel environment by finetuning the parameter values of the last output layer with the same or different output vector size from the previously trained CSI model while fixing the structure and parameter values for the rest of the layers).” Therefore, this claim is rejected with the similar reasons and motivation set forth in the rejection of claim 2.
Claims 6, 16, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Arfaoui in view of Sangdeh and further in view of Chai et al. (US 2025/0379628, “Chai”).
Regarding claim 6, Arfaoui and Chai disclose “the at least one processor is provided with configuration parameters associated with the encoder, the configuration parameters comprising a threshold parameter for a size of the encoded output data (Arfaoui, See ¶.79, a quantized-binary representation with a predefined feedback size), a target size of the encoded output data, a target accuracy of the encoded output data, an indication of an importance of an accuracy of the encoded output data, an overhead cost, or a combination thereof (Chai, See ¶.172, a size of the CSI reports to be transmitted exceeds a specified threshold).” Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to apply “a threshold parameter for a size of the encoded output data” as taught by Chai into the system of Arfaoui and Sangdeh, so that it provides a way of performing transmission of the CSI report according to the specified threshold (Chai, See ¶.468).
Regarding claim 16, Arfaoui and Sangdeh do not explicitly disclose what Chai discloses “the encoder is trained to minimize a loss function (Chai, See ¶.147, a process of minimizing the loss to make a value of the loss function), the loss function based on a weighted sum of at least one component comprising (Chai, See ¶.147, a weight vector is adjust; See ¶.220, weighting coefficients and quantization precision of a weighting coefficient): mutual information between the input data and the encoded output data; negative of mutual information between the encoded output data and an expected target output; and difference between output of the encoder and the expected target output (See .147, the loss function and the objective function are important equations that measure the difference between the predicted value and the target value. The loss function is used as an example. A larger output value (loss) of the loss function indicates a greater difference. In this case, the training for the AI model is a process of minimizing the loss, to make a value of the loss function less than a threshold, or make a value of the loss function satisfy a target requirement).” Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to apply the method of “the encoder is trained to minimize a loss function, the loss function based on a weighted sum of at least one component comprising: mutual information between the input data and the encoded output data; negative of mutual information between the encoded output data and an expected target output; and difference between output of the encoder and the expected target output” as taught by Chai into the system of Arfaoui and Sangdeh, so that it provides a way for a weight vector of each layer of the AI model to be updated based on a difference between the predicted value and the target value (Chai, See ¶.147).
Regarding claim 17, Arfaoui discloses “weights of the at least one component of the loss function are configurable, with a minimum weight being zero (See ¶.93, the WTRU may employ the estimated DL effective channel to equalize the received DL data using a certain equalizer (e.g., zero-forcing (ZF), minimum mean square error (MMSE), etc.) to estimate the received training data, which are denoted by Xe; See ¶.172, Table #2, Proposed: End-to-End AI/ML End-to-end learning of the AI/ML Learning (AI/ML − model where the AI/ML model at CSF) the NW reconstructs the DL CSI and then the NW determines the DL precoder based on the reconstructed CSI (e.g., zero-forcing precoding).”
Claims 8-11 are rejected under 35 U.S.C. 103 as being unpatentable over Arfaoui in view of Sangdeh and further in view of Agrawal et al. (US 2021/0142158, “Agrawal”).
Regarding claim 8, Arfaoui discloses “the encoded output data is generated based on a latent representation of the input data generated using the neural network-based latent-generator (Arfaoui, See ¶.74, a lower dimensional latent vector by using latent representation), but Arfaoui and Sangdeh do not explicitly disclose what Agrawal discloses “masking information (Agrawal, See Fig.1, ¶.138, and ¶.177, a set of masking {0, 1}k in an Encoder side).” Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to apply the method of “masking information” as taught by Agrawal into the system of Arfaoui and Sangdeh, so that it provides a way for loss function to penalize the weights based on the probability of the target bit (Agrawal, See ¶.138).
Regarding claim 9, Arfaoui and Sangdeh do not explicitly disclose what Agrawal discloses “wherein the encoder comprises a neural network-based mask generator and the masking information comprises a masking vector that is generated by providing the input data to the neural network-based mask generator (Agrawal, See Fig.1 and ¶.138, a target vector by using masking information; See Fig.4, an input layer of the NN).” Therefore, this claim is rejected with the similar reasons and motivation set forth in the rejection of claim 8.
Regarding claim 10, Arfaoui and Sangdeh do not explicitly disclose what Agrawal discloses “wherein the masking information relates to entries of the latent representation of the input data that are selected to generate the encoded output data (Agrawal, See ¶.15, optimizing trainable parameters of the NN to minimize a loss function may comprise minimizing the loss function calculated on the basis of at least one intermediate output representation selected from the set comprising intermediate output representations available at layers up to and including the layer at which the syndrome check is satisfied).” Therefore, this claim is rejected with the similar reasons and motivation set forth in the rejection of claim 8.
Regarding claim 11, Arfaoui discloses “wherein the masking information indicates a number of entries of the latent representation of the input data that are used to generate the encoded output data (See ¶.171, Table #1, ‘Number of latent’), but does not explicitly disclose what Sangdeh discloses the limitations “a number of entries of the latent representation” (Sangdeh, See Fig. 7 and ¶.61, the output of the ML-based encoder 714 is a latent representation of the compressed DL CSI. This latent representation may a vector with a predefined dimension, typically much shorter than the input CSI. For example, if the input to the ML-based encoder is a 32 by 32 matrix of complex numbers, the output latent representation could be a vector with 28, 52, or 56 elements. This compression reduces the amount of data that needs to be transmitted from the UE to the base station 702).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to apply the method of “a number of entries of the latent representation” as taught by Sangdeh into the system of Arfaoui, so that it provides a way of reducing the amount of data that needs to be transmitted from the UE to the base station (Sangdeh, See ¶.61).
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over Arfaoui in view of Sangdeh and Agrawal and further in view of LI et al. (US 2024/0030941, “LI”).
Regarding claim 12, Arfaoui, Sangdeh, and Agrawal do not explicitly disclose what LI discloses “wherein the at least one processor is configured to cause the apparatus to transmit decoder assistance information comprising an indication of the encoded output data that are selected from a latent representation of the input data (LI, See ¶.132, selection is performed based on row weights of generator matrices, that is, information bits and fixed bits are selected based on the row weights of corresponding generator matrices for the to-be-encoded bits. The row weights of the generator matrices corresponding to the information bits are greater than or equal to the row weights of the generator matrices corresponding to the fixed bits; See ¶.133, if K information bits need to be selected from N to-be-encoded bits, after the to-be-encoded bits are sorted in descending order).” Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to apply the method of “an indication of the encoded output data that are selected from a latent representation of the input data” as taught by LI into the system of Arfaoui, Sangdeh, and Agrawal, so that it provides a way of selecting the to-be-encoded bits (LI, See ¶.27).
Contact Information
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Jung H Park whose telephone number is 571-272-8565. The examiner can normally be reached M-F: 7:00 AM-3:00 PM.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Derrick Ferris can be reached on 571-272-3123. 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.
/JUNG H PARK/
Primary Examiner, Art Unit 2411
Prosecution Insights