DETAILED ACTION
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 .
Election/Restrictions
Applicant’s election without traverse of Group I (Claims 1-9 and 17-20) in the reply filed on 05/08/2026 is acknowledged. Claims 10-16 are cancelled.
Oath/Declaration
The receipt of oath/declaration is acknowledged.
Drawings
The drawings were received on 01/29/2024. These drawings are reviewed and accepted by the Examiner.
Specification
The lengthy specification has not been checked to the extent necessary to determine the presence of all possible minor errors. Applicant’s cooperation is requested in correcting any errors of which applicant may become aware in the specification.
Information Disclosure Statement
The information disclosure statement (IDS), submitted on 02/03/2025, is in compliance with the provisions of 37 CRR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
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.
The factual inquiries set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied 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.
Claims 1-2, 4, 6-7, 17-19 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over anticipated Pourahmadi et al. (U.S 2026/0046642; hereinafter “Pourahmadi” cited with provisional 63/396,596) in view of NPL- Vivo; "Evaluation on AI/ML for CSI feedback enhancement"; 3GPP TSG RAN WG1 #111; R1-2210998; hereinafter “Vivo- R1-2210998”; November 14, 2022.
For citation purposes, hereinafter, the Office Action refers to the cited by Pourahmadi et al. with provisional 63/396,596, which qualifies as prior art date.
For claim 1:
Pourahmadi discloses a baseband processor configured to, when executing instructions stored in a memory (see Pourahmadi, at least paragraph [0271]; the processor, the memory, the transceiver, or various combinations or components thereof may be implemented in hardware), perform operations comprising:
receiving a set of encoder outputs from an artificial intelligence (AI)-based encoder that generates compressed CSI feedback (see Pourahmadi, at least paragraph [0272]; [0236]; the processor 804 may be configured as and/or otherwise support a means to receive a set of parameters from a second apparatus, the set of parameters including a set of encoder parameters for an encoder of a two-sided model and the techniques discussed herein provide an AI-based CSI feedback mechanism that provides a channel-matching precoder under different channel conditions and [0029]; provides for techniques that support performance monitoring of a two-sided model. For instance, implementations provide an architecture and associated signaling for compressing an input (e.g., 3-dimensional (3D) input, CSI at a UE, etc.), quantizing the compressed input, transmitting the quantized compressed input); and
based on the set of encoder outputs, optimizing a per-segment vector quantization (VQ) codebook for use in quantizing respective segments of encoder outputs (see Pourahmadi, at least paragraph [0062]; [0104]-[0105]; training a single model with inputs having different statistics may result in a model with average and sub-optimal performance over different UE types),
Pourahmadi, does not explicitly disclose wherein each segment of encoder outputs comprises a subset of the set of encoder outputs, wherein a number of inputs of the VQ codebook and a number of outputs of the VQ codebook are based on a number of bits configured for uplink channel information (UCI) and a number of segments in the set of encoder outputs.
Vivo-R1-2210998, from the same or similar fields of endeavor, discloses what Pourahmadi fails: AI models with different output dimensions also need to be trained independently. To generalize different payload without training a new AI model, we use payload truncation for different payload so the length of encoder output can be fixed (see Vivo-R1-2210998, at least section 2, 2.3.2, figure 11) and proposal 6; CSI payload truncation for the generalization of UCI payload; section 2.5; CSI feedback quantization and figure15, observation 21-22; vector quantization with optimized codebook).
Therefore, it would have been obvious statement before the effective filing date of the claimed invention to have a system comprises a method as taught by Vivo-R1-2210998. The motivation for doing this is to provide a system networks can provide possible benefits where in a reasonable span of decoder input size, one common encoder can be utilized and corresponds to serval decoders based on payioad truncation to save the overhead of AI model transmission and switching complexity.
For claims 2 and 18:
In addition to rejection in claim 2, Pourahmadi- Vivo-R1-2210998 further discloses wherein the operations comprise receiving configuration of the number of inputs of the VQ codebook, the number of output bits of the VQ codebook, or the number of segments (see Pourahmadi, at least paragraph [0116]; [0154]; c) a number of vectors that will be quantized using the quantization codebook 710 ( e.g. the output 714), d) the number of bits used to select a quantization codebook codeword).
For claim 4:
In addition to rejection in claim 4, Pourahmadi- Vivo-R1-2210998 further discloses wherein the operations comprise de-quantizing respective segments of UCI encoding compressed CSI feedback based on the VQ codebook; combining the de-quantized segments of the UCI to generate estimated encoder output values; and decoding the estimated encoder output values to re-construct CSI feedback (see Pourahmadi, at least paragraph [0110]; the network subsystem 700b can be trained to use the bits received from the UE subsystem 700a (e.g., feedback CSI bits such as those corresponding to the two latent representations) to generate a desired output. In at least some examples, a training objective is to have the output data (e.g., reconstructed data) as similar as possible to the input data).
For claim 6:
In addition to rejection in claim 6, Pourahmadi- Vivo-R1-2210998 further discloses wherein the operations comprise re-training a trained AI-based encoder, a trained AI-based decoder, or both, based on the optimized VQ codebook (see Vivo-R1-2210998, at least figure11, section 2, section 2.3.2; AI models with different output dimensions also need to be trained independently. To generalize different payload without training a new AI model, we use payload truncation for different payload of the length of encoder output can be fixed and see observation 22; Vector· quantization with optimized codebook can achieve slightly better performance).
For claim 7:
In addition to rejection in claim 7, Pourahmadi- Vivo-R1-2210998 further discloses wherein the operations comprise re-training the trained AI-based encoder or the trained AI-based decoder using a loss function that includes a first loss term that optimizes encoder weights toward the optimized VQ codebook (see Vivo-R1-2210998, at least section 2.3.2; the Al model is trained, the loss function is set to include the correlation of all decoder output and a weight for each decoder and accumulate the correla1ion of each decoder output with the weight as a total correlation, The weight is trained with the decoder. We choose four different payload and use the dedicated model for each payload as baseline, We train the joint encoder with different c:ombination of payloads, Por each payload combination, only the decoder corresponding lo the given payload is used).
For claim 17:
Pourahmadi discloses a network device, comprising: a memory; and a baseband processor coupled to the memory, the processor configured to, when executing instructions stored in the memory (see Pourahmadi, at least paragraph [0271]; the processor 804, the memory 806, the transceiver 808, or various combinations or components thereof may be implemented in hardware), cause the network device to: receive a set of encoder outputs from an artificial intelligence (AI)-based encoder that generates compressed CSI feedback (see Pourahmadi, at least figure 11, paragraph [0272]; [0236]; the processor 804 may be configured as and/or otherwise support a means to receive a set of parameters from a second apparatus, the set of parameters including a set of encoder parameters for an encoder of a two-sided model and the techniques discussed herein provide an AI-based CSI feedback mechanism that provides a channel-matching precoder under different channel conditions and [0029]; provides for techniques that support performance monitoring of a two-sided model. For instance, implementations provide an architecture and associated signaling for compressing an input (e.g., 3-dimensional (3D) input, CSI at a UE, etc.), quantizing the compressed input, transmitting the quantized compressed input); based on the set of encoder outputs, optimize a per-segment vector quantization (VQ) codebook for use in quantizing respective segments of encoder outputs (see Pourahmadi, at least paragraph [0062]; [0104]-[0105]; training a single model with inputs having different statistics may result in a model with average and sub-optimal performance over different UE types),
encode the optimized VQ codebook for transmission to another network device using a physical uplink shared channel (PUSCH) or a physical downlink shared channel (PDSCH) (see Pourahmadi, at least paragraph [0185]; [0207]-[0208]; the UE can feedback information which could help the network node adjust/update the AI/ML model parameters if needed. The feedback can be sent using configured grant PUSCH transmissions and [0203]; the network signals the set of parameters as part of an AI based report over at least one of PUSCH,).
Pourahmadi, does not explicitly disclose wherein each segment of encoder outputs comprises a subset of the set of encoder outputs, wherein a number of inputs of the VQ codebook and a number of outputs of the VQ codebook are based on a number of bits configured for uplink channel information (UCI) and a number of segments in the set of encoder outputs; and
Vivo-R1-2210998, from the same or similar fields of endeavor, discloses what Pourahmadi fails: AI models with different output dimensions also need to be trained independently. To generalize different payload without training a new AI model, we use payload truncation for different payload so the length of encoder output can be fixed (see Vivo-R1-2210998, at least section 2, 2.3.2, figure 11) and proposal 6; CSI payload truncation for the generalization of UCI payload; section 2.5; CSI feedback quantization and figure15, observation 21-22; vector quantization with optimized codebook).
Therefore, it would have been obvious statement before the effective filing date of the claimed invention to have a system comprises a method as taught by Vivo-R1-2210998. The motivation for doing this is to provide a system networks can provide possible benefits where in a reasonable span of decoder input size, one common encoder can be utilized and corresponds to serval decoders based on payload truncation to save the overhead of AI model transmission and switching complexity.
For claim 19:
In addition to rejection in claim 19, Pourahmadi- Vivo-R1-2210998 further discloses wherein the baseband processor is configured to cause the network device to train an AI-based encoder neural network (NN) in the AI-based encoder based on a training data set (see Pourahmadi, at least paragraph [0010]; [0014]; [0032]-[0033]; at least one of a structure of a neural network of the set of neural network models, or a weight of a neural network of the set of neural network models and [0112]-[0113]; input data 702 is input to a neural network 704); generate a decoder data set by inputting the training data set to the trained NN (see Pourahmadi, at least paragraph [0128]-[0130]; [0207]; neural network weights are initialized randomly for training. The neural network weights can be changed during the training phase in a way that reduces the loss function; encode the decoder data set for transmission; and cause transmission of the VQ codebook and the decoder data set to a base station (see Pourahmadi, at least paragraph [0014]; a set of encoder parameters for an encoder of a two-sided model and [0102]; transmitted to the gNB side. In a related proposal, a vector quantization scheme is presented using neural networks where the prior is learnt from the data rather than being static).
For claim 20:
In addition to rejection in claim 20, Pourahmadi- Vivo-R1-2210998 further discloses wherein the baseband processor is configured to cause the network device to train an AI-based decoder NN in the AI-based decoder based on a decoder data set (see Pourahmadi, at least paragraph [0032]-[0033]; [0274]; the set of parameters include at least one set of decoder parameters for a decoder of the two-sided model; the at least one model metric is based at least in part on the decoder parameters); generate an encoder data set generated by the trained AI-based decoder NN; encode the encoder data set for transmission; and cause transmission of the VQ codebook and the encoder data set to a UE (see Pourahmadi, at least paragraph [0172]-[0173]; [0207]; AI/ML model training is performed at the UE side and shared with the network. The UE signals a set of parameters corresponding to multiple AI/ML models, e.g., decoder functions of a CSI auto-encoder, where the set of parameters are partitioned into multiple subsets of parameters, each subset of parameters is associated with an AI/ML model of the multiple AI/ML models).
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over anticipated Pourahmadi et al. (U.S 2026/0046642; hereinafter “Pourahmadi” cited with provisional 63/396,596) in view of NPL- Vivo; "Evaluation on AI/ML for CSI feedback enhancement"; 3GPP TSG RAN WG1 #111; R1-2210998; hereinafter “Vivo- R1-2210998”; November 14, 2022 further in view of Ramo et al. AU-2006286177-C1.
For citation purposes, hereinafter, the Office Action refers to the cited by Pourahmadi et al. with provisional 63/396,596, which qualifies as prior art date.
For claim 5:
In addition to rejection in claim 5, Pourahmadi- Vivo-R1-2210998 does not explicitly disclose wherein the set of encoder outputs correspond to inference outputs of a trained AI-based encoder, wherein the processor is configured to optimize the VQ codebook using a Linde-Buzo-Gray (LBG) based algorithm.
Ramo, from the same or similar fields of endeavor, further discloses what Pourahmadi- Vivo-R1-2210998 fails: Linde, Buzo, and Gray (LBG) proposed the so called LBG algorithm for generating codebooks based on 25 a training sequence of vectors (see Linde, Y., Buzo, A. and Gray, R. M., "An algorithm or Vector Quantization", which a respective codebook has to be trained, and stored at both the quantizer and a unit that is used to retrieve the reproduction vectors from 10 identifiers of the reproduction vectors (see Ramo, at least page 3)
Therefore, it would have been obvious statement before the effective filing date of the claimed invention to have a system comprises a method as taught by Ramo. The motivation for doing this is to provide a system networks in order to allows for a particularly flexible choice of 10 the number of quantization levels.
Allowable Subject Matter
Claims 3, 7 and 8 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in all independents form including all of the limitations of the base claim and any intervening claims and if rewritten or amended to overcome any objection claims set forth in this Office action.
Conclusion
The prior arts made or record and not relied upon are considered pertinent to applicant's disclosures. Gu (U.S 20250379796), discloses a basic model of auto-encoder is shown as follows. The encoder compressed the raw CSI-RS values (in short, raw CSI)/maximum Eigen vector and reports its output to the gNB. The gNB will decompress it. A new CSI report is the CSI report that contains the enhanced CSI feedback by an AI/ML model.
Hindy et al. (U.S 2026/0040120), discloses a method transmit, to the network entity, a third signaling indicating at least one parameter corresponding to the selection of the AI model from the multiple AI models. Additionally, or alternatively, the third signaling includes at least one of a CSI report or an AI-based report, and wherein the third signaling is transmitted over multiple time units.
Kim et al. (U.S 2016/0093311), discloses Vector quantization (VQ) of a vector that does not depend on past quantized vectors stored in memory of an
encoder or decoder from a previous time segment ( e.g. a frame), and The weight determination unit 572 may represent a unit configured to determine the 32 weights 503 ( or another number of a plurality of weights 503) for a current time
segment (e.g., an ith audio frame) corresponding to the 32 volume AE vectors 501 defined in a higher order ambisonics domain and indicative of the input V-vector 55(i)).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LAN HUONG TRUONG whose telephone number is (571)270-5829. The examiner can normally be reached on Mon-Fri (7:30AM-5:00PM).
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, RICKY NGO can be reached on 571-272-3139. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov/. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/Lan-Huong Truong/
Primary Examiner, Art Unit: 2464
07/28/2026
Prosecution Insights