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(s) 1-11, 13-15 are pending.
Drawings are accepted.
IDS are accepted.
Response to Arguments
Remarks dated 10/13/2025 is/are fully considered. The arguments are reviewed, but they are found not persuasive. Specifically:
Applicant argues Damm does not teach "the channel frequency response is generated in response to a transmission over the channel or a simulation thereof" as recited in claim 1, and either does Ali teach this missing limitation, per Remarks – page 8.
Applicant however appears to contradict that very assertion in the Remarks by stating “the cited portions of Damm teach a frequency response of transmission lines that generates frequency response data from which Hlog data is created ([033]), the frequency data may be generated from simulated impairments or transmission lines ([015], [044]), or be retrieving magnitude-frequency response data in relation to the transmission line ([054])”. The argument is therefore not persuasive.
Damm indeed shows in ¶0032-0034 that frequency response system 18 includes a measurement module 20, which measures the frequency response of transmission lines 12, which generates frequency response data 22 and from which Hlog data 24 is created, which couldn’t be clearer and expressively disclosing the generated frequency response data is generated in responsive the transmissions in lines 12. In fact, the very term “frequency response” in Damm is self-explanatory, i.e. a frequency response to signal transmission over the line. ¶0034 further discloses the dotted line 28 showing the ideal frequency response and solid line 30 showing the actual frequency response of this particular transmission line 12, which is an explicit disclosure for the limitation at issue.
Applicant further argues references Damm and Ali does not disclose “the channel response data comprises a one-dimensional channel response vector”.
The examiner respectfully disagrees. The reference Damm is quite explicit in disclosing this feature. Specifically:
Damm in ¶0014 discloses “In some embodiments, the dimensionality reduction system is an encoding portion of an autoencoder and wherein the autoencoder is a one dimensional (one channel) unsupervised convolutional neural network”. The teaching is rather explicit that the system employing one-dimensional model (thus employing one-dimensional input vector), with the input layer of fixed number of nodes. An input data set cannot be in multi-dimensional format for compatibility purposes, as a multi-dimensional input vector would not function with the system of single-dimensional input layer.
Furthermore, one-dimensional vector has a format of a V = [ x, y, z.., n], with each member is a single value. In a multi-dimensional vector, each member comprises multiple values indicative of various aspects such as magnitude, direction, depth, location etc. Damm in at least ¶0014, 0040, disclose a single-channel input as a 400 discrete sampled magnitudes of channel responses, without any other dimensions of direction, depth, etc. By treating the frequency response data as 400 sampled inputs, the system fundamentally receiving and processing as a one-dimensional vector.
¶0048 further discloses “An input layer of 50 inputs is used in the above to correlate to the 50 nodes of the encoder 62 above, representing an encoding of the original 400 input points of the original magnitude frequency response data”. This confirms the initial “channel response data” is comprised of a specific set of 400 values (vector) that is no other dimensions other than magnitude values.
The argument is therefore not persuasive.
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.
Claim(s) 1-11, 13-15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Damm et al. (AU 2020256389) in view of Ali et al. NPL titled “Escaping the Big Data Paradigm with Compact Transformers” (06/2022) – IDS.
As to claim 1:
Damm discloses:
An apparatus comprising: at least one memory including computer program code; at least one processor configured to execute the computer program code (¶011, 023, 054, computer executing program) and cause the apparatus to perform, obtaining channel response data comprising a channel frequency response of a channel over a frequency spectrum, (See at least ¶032-035, 038, measure frequency response of at least a frequency channel . Response data is obtained, i.e. frequency response data in form of Hlog)
wherein the channel frequency response is generated in response to a transmission over the channel or a simulation thereof, (¶015, 044, 054, system processes both real (measured) and simulated channel response data), and wherein the channel response data comprises one-dimensional response vector (See ¶0040, 0048, 0014, “In some embodiments, the dimensionality reduction system is an encoding portion of an autoencoder and wherein the autoencoder is a one dimensional (one channel) unsupervised convolutional neural network”. The teaching is rather explicit that the system employing one-dimensional model (thus employing one-dimensional input vector), with the input layer of fixed number of nodes. An input data set cannot be in multi-dimensional format for compatibility purposes, as a multi-dimensional input vector would not function with the system of single-dimensional input layer)
and generating an indication of channel impairments in response to applying the channel response data to a machine-learning, ML, model trained to predict a channel impairment estimate. (¶036, 046-056, outputting specific predictions of what impairment exists using a machine learning model, such as a fully connected neural network/convolutional neural network)
Except that Damm does not explicitly indicates the machine learning model being a transformer-based ML model to perform the generation step above.
Ali, in a related field of machine-learning data processing and training computer vision, discloses in Section I, page 2, and 4.3 page 8, that transformer neural networks, such as ViT can be desirable as an alternative in data interference and prediction as opposed to a convolutional network (such as the channel impairment inference of Damm), which comes with its own set of benefit.
It would have been obvious to one of ordinary skill in the art before the effective filing time of the invention that Damm’s system of using neural network to infer channel impairments can incorporate the specific type of neural network, namely the transformer neural network, to process and infer statistical data. Both Damm and Ali advocate the use of machine learning to process and infer data. Damm in ¶0046-0056 that neural networks can be effectively used to infer channel impairment. Ali in page 8 indicates the specific transform type of neural network can be implemented to process any size of data with high confidence. In page 2-3, Ali again stressed the advantages of transformer network as being light weight, data-efficient and highly desirable with its own sets of advantages.
As to claim 2:
Damm in view of Ali discloses all limitations of claim 1, wherein the a one-dimensional channel response vector comprising data representative of a Hlog channel response. (See Damm, ¶032-038, generate frequency response data from which Hlog is created)
As to claim 3:
Damm in view of Ali discloses all limitations of claim 1, wherein Damm further discloses: the transformer-based ML model further comprises a pre-processing component, coupled to a transformer encoder neural network and multiclass classifier; (¶040-046, 014, autoencoder 62 coupled to the inference neural network (transformer in view of Ali), multi-classifier read as the machine learning 52 with output layers of nodes being number of output classes per ¶047-048)
the pre-processing component is configured for pre-processing the channel response data into a multi-dimensional embedding for input to a transformer encoder neural network; (¶014, 041-046, NN encoder compressed the Hlog vector to produces multi-node embedding)
the transformer encoder neural network is configured for processing the multi-dimensional embedding and outputting a multi-dimensional encoded signal of the channel response data; (¶046-056, outputting specific predictions of what impairment exists using a machine learning model, such as a fully connected neural network/convolutional neural network. The encoder is a transformer in view of Ali) the multi-class classifier is configured for processing the multi-dimensional encoded signal and predicting a multiclass channel impairment estimate.(¶046-054, output a predicted impairment or impairment class, using output layers of multiple nodes)
As to claim 4:
Damm in view of Ali discloses all limitations of claim 3, wherein the transformer encoder neural network is a visualization transformer ML model and the pre-processing component is configured to encode the channel response data into a multi-dimensional embedding for input to the visualization transformer ML model. (Ali in 3.2 discloses ViT as discussed in claim 1 above, also page 3, ViT subdivides an input into non-overlapping square patches in raster-scan order. The sequence of patches, xp ∈ R H×(P 2C) with patch size P, are flattened into 1D vectors and transformed into latent vectors of dimension d”. Damm disclose in at least ¶041, the encoder has an output of multiple nodes that represent the plurality of input points. )
As to claim 5:
Damm in view of Ali discloses all limitations of claim 4, wherein the pre-processing component is further configured to group a plurality of data elements of the input channel response data into patches and generating the multi-dimensional embedding that projects each of the patches along a projection dimension of length pdim. (Damm, ¶040, grouping of data element, Ali page 3, ViT subdivides an input into non-overlapping square patches in raster-scan order. The sequence of patches, xp ∈ R H×(P 2C) with patch size P, are flattened into 1D vectors and transformed into latent vectors of dimension d”)
As to claim 6:
Damm in view of Ali discloses all limitations of claim 3, wherein the pre-processing component is a neural network ML model configured for feature extraction and encoding of the channel response data into a multi-dimensional embedding for input to the transformer encoder neural network. (See at least ¶014, 040-046, autoencoder is a one-dimensional neural network, which compresses input frequency response data into a lower dimensional encoding. In other words, autoencoder extracts features and encodes the Hlog input)
As to claim 7:
Damm in view of Ali discloses all limitations of claim 6, wherein the neural network ML model is configured to process groupings of a plurality of elements of the input channel response data, perform feature extraction of the groupings, (Damm, ¶014, 040, 047, the 1D convolutional layer slide over the Hlog vector and processes in it overlap patches (5-point window), which matches the processing groupings of data elements. See also 0040-0042, encoder maps inputs into codes, i.e. extracting feature elements)and generate a multi-dimensional embedding that projects each of the data elements of the input channel response along a projection dimension of length pdim. (Ali page 3, ViT subdivides an input into non-overlapping square patches in raster-scan order. The sequence of patches, xp ∈ R H×(P 2C) with patch size P, are flattened into 1D vectors and transformed into latent vectors of dimension d”)
As to claim 8:
Damm in view of Ali discloses all limitations of claim 6, wherein the neural network ML model is a convolutional encoder neural network ML model. (Damm, ¶014, CNN model)
As to claim 9:
Damm in view of Ali discloses all limitations of claim 8, the convolutional encoder neural network ML model further comprises a neural network of one or more convolution layers, one or more pooling layers, and one or more fully-connected layers configured for extracting a channel response feature set and outputting the multi-dimensional embedding of said channel response feature set for input to the transformer encoder neural network. (See Damm, ¶0049-0050, fully connected, convolutional, pooling layers included. ¶046-056, outputting specific predictions of what impairment exists using a machine learning model, such as a fully connected neural network/convolutional neural network
As to claim 10:
Damm in view of Ali discloses all limitations of claim 3, wherein the transformer encoder neural network comprises one or more transformer encoders coupled together, wherein each transformer encoder comprises one or more multi-headed attention layers, one or more normalization layers, and wherein at least the final transformer encoder includes one or more multi-layer perceptron layers for outputting the multi-dimensional encoding of the channel response data. (Ali, page 5, “transformer blocks, each including an MHSA layer and an MLP block. The encoder also applies Layer Normalization, GELU activation, and dropout”, page 3 “A transformer encoder consists of a series of stacked encoding layers. Each encoder layer is comprised of two sub-layers: Multi-Headed Self-Attention (MHSA) and a Multi-Layer Perceptron (MLP) head. Each sub-layer is preceded by a layer normalization (LN), and followed by a residual connection to the next sub-layer”)
As to claim 11:
Damm in view of Ali discloses all limitations of claim 1, wherein the apparatus is further caused to perform training of the transformer-based ML model based on, obtaining training data instances, each training data instance comprising data representative of a channel response and data representative of a target channel impairment associated with the channel response; applying a training data instance to the transformer-based ML model; estimating a loss based on a difference between the estimated channel impairment(s) output by the transformer-based ML model and the target channel impairment(s) of each training data instance; and updating a set of weights associated with the transformer-based ML model based on the estimated loss
(Claim 11 is directed to general standard training practice for machine learning model, i.e. using training data to receive output for each iteration to produce a loss function. Offset are determined to retune (adjustment of weights of each nodes) for next iteration, thus is not novel. This practice is disclosed in Damm, ¶0044-0056, each training instance include obtaining channel response impairment output for each training data iteration, comparing against known results (to determine error), and retrain the model based on such offset. “Backpropagation”, i.e. practice of updating weights after determining computed loss.)
As to claim 13:
Damm in view of Ali discloses all limitations of claim 1, wherein the channel is a communications medium comprising a wired communications medium or, a wireless communications medium, or a combination of both. (See Damm, ¶031)
As to claim 14:
Damm discloses:
A method comprising: obtaining channel response data comprising a channel frequency response of a channel over a frequency spectrum, (See at least ¶032-035, 038, measure frequency response of at least a frequency channel . Response data is obtained, i.e. frequency response data in form of Hlog)
wherein the channel frequency response is generated in response to a transmission over the channel or a simulation thereof; (¶015, 044, 054, system processes both real (measured) and simulated channel response data)
and generating an indication of channel impairments in response to applying the channel response data to a transformer-based machine-learning, ML, model trained to predicting a channel impairment estimate. (¶046-056, outputting specific predictions of what impairment exists using a trained machine learning model, such as a fully connected neural network/convolutional neural network)
Except that Damm does not explicitly indicates the machine learning model being a transformer-based ML model to perform the generation step above.
Ali, in a related field of machine-learning data processing and training computer vision, discloses in Section I, page 2, and 4.3 page 8, that transformer neural networks, such as ViT can be desirable as an alternative in data interference and prediction as opposed to a convolutional network (such as the channel impairment inference of Damm), which comes with its own set of benefit.
It would have been obvious to one of ordinary skill in the art before the effective filing time of the invention that Damm’s system of using neural network to infer channel impairments can incorporate the specific type of neural network, namely the transformer neural network, to process and infer statistical data. Both Damm and Ali advocate the use of machine learning to process and infer data. Damm in ¶0046-0056 that neural networks can be effectively used to infer channel impairment. Ali in page 8 indicates the specific transform type of neural network can be implemented to process any size of data with high confidence. In page 2-3, Ali again stressed the advantages of transformer network as being light weight, data-efficient and highly desirable with its own sets of advantages.
As to claim 15:
Damm discloses:
A non-transitory computer readable medium storing computer program code that when executed by a processor causes and apparatus including the processor (¶011, 023, 054, computer executing program) to perform, obtaining channel response data comprising a channel frequency response of a channel over a frequency spectrum, (See at least ¶032-035, 038, measure frequency response of at least a frequency channel . Response data is obtained, i.e. frequency response data in form of Hlog)
wherein the channel frequency response is generated in response to a transmission over the channel or a simulation thereof; (¶015, 044, 054, system processes both real (measured) and simulated channel response data)
and generating an indication of channel impairments in response to applying the channel response data to a transformer-based machine-learning, ML, model trained to predicting a channel impairment estimate. (¶046-056, outputting specific predictions of what impairment exists using a machine learning model, such as a fully connected neural network/convolutional neural network)
Except that Damm does not explicitly indicates the machine learning model being a transformer-based ML model to perform the generation step above.
Ali, in a related field of machine-learning data processing and training computer vision, discloses in Section I, page 2, and 4.3 page 8, that transformer neural networks, such as ViT can be desirable as an alternative in data interference and prediction as opposed to a convolutional network (such as the channel impairment inference of Damm), which comes with its own set of benefit.
It would have been obvious to one of ordinary skill in the art before the effective filing time of the invention that Damm’s system of using neural network to infer channel impairments can incorporate the specific type of neural network, namely the transformer neural network, to process and infer statistical data. Both Damm and Ali advocate the use of machine learning to process and infer data. Damm in ¶0046-0056 that neural networks can be effectively used to infer channel impairment. Ali in page 8 indicates the specific transform type of neural network can be implemented to process any size of data with high confidence. In page 2-3, Ali again stressed the advantages of transformer network as being light weight, data-efficient and highly desirable with its own sets of advantages.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 2006/0251156 - A wireless communication receiver comprising one or more processing circuits configured to generate impairment correlations for one or more data signals transmitted in conjunction with pilot signals from a transmitter having multiple transmit antennas by: determining a data-to-pilot signal transmit power ratio and transmit antenna power distributions for the data and pilot signals; and calculating the impairment correlations as a function of the data-to-pilot signal transmit power ratio and the transmit antenna power distributions for the data and pilot signals..
THIS ACTION IS MADE FINAL. 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 QUAN M HUA whose telephone number is (571)270-7232. The examiner can normally be reached 10:30-6:30.
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, Anthony Addy can be reached at 571-272-7795. 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.
/QUAN M HUA/Primary Examiner, Art Unit 2645
Prosecution Insights