DETAILED ACTION
This action is in response to the application filed 02/01/2024. Claims 1-20 are pending and have been examined.
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 Objections
Claim 1 is objected to because of the following informalities:
Regarding claim 1, “a pair of decoder neural networks of the plurality of decoder neural networks are replaced by another pair of neural networks of the plurality of decoder neural networks are” in lines 9-11 should read “a pair of decoder neural networks of the plurality of decoder neural networks are replaced by another pair of neural networks of the plurality of decoder neural networks”
Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 10 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claim 10, claim 10 recites “wherein the second signal-to-noise ratio is a wider range than the first signal-to-noise ratio”. It is unclear as to how a ratio can be a range. For purposes of examination, Examiner has interpreted this limitation to mean that the signal-to-noise ration has a range that is wider than the range of another signal-to-noise ratio.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1, 4, 8-11, 13-15, and 20 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Jamali et al. (“ProductAE: Towards Training Larger Channel Codes based on Neural Product Codes”) (hereafter referred to as Jamali).
Regarding claim 1, Jamali teaches
A method of training an autoencoder comprising a plurality of encoder neural networks and a plurality of decoder neural networks, the method comprising (Jamali, page 2, Figure 1
PNG
media_image1.png
216
1017
media_image1.png
Greyscale
):
training the plurality of encoder neural networks, wherein weights of the plurality of decoder neural networks are fixed during the training of the plurality of encoder neural networks (Jamali, page 3, 1st column, 1st paragraph, “during each training epoch, we first train the decoder several times while keeping the encoder network fixed, and then we train the encoder multiple times while keeping the decoder weights unchanged.”);
iteratively training the plurality of decoder neural networks for a plurality of iterations, wherein, for each iteration of training the plurality of decoder neural networks (Jamali, page 2, Figure 1
PNG
media_image1.png
216
1017
media_image1.png
Greyscale
):
a pair of decoder neural networks of the plurality of decoder neural networks are replaced by another pair of neural networks of the plurality of decoder neural networks are (Jamali, page 2, 2nd column, 3rd paragraph, “each pair of decoders at each iteration is replaced by a distinct pair of NN decoders resulting in 2I NN decoders in total.”),
and a second decoder neural network of the pair of decoder neural networks utilizes different parameters than a first decoder neural network of the pair of decoder neural networks (Jamali, page 1, Abstract, “instead of directly training the encoder and decoder for a large neural code of dimension k and blocklength n, we provide a framework that requires training neural encoders and decoders for the code parameters (k1-, n-1) and (k2, n2) such that k1k2 = k and n1n2 = n” where “Particular, instead of directly training the encoder and decoder for a large code of dimension k and blocklength n, we provide a framework that requires training smaller neural encoder and decoder components of parameters (k1-, n-1), (k2, n2), … ,(kM, nM) such that k1k2l…k-M = k and n1n2…nM = n for some positive integer M” (Jamali, page 1, 2nd column, 3rd paragraph) and “In this case, the first decoder
D
2
(
1
)
only takes the channel output as the input. All the next 2I -2 decoders take the channel output besides the output of the previous decoder, while the last decoder
D
1
(
1
)
only takes the output of the previous decoder as input (due to some issues with the size of the tensors being concatenated at the input of this decoder”. Examiner notes that inputs of the different decoders are the different parameters.).
Regarding claim 4, Jamali teaches
The method of claim 1, wherein the iteratively training the plurality of decoder neural networks comprises: training all of the plurality of decoder neural networks for each of a first number of iterations of the plurality of iterations; training only a single pair of decoder neural networks for each of a second number of iterations of the plurality of iterations after the first number of iterations; and training all of the plurality of decoder neural networks for each of a third number of iterations of the plurality of iterations after the second number of iterations (Jamali, page 4, Figure 3.
PNG
media_image2.png
326
823
media_image2.png
Greyscale
Examiner notes that the iterations are being taught by the flow of arrows in Figure 3. Examiner notes that training all the decoder neural networks occurs in the first iteration by starting from the far right decoder and working left to the decoder optimizer. Examiner notes that as Figure 3 shows, the next iteration can encompass training only the single pair of decoders by looping back to the third decoder from the right. Examiner further notes that the last iteration loops back to train all the decoders again.).
Regarding claim 8, Jamali teaches
The method of claim 1, wherein the plurality of encoder neural networks is configured to map a message to a codeword and to transmit the codeword over a noisy channel having a signal-to-noise ratio, wherein the signal-to-noise ratio is a first signal-to-noise ratio in a first range, and wherein the method further comprises retraining the autoencoder for a plurality of epochs over the noisy channel having a second signal-to-noise ratio different than the first signal-to-noise ratio (Jamali, page 4, Section C. 3, “The encoder and decoder training schedules were discussed in Section III-A. Throughout our experiments, we used Tenc = 100 and Tdec = 500. Also at each encoder or decoder training iteration, the training SNR (equivalently the noise variance of the channel) is carefully chosen. In particular, we used a single-point of γ dB for the encoder training SNR but a range of [γ-2.5, γ+1] dB for the decoder training SNR. Therefore, during each encoder training iteration, a single SNR is used to generate all noise vectors of the batch. However, during each iteration of the decoder training schedule, B random values are picked uniformly from the interval [γ-2.5, γ+1] dB to generate B noise vectors of appropriate variances” where “Consider transmission of a length-k sequence of information bits u across a noisy channel. The channel encoder ε(•) maps u to a length-n sequence of coded symbols c = ε(u) called a codeword….In the case of transmission over the additive white Gaussian noise (AWGN) channel, which is considered in the paper, the received noisy signal can be expressed as y = c+n, where n is the channel noise vector whose components are Gaussian random variables with mean zero and variance σ2” (Jamali, page 2, 1st column, 1st paragraph). Examiner notes that the first range is just the ratio. Examiner further notes that retraining the autoencoder encompasses the 500 iterations of training the decoder under BRI.).
Regarding claim 9, Jamali teaches
The method of claim 8, wherein the second signal-to-noise ratio is larger than the first signal-to-noise ratio (Jamali, page 4, Section C. 3, “The encoder and decoder training schedules were discussed in Section III-A. Throughout our experiments, we used Tenc = 100 and Tdec = 500. Also at each encoder or decoder training iteration, the training SNR (equivalently the noise variance of the channel) is carefully chosen. In particular, we used a single-point of γ dB for the encoder training SNR but a range of [γ-2.5, γ+1] dB for the decoder training SNR. Therefore, during each encoder training iteration, a single SNR is used to generate all noise vectors of the batch. However, during each iteration of the decoder training schedule, B random values are picked uniformly from the interval [γ-2.5, γ+1] dB to generate B noise vectors of appropriate variances.” Examiner notes that a random value could include γ+1 which is larger than γ.).
Regarding claim 10, Jamali teaches
The method of claim 9, wherein the second signal-to-noise ratio is a wider range than the first signal-to-noise ratio (Jamali, page 4, Section C. 3, “The encoder and decoder training schedules were discussed in Section III-A. Throughout our experiments, we used Tenc = 100 and Tdec = 500. Also at each encoder or decoder training iteration, the training SNR (equivalently the noise variance of the channel) is carefully chosen. In particular, we used a single-point of γ dB for the encoder training SNR but a range of [γ-2.5, γ+1] dB for the decoder training SNR. Therefore, during each encoder training iteration, a single SNR is used to generate all noise vectors of the batch. However, during each iteration of the decoder training schedule, B random values are picked uniformly from the interval [γ-2.5, γ+1] dB to generate B noise vectors of appropriate variances”.
Regarding claim 11, Jamali teaches
The method of claim 8, wherein the plurality of epochs comprises 11 epochs (Jamali, page 4, Section C. 3, “The encoder and decoder training schedules were discussed in Section III-A. Throughout our experiments, we used Tenc = 100 and Tdec = 500. Also at each encoder or decoder training iteration, the training SNR (equivalently the noise variance of the channel) is carefully chosen. In particular, we used a single-point of γ dB for the encoder training SNR but a range of [γ-2.5, γ+1] dB for the decoder training SNR. Therefore, during each encoder training iteration, a single SNR is used to generate all noise vectors of the batch. However, during each iteration of the decoder training schedule, B random values are picked uniformly from the interval [γ-2.5, γ+1] dB to generate B noise vectors of appropriate variances.” Examiner notes that the amount of epochs is 100 for the encoder and 500 for the decoder which are both greater than 11 and thus comprise 11 epochs.).
Regarding claim 13, Jamali teaches
The method of claim 1, training the plurality of encoder neural networks comprises applying power normalization (Jamali, page 3, 1st column, 2nd paragraph, Encoder Architecture and Its Training Schedule, “At the end, the power normalization, to be clarified in Section III-C, will be applied to the codewords.”).
Regarding claim 14, Jamali teaches
The method of claim 1, training the plurality of decoder neural networks comprises applying power normalization (Jamali, page 3, 1st column, 2nd paragraph, Encoder Architecture and Its Training Schedule, “At the end, the power normalization, to be clarified in Section III-C, will be applied to the codewords.” Examiner notes that since the power normalization is part of training the encoders, the power normalization is part of the training of the autoencoder which includes the decoder neural networks.).
Regarding claim 15, Jamali teaches
An autoencoder comprising: a plurality of encoder neural networks configured to map a message to a codeword and to transmit the codeword over a noisy channel having a signal-to-noise ratio (Jamali, page 2, 1st column, 1st paragraph, “Consider a transmission of a length-k sequence of information bits u across a noisy channel. The channel encoder ε(•) maps u to a length-n sequence of coded symbols c = ε(u) called a codeword. Here, k and n are the code dimension and blocklength, respectively, and the resulting code is denoted by a (k, n) code….The ratio of the average energy per coded symbol to the noise variance is called the signal-to-noise ratio (SNR).”)
A plurality of decoder neural networks configured to decode the message (Jamali, page 2, Figure 1
PNG
media_image1.png
216
1017
media_image1.png
Greyscale
)
Wherein a second decoder neural network of the pair of decode neural networks utilizes different parameters than a first decode neural network of the pair of decoder neural networks(Jamali, page 1, Abstract, “instead of directly training the encoder and decoder for a large neural code of dimension k and blocklength n, we provide a framework that requires training neural encoders and decoders for the code parameters (k1-, n-1) and (k2, n2) such that k1k2 = k and n1n2 = n” where “Particular, instead of directly training the encoder and decoder for a large code of dimension k and blocklength n, we provide a framework that requires training smaller neural encoder and decoder components of parameters (k1-, n-1), (k2, n2), … ,(kM, nM) such that k1k2l…k-M = k and n1n2…nM = n for some positive integer M” (Jamali, page 1, 2nd column, 3rd paragraph) and “In this case, the first decoder
D
2
(
1
)
only takes the channel output as the input. All the next 2I -2 decoders take the channel output besides the output of the previous decoder, while the last decoder
D
1
(
1
)
only takes the output of the previous decoder as input (due to some issues with the size of the tensors being concatenated at the input of this decoder”. Examiner notes that inputs of the different decoders are the different parameters.).
Regarding claim 20, Jamali teaches
The autoencoder of claim 15, wherein the autoencoder is trained on a noisy channel having a first signal-to-noise ratio and a second signal-to-noise ratio different than the first signal-to-noise ratio (Jamali, page 4, Section C. 3, “The encoder and decoder training schedules were discussed in Section III-A. Throughout our experiments, we used Tenc = 100 and Tdec = 500. Also at each encoder or decoder training iteration, the training SNR (equivalently the noise variance of the channel) is carefully chosen. In particular, we used a single-point of γ dB for the encoder training SNR but a range of [γ-2.5, γ+1] dB for the decoder training SNR. Therefore, during each encoder training iteration, a single SNR is used to generate all noise vectors of the batch. However, during each iteration of the decoder training schedule, B random values are picked uniformly from the interval [γ-2.5, γ+1] dB to generate B noise vectors of appropriate variances” where “Consider transmission of a length-k sequence of information bits u across a noisy channel. The channel encoder ε(•) maps u to a length-n sequence of coded symbols c = ε(u) called a codeword….In the case of transmission over the additive white Gaussian noise (AWGN) channel, which is considered in the paper, the received noisy signal can be expressed as y = c+n, where n is the channel noise vector whose components are Gaussian random variables with mean zero and variance σ2” (Jamali, page 2, 1st column, 1st paragraph).).
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.
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.
Claim(s) 2-3, and 16-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Jamali in view of Lin et al. (“Early Decoding for Gaussian Broadcast Channels with Heterogeneous Blocklength Constraints”) (hereafter referred to as Lin).
Regarding claim 2, Jamali teaches the method of claim 1. Jamali does not teach, but Lin does teach
wherein the second decoder neural network utilizes a larger blocklength than the first decoder neural network (Lin, page 2, 1st column, 3rd bullet point, “Two decoders, where decoder k assigns an estimate
m
k
^
∈
M
k
, or an error message to each received sequence
y
k
n
k
, k = 1, 2” where “Unlike traditional GBC [Gaussian broadcast channel] where two users have the same blocklength constraints, here the user with higher output SNR may have a shorter blocklength constraint” (Lin, page 1, abstract) and “when n1=n2, user 2 can perform the traditional two-step SIC. The first step is decoding user 1’s codeword and removing this interference. Next, user 2 decodes his own codeword. However, with n1>n2, it is not clear that whether the first step of SIC is still feasible. This motivates the early decoding for this step” (Lin, page 2, 1st column, 2nd to last paragraph). Examiner notes that the second decoder is the decoder with the blocklength of n1--. Examiner further notes that the SNR is the rate)
Jamali and Lin are considered analogous to the claimed invention because they both use multiple decoders to decode messages sent across channels. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Jamali to use a decoder that utilizes a larger blocklength than another decoder such as in Lin’s early decoding. Doing so is advantageous because “Numerical results show that the successful early decoding significantly reduces blocklength in the finite blocklength regime, if the channel advantage is sufficiently large” (Lin, page 4, Conclusion).
Regarding claim 3, Jamali teaches the method of claim 1. Jamali does not teach, but Lin does teach
wherein the second decoder neural network utilizes a smaller rate than the first decoder neural network (Lin, page 2, 1st column, 3rd bullet point, “Two decoders, where decoder k assigns an estimate
m
k
^
∈
M
k
, or an error message to each received sequence
y
k
n
k
, k = 1, 2” where “Unlike traditional GBC [Gaussian broadcast channel] where two users have the same blocklength constraints, here the user with higher output SNR may have a shorter blocklength constraint” (Lin, page 1, abstract). (Lin, page 1, abstract) and “when n1=n2, user 2 can perform the traditional two-step SIC. The first step is decoding user 1’s codeword and removing this interference. Next, user 2 decodes his own codeword. However, with n1>n2, it is not clear that whether the first step of SIC is still feasible. This motivates the early decoding for this step” (Lin, page 2, 1st column, 2nd to last paragraph). Examiner notes that the second decoder is the decoder with the blocklength of n1--. Examiner further notes that the SNR is the rate) .
Jamali and Lin are considered analogous to the claimed invention because they both use multiple decoders to decode messages sent across channels. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Jamali to use a decoder that utilizes a smaller rate than another decoder such as in Lin’s early decoding. Doing so is advantageous because “Numerical results show that the successful early decoding significantly reduces blocklength in the finite blocklength regime, if the channel advantage is sufficiently large” (Lin, page 4, Conclusion).
Regarding claim 16, claim 16 recites substantially similar limitations to claim 2, and is therefore rejected under the same analysis.
Regarding claim 17, claim 17 recites substantially similar limitations to claim 3, and is therefore rejected under the same analysis.
Claim(s) 5-7, and 18-19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Jamali in view of Aoudia et al. (“End-to-End Learning of Communications Systems Without a Channel Model”) (hereafter referred to as Aoudia).
Regarding claim 5, Jamali teaches
The method of claim 1, wherein the plurality of encoder neural networks is configured to map a message to a codeword and to transmit the codeword over a noisy channel (Jamali, page 2, 1st column, 1st paragraph, “Consider a transmission of a length-k sequence of information bits u across a noisy channel. The channel encoder ε(•) maps u to a length-n sequence of coded symbols c = ε(u) called a codeword. Here, k and n are the code dimension and blocklength, respectively, and the resulting code is denoted by a (k, n) code”)
Jamali does not teach, but Aoudia does teach
wherein the noisy channel is a first type of channel, and wherein the method further comprises retraining the autoencoder on a second type of channel different than the first type of channel (Aoudia, page 1, 1st column, last paragraph – 2nd column, 1st paragraph, “Inspired by this technique, we develop a novel algorithm for end-to-end training which iterates between two phases: (i) supervised training of the receiver and (ii) RL-based training of the transmitter based on an estimated gradient of the loss. Comparison with the fully supervised training method of [1] on additive white Gaussian noise (AWGN) and Rayleigh block-fading (RBF) channels reveals essentially identical performance. Although our method requires considerably more training iterations on AWGN channels, it converges faster on RBF channels.” Examiner notes that by training on both AWGN and RBF channels, the method comprises retraining the autoencoder on a different type of channel.).
Jamali and Aoudia are considered analogous to the claimed invention because they both use autoencoders to encode and decode messages that are passed through channels. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Jamali to use two different types of channels like in Aoudia. Doing so is advantageous because “the main contribution of this paper is to show that knowledge of the channel model and the instantaneous channel transfer function is indeed not needed. This implies that the autoencoder can be trained from pure observations alone without any knowledge of the underlying channel model” (Aoudia, page 1, 1st column, last paragraph).
Regarding claim 6, Jamali in view Aoudia teaches the method of claim 5. Jamali in view of Aoudia further teaches
wherein the first type of channel is an additive white Gaussian noise (AWGN) channel, and wherein the second type of channel is a Rayleigh fading channel (Aoudia, page 1, 1st column, last paragraph – 2nd column, 1st paragraph, “Inspired by this technique, we develop a novel algorithm for end-to-end training which iterates between two phases: (i) supervised training of the receiver and (ii) RL-based training of the transmitter based on an estimated gradient of the loss. Comparison with the fully supervised training method of [1] on additive white Gaussian noise (AWGN) and Rayleigh block-fading (RBF) channels reveals essentially identical performance. Although our method requires considerably more training iterations on AWGN channels, it converges faster on RBF channels.”)
Jamali and Aoudia are considered analogous to the claimed invention because they both use autoencoders to encode and decode messages that are passed through channels. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Jamali to use two different types of channels like in Aoudia. Doing so is advantageous because “the main contribution of this paper is to show that knowledge of the channel model and the instantaneous channel transfer function is indeed not needed. This implies that the autoencoder can be trained from pure observations alone without any knowledge of the underlying channel model” (Aoudia, page 1, 1st column, last paragraph).
Regarding claim 7, Jamali in view Aoudia teaches the method of claim 5. Jamali in view of Aoudia further teaches
wherein the retraining of the autoencoder comprises performing a single training epoch on the second type of channel (Aoudia, page 1, 1st column, last paragraph – 2nd column, 1st paragraph, “Inspired by this technique, we develop a novel algorithm for end-to-end training which iterates between two phases: (i) supervised training of the receiver and (ii) RL-based training of the transmitter based on an estimated gradient of the loss. Comparison with the fully supervised training method of [1] on additive white Gaussian noise (AWGN) and Rayleigh block-fading (RBF) channels reveals essentially identical performance. Although our method requires considerably more training iterations on AWGN channels, it converges faster on RBF channels.” Examiner notes that by training on RBF channels, the retraining comprises performing a single training epoch on the second type of channel.).
Jamali and Aoudia are considered analogous to the claimed invention because they both use autoencoders to encode and decode messages that are passed through channels. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Jamali to train on two different types of channels like in Aoudia. Doing so is advantageous because “the main contribution of this paper is to show that knowledge of the channel model and the instantaneous channel transfer function is indeed not needed. This implies that the autoencoder can be trained from pure observations alone without any knowledge of the underlying channel model” (Aoudia, page 1, 1st column, last paragraph).
Regarding claim 18, Jamali teaches the autoencoder of claim 15. Jamali does not teach, but Aoudia does teach
wherein the autoencoder is trained on a first type of channel and on a second type of channel different than the first type of channel (Aoudia, page 1, 1st column, last paragraph – 2nd column, 1st paragraph, “Inspired by this technique, we develop a novel algorithm for end-to-end training which iterates between two phases: (i) supervised training of the receiver and (ii) RL-based training of the transmitter based on an estimated gradient of the loss. Comparison with the fully supervised training method of [1] on additive white Gaussian noise (AWGN) and Rayleigh block-fading (RBF) channels reveals essentially identical performance. Although our method requires considerably more training iterations on AWGN channels, it converges faster on RBF channels.” Examiner notes that by training on both AWGN and RBF channels, the method comprises retraining the autoencoder on a different type of channel.).
Jamali and Aoudia are considered analogous to the claimed invention because they both use autoencoders to encode and decode messages that are passed through channels. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Jamali to use two different types of channels like in Aoudia. Doing so is advantageous because “the main contribution of this paper is to show that knowledge of the channel model and the instantaneous channel transfer function is indeed not needed. This implies that the autoencoder can be trained from pure observations alone without any knowledge of the underlying channel model” (Aoudia, page 1, 1st column, last paragraph).
Regarding claim 19, claim 19 recites substantially similar limitations to claim 6, and is therefore rejected under the same analysis.
Claim(s) 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over Jamali in view of Meller et al. (US 2020/0259510 A1) (hereafter referred to as Meller).
Regarding claim 12, Jamali teaches
The method of claim 1, wherein the plurality of encoder neural networks is configured to map a message to a codeword and to transmit the codeword over a noisy channel having a signal-to-noise ratio (Jamali, page 2, 1st column, 1st paragraph, “Consider a transmission of a length-k sequence of information bits u across a noisy channel. The channel encoder ε(•) maps u to a length-n sequence of coded symbols c = ε(u) called a codeword. Here, k and n are the code dimension and blocklength, respectively, and the resulting code is denoted by a (k, n) code….The ratio of the average energy per coded symbol to the noise variance is called the signal-to-noise ratio (SNR).”),
Jamali does not teach, but Meller does teach
and wherein the message has a code dimension of at least 300 bits (Meller, page 24, paragraph 0260-0262, “The codes are described by their information payload size in bits (i.e. code dimension), their code rates and codeword length in bits e.g.: [0261] code #1 as length N=1200 bits, information size=200 bits, and therefore code rate R = 200/1200 = 1/6 [0262] code #2 has code length 1800 bits, information size=600 bits and therefore code rate R = 600/1800=1/3.” Examiner notes that the information size is the code dimension.)
Jamali and Meller are considered analogous to the claimed invention because they both use autoencoders to encode and decode messages. It would have been obvious to one having ordinary skill in the art prior to the effective filing date to have modified Jamali to use the code dimension set in Meller. Doing so is advantageous because “among the advantages of certain embodiments of the presently disclosed subject matter are low latency decoding, low power consumption, and better error-correction performance (lower frame-error-rate or bit-error-rate) compared to prior art solutions” (Meller, page 14, paragraph 0055).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Jiang et al. (“Turbo Autoencoder: Deep learning based channel codes for point-to-point communication channels”) also discloses transmitting messages from encoders to decoders through noisy channels.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KAITLYN R LAU whose telephone number is (571)272-1429. The examiner can normally be reached Monday - Thursday: 8:00 am - 6:00 pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Michelle Bechtold can be reached at (571) 431-0762. 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.
/K.R.L./Examiner, Art Unit 2148
/MICHELLE T BECHTOLD/Supervisory Patent Examiner, Art Unit 2148
Prosecution Insights