Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 09/11/2025 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Objections
Claim 15 is objected to because of the following informalities:
In claim 15, “ReLU” should be “Rectified Linear Unit (ReLU)”.
Appropriate correction is required.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
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.
Claims 1-11, 13, 17-20, and 22 are rejected under U.S.C. 103 as being unpatentable over Choukroun et al. (“Error Correction Code Transformer” published 03/27/2022) in view of Bennatan et al. (“Deep Learning for Decoding Linear Codes – A Syndrome-Based Approach” published 02/13/2018).
As per claims 1 and 22:
Choukroun teaches a method and a system for decoding signals encoded with error correction codes (Choukroun pg. 1, transformer architecture to the soft decoding of linear codes), comprising:
a memory storing computer-readable instructions; and
at least one processor configured to execute the instructions to (Choukroun pg. 5 section 4.3, use of an Adam optimizer. Please note running an Adam optimizer requires the use of a processor and memory):
input a first error correction code, comprising a first parity check matrix (Choukroun pg. 2 section 3.1, the linear code is defined by a binary parity check matrix H), into a pre-trained model having a transformer architecture (Choukroun pg. 1 section 1, employ a model free decoder built upon Transformer architecture), the pre-trained model having been trained on a plurality of error correction codes (Choukroun pg. 6 section 5, the proposed architecture is trained with LDPC codes, Polar codes, and BCH codes);
incorporate relative position information into the high-dimensional representation of the first error correction code based on the first parity check matrix (Choukroun pg. 5 section 4.2, build a symmetric mask such that it contains information about every pairwise bit relations); and
predict a noise estimate for decoding based on the first parity check matrix and the high-dimensional representation of the first error correction code (Choukroun pg. 5 section 4.3, training objective is to predict the multiplicative noise).
Choukroun does not explicitly disclose to generate a position-invariant high-dimensional representation of the first error correction code based on the pre-trained model. However, Bennatan discloses to generate a position-invariant high-dimensional representation of the first error correction code based on the pre-trained model (Bennatan pg. 3 section IV, “For any choice of F, the decoder’s BER and MSE, conditioned on transmission of any codeword x, are both invariant to x.” Please note codeword invariance is interpreted as position invariance by discarding the absolute bit positions).
Choukroun and Bennatan are analogous arts because they are in the same field of endeavor of neural decoding. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Bennatan’s position invariance with the method and system of Choukroun. This modification would have been obvious to one of ordinary skill in the art at the time of filing because it guarantees the generalization of performance obtained during training (Bennatan pg. 1 abstract).
As per claim 2:
Choukroun and Bennatan further teach the system according to claim 1, wherein generating a position-invariant high-dimensional representation of the first error correction code comprises applying a code-invariant initial embedding based on the pre-trained model to a representation derived from the first parity check matrix (Bennatan pg. 3 section IV, codeword invariance).
As per claim 3:
Choukroun and Bennatan further teach the system according to claim 1, wherein incorporating relative position information into the high-dimensional representation of the first error correction code comprises: constructing a Tanner graph from the first parity check matrix (Choukroun pg. 5 Fig. 2, induced Tanner graph); computing a first distance matrix from the Tanner graph (Choukroun pg. 5 section 4.2, build an adjacency matrix of the Tanner graph); and modulating the pre-trained model’s self-attention map with the first distance matrix (Choukroun pg. 5 section 4.2, masked self-attention with the adjacency matrix).
As per claim 4:
Choukroun and Bennatan further teach the system according to claim 1, wherein predicting a noise estimate for decoding comprises applying a size-invariant transformation informed by the first parity check matrix to the refined high-dimensional representation of the first error correction code (Choukroun pg. 1 abstract, soft decoding of linear codes at arbitrary block lengths).
As per claim 5:
Choukroun and Bennatan further teach the system according to claim 1, wherein the at least one processor is further configured to execute the instructions to: receive a signal encoded with the first error correction code; and decode the received signal, thereby generating a decoded output, wherein decoding the received signal comprises applying the noise prediction to the received signal (Choukroun pg. 2-3 section 3.1, receive channel output and decode to provide a soft approximation).
As per claim 6:
Choukroun and Bennatan further teach the system according to claim 2, wherein the code-invariant initial embedding is configured to be length-invariant (Choukroun pg. 1 abstract, soft decoding of linear codes at arbitrary block lengths).
As per claim 7:
Choukroun and Bennatan further teach the system according to claim 4, wherein the size-invariant transformation is pre-trained on the plurality of error correction codes, each error correction code in the plurality having a block length less than a predetermined threshold (Choukroun pg. 7 Table 1, all block lengths of each error correction code is less than 128).
As per claim 8:
Choukroun and Bennatan further teach the system according to claim 4, wherein the size-invariant transformation comprises a learned aggregation function (Choukroun pg. 8 section 6.1, learned self-attention aggregation).
As per claim 9:
Choukroun and Bennatan further teach the system according to claim 1, wherein each of the plurality of error correction codes is a linear code (Choukroun pg. 6 section 5).
As per claim 10:
Choukroun and Bennatan further teach the system according to claim 9, wherein a linear code is selected from the group comprising a Low-Density Parity Check (LDPC) code, a Polar code, a Reed Solomon code, and a Bose-Chaudhuri-Hocquenghem (BCH) code (Choukroun pg. 6 section 5).
As per claim 11:
Choukroun and Bennatan further teach the system according to claim 5, wherein decoding the received signal further comprises processing the received signal using a plurality of self-attention layers and feed-forward layers, and a plurality of normalization layers (Choukroun pg. 5 section 4.3, decoder has a concatenation of N decoding layers composed of self-attention and feed-forward layers interleaved by normalization layers).
As per claim 13:
Choukroun and Bennatan further teach the system according to claim 11, wherein the plurality of self-attention layers and feed-forward layers comprises at least 6 layers (Choukroun pg. 6 section 5, at least 6 encoding layers N).
As per claim 17:
Choukroun and Bennatan further teach the system according to claim 1, wherein the high-dimensional representation comprises at least 128 dimensions (Choukroun pg. 6 section 5, at least 128 dimensions d).
As per claim 18:
Choukroun and Bennatan further teach the system according to claim 7, wherein the predetermined threshold is 150 (Choukroun pg. 7 Table 1).
As per claim 19:
Choukroun and Bennatan further teach the system according to claim 1, wherein each error correction code in the plurality of error correction codes comprises a generator matrix and a parity check matrix (Choukroun pg. 2 section 3.1, each linear code is defined by a binary generator matrix and a binary parity check matrix), and wherein the pre-trained model was trained on the plurality of error correction codes using a plurality of differentiable masks, each differentiable mask being derived from the parity check matrix of a corresponding error correction code (Choukroun pg. 4 section 4.2, defining a function g(H) which defines masks to be applied to the self-attention mechanisms).
As per claim 20:
Choukroun and Bennatan further teach the system according to claim 5, wherein the noise prediction is based on one or more of the following noise models: additive white Gaussian noise, Rayleigh fading, or burst-error channels (Choukroun pg. 6 section 4.3, additive Gaussian noise is sampled randomly).
Claims 12 and 14-16 are rejected under U.S.C. 103 as being unpatentable over Choukroun-Bennatan in further view of Beery et al. (US Pat. Pub. 20220231785; hereinafter referred to as Beery).
As per claim 12:
Choukroun and Bennatan teach the system according to claim 11.
Choukroun and Bennatan do not explicitly disclose wherein decoding the received signal further comprises applying a distance embedding function, the distance embedding function being implemented as a fully connected neural network trained to learn a mapping from a number of paths in a Tanner graph to a scalar, the neural network comprising a multi-dimensional hidden layer and a plurality of nonlinear activation functions. However, Beery discloses wherein decoding the received signal further comprises applying a distance embedding function (Beery par. 0123, permutation embedding engine computes embedding vectors based on node embeddings of the Tanner graph), the distance embedding function being implemented as a fully connected neural network (Beery par. 0139, permutation classifier is implemented as a Neural Multiplayer Perception with fully connected layers) trained to learn a mapping from a number of paths in a Tanner graph to a scalar (Beery par. 0130, embeddings model is trained over the corresponding Tanner graph of the ECC), the neural network comprising a multi-dimensional hidden layer and a plurality of nonlinear activation functions (Beery par. 0148-0149, permutation classifier has hidden layers with multiple dimensions and uses LeakyReLU activation functions).
Choukroun, Bennatan, and Beery are analogous arts because they are in the same field of endeavor of neural decoding. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Beery’s permutation embedding engine and permutation classifier with the system of Choukroun-Bennatan. This modification would have been obvious to one of ordinary skill in the art at the time of filing because it allows for all embeddings to be computed once and stored in memory (Beery par. 0149), thus reducing the inference cost.
As per claim 14:
Choukroun, Bennatan, and Beery further teach the system according to claim 12, wherein the multi-dimensional hidden layer possesses at least 50 dimensions (Beery par. 0149, embedding size is set to 80 and the hidden layer dimensions are multiples of 80).
As per claim 15:
Choukroun, Bennatan, and Beery further teach the system according to claim 12, wherein the plurality of nonlinear activation functions comprises a ReLU activation function (Beery par. 0148).
As per claim 16:
Choukroun, Bennatan, and Beery further teach the system according to claim 12, wherein the learned mapping is represented as a fixed tensor at inference time (Beery par. 0195, neural network weights are learned during training and fixed during inference).
Claim 21 is rejected under U.S.C. 103 as being unpatentable over Choukroun-Bennatan in further view of Luo et al. (US Pat. Pub. 20220368349; hereinafter referred to as Luo).
As per claim 21:
Choukroun and Bennatan teach the system according to claim 1.
Choukroun and Bennatan do not explicitly disclose wherein the system is applied to one or more of the following: 5G NR wireless communication networks, Wi-Fi, satellite communications, or low-power IoT devices. However, Luo discloses wherein the system is applied to one or more of the following: 5G NR wireless communication networks, Wi-Fi, satellite communications, or low-power IoT devices (Luo par. 0038, encoded data received in a 5G wireless transmission to be decoded).
Choukroun, Bennatan, and Luo are analogous arts because they are in the same field of endeavor of neural decoding. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Luo’s 5G wireless communication with the system of Choukroun-Bennatan because the claimed invention is merely a combination of old elements, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of this combination would have been predictable since it would allow for increased reliability and provide ultra-low latency and high data throughput.
Conclusion
The additional prior arts not made of record and have not been relied upon are considered pertinent to applicant’s disclosure as follows:
Choukroun et al. (US Pat. Pub. 20240039559) discloses a system and method for training neural network based decoding for decoding error correction codes, comprising obtaining a plurality of training samples comprising one or more codewords encoded using an error correction code.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JEFFREY A YANG whose telephone number is (703)756-1447. The examiner can normally be reached Monday - Friday 8:30 a.m. - 5:30 p.m. PST.
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, Mark Featherstone can be reached at (571) 270-3750. 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.
/JEFFREY ANDREW YANG/Examiner, Art Unit 2111
/MARK D FEATHERSTONE/Supervisory Patent Examiner, Art Unit 2111
Prosecution Insights