DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
This non-final office action is in response to the application filed 29 June 2023.
Claims 1-20 are pending. Claims 1, 12, and 16 are independent claims.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 29 June 2023 and 16 December 2024 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
Drawings
The examiner accepts the drawings filed 29 June 2023.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 12-15 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter.
With respect to independent claim 12, the claim recites a “system for efficiently storing a sparse neural network, the system comprising: one or more memories configured to store (lines 1-2).”
The applicant’s specification states, embodiments of the invention “may include an article such as a non-transitory computer or processor readable medium, or a computer or processor non-transitory storage medium, such as for example a memory (paragraph 0088; emphasis added).” However, these are non-limited examples. The broadest reasonable interpretation of the term “one or more memories” includes embodiments where the memories are transitory in nature.
In such embodiments, the claimed “system” fails to recite any hardware component, and therefore, does not recite a machine. Similarly, the claim fails to recite a process, manufacture, or composition of matter. For this reason, claim 12 is non-statutory.
The examiner recommends amending the claim to further recite “a processor or controller.”
Claims 13-15 fail to cure the deficiencies of independent claim 12. Claims 13-15 are rejected under similar rationale.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
According to Step 1 of the two Step analysis, claims 1-11 are directed toward a method (process). Claims 16-20 are directed toward a non-transitory computer-readable storage medium (manufacture). Therefore, each of these claims falls within one of the four statutory categories.
As noted above, claims 12-15 are directed toward a system (machine). As noted above, the claims fail to define a statutory system (machine). However, for the purpose of providing a complete examination, the examiner will treat claims 12-15 as though the claims recite a statutory system (machine).
Claim 1:
Step 2A, Prong 1:
The claim recites in part:
encoding the initial instances of the plurality of recorded input samples, according to an input map, from the training dataset to a plurality of respective nodes in an input layer of the plurality of layers of the neural network (mental process; As drafted and under its broadest reasonable interpretation, this limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, this limitation encompasses performing a judgment based upon the input samples and the input map, to encode the instances)
filtering the training dataset to exclude subsequent instances of one or more of the recorded input samples from one or more of the source recording devices encoded by the input map to the eliminated nodes of the sparsified neural network (mental process; As drafted and under its broadest reasonable interpretation, this limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, this limitation encompasses performing an evaluation to filter the training data to exclude certain instances)
encoding subsequent instances of the plurality of recorded input samples of the filtered training dataset to remaining nodes not eliminated in the input layer of the sparsified neural network to train the neural network in a subsequent training phase or to predict an output of the neural network in a prediction phase (mental process; As drafted and under its broadest reasonable interpretation, this limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, this limitation encompasses performing a judgement, based upon the instances of input samples that remain after filtering, to encode instances)
Step 2A, Prong 2:
The judicial exception is not integrated into a practical application.
The claim recites the additional elements:
receiving a neural network comprising a plurality of layers, each layer comprising a plurality of nodes, each node connected by a plurality of weights to a respective plurality of nodes in one or more different layers of the plurality of layers
receiving a training dataset comprising initial instances of a plurality of recorded input samples from one or more source recording devices
The additional elements amount to steps for data gathering, which is recited at a high level of generality and amounts to extra-solution activity of receiving data i.e. pre-solution activity of gathering data for use in the claimed process. The courts have found limitations directed to obtaining information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”, "electronic record keeping," and "storing and retrieving information in memory").
The claim recites the additional element:
sparsifying the neural network by eliminating one or more nodes in the input layer of the neural network during a training phase
The use of the neural network is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Accordingly, at Step 2A, prong two, the additional elements individually or in combination do no integrate the judicial exception into a practical application.
Step 2B:
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception.
The claim recites the additional elements:
receiving a neural network comprising a plurality of layers, each layer comprising a plurality of nodes, each node connected by a plurality of weights to a respective plurality of nodes in one or more different layers of the plurality of layers
receiving a training dataset comprising initial instances of a plurality of recorded input samples from one or more source recording devices
The additional elements amount to steps for data gathering, which is recited at a high level of generality and amounts to extra-solution activity of receiving data i.e. pre-solution activity of gathering data for use in the claimed process. The courts have found limitations directed to obtaining information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”, "electronic record keeping," and "storing and retrieving information in memory").
The claim recites the additional element:
sparsifying the neural network by eliminating one or more nodes in the input layer of the neural network during a training phase
The use of the neural network is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Accordingly, at Step 2B the additional elements individually or in combination do not amount to significantly more than the judicial exception.
Claim 2:
With respect to claim 2, the claim depends upon claim 1. The analysis of claim 1 is incorporated herein by reference.
Step 2A, Prong 1:
The claim recites in part:
reactivating the eliminated nodes by encoding the modified recorded input samples therein for further training or prediction (mental process; As drafted and under its broadest reasonable interpretation, this limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, this limitation encompasses performing a judgment to reconsider eliminated nodes by further encoding the modified recorded input samples)
Step 2A, Prong 2:
The judicial exception is not integrated into a practical application.
The claim recites the additional element
adjusting recording parameters of one or more source recording devices to modify one or more of the excluded recorded input samples encoded in the eliminated nodes
The additional elements amounts to data gathering, based upon adjusted parameters, which is recited at a high level of generality and amounts to extra-solution activity of receiving data i.e. pre-solution activity of gathering data for use in the claimed process. The courts have found limitations directed to obtaining information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”, "electronic record keeping," and "storing and retrieving information in memory").
Accordingly, at Step 2A, prong two, the additional elements individually or in combination do no integrate the judicial exception into a practical application.
Step 2B:
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception.
The claim recites the additional element
adjusting recording parameters of one or more source recording devices to modify one or more of the excluded recorded input samples encoded in the eliminated nodes
The additional elements amounts to data gathering, based upon adjusted parameters, which is recited at a high level of generality and amounts to extra-solution activity of receiving data i.e. pre-solution activity of gathering data for use in the claimed process. The courts have found limitations directed to obtaining information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”, "electronic record keeping," and "storing and retrieving information in memory").
Accordingly, at Step 2B the additional elements individually or in combination do not amount to significantly more than the judicial exception.
Claim 3:
As per dependent claim 3, the claim depends upon claim 1. The analysis of claim 1 is incorporated herein.
Step 2A, Prong 1:
The claim recites in part:
the training dataset is filtered to exclude subsequent instances directly encoded according to the input map, only in the eliminated nodes of the input layer (mental process; As drafted and under its broadest reasonable interpretation, this limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, this limitation encompasses performing an evaluation to filter the training data to exclude certain instances)
Step 2A, Prong 2:
The judicial exception is not integrated into a practical application.
The claim recites the additional element:
wherein sparsifying occurs in the input layer but not in hidden layers in the neural network
The use of the neural network is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Accordingly, at Step 2A, prong two, the additional elements individually or in combination do no integrate the judicial exception into a practical application.
Step 2B:
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception.
The claim recites the additional element:
wherein sparsifying occurs in the input layer but not in hidden layers in the neural network
The use of the neural network is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Accordingly, at Step 2B the additional elements individually or in combination do not amount to significantly more than the judicial exception.
Claim 4:
As per dependent claim 4, the claim depends upon claim 1. The analysis of claim 1 is incorporated herein.
Step 2A, Prong 1:
The claim recites in part:
wherein the training data is filtered to exclude subsequent instances encoded in a root input layer node that is not eliminated, which has a plurality of hidden layer nodes branching therefrom that are eliminated or have cumulative unsatisfactory strength of connection (mental process; As drafted and under its broadest reasonable interpretation, this limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, this limitation encompasses performing an evaluation to filter the training data to exclude certain instances)
Step 2A, Prong 2:
The judicial exception is not integrated into a practical application.
The claim recites the additional element:
sparsifying one or more hidden layers in the neural network
The use of the neural network is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Accordingly, at Step 2A, prong two, the additional elements individually or in combination do no integrate the judicial exception into a practical application.
Step 2B:
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception.
The claim recites the additional element:
sparsifying one or more hidden layers in the neural network
The use of the neural network is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Accordingly, at Step 2B the additional elements individually or in combination do not amount to significantly more than the judicial exception.
Claim 5:
As per dependent claim 5, the claim depends upon claim 4. The analysis of claim 4 is incorporated herein.
Step 2A, Prong 1:
The claim recites in part:
wherein the cumulative strength of connection of the plurality of branching hidden layer nodes is a weighted sum of neuron weights or channel filters along paths originating at the root input layer node and connecting the hidden layer nodes branching therefore (mental process; As drafted and under its broadest reasonable interpretation, this limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, this limitation encompasses performing an evaluation to determine a cumulate strength by calculate a weighted sum or observation of filters along the path)
Step 2A, Prong 2:
There are no additional elements considered under Step 2A, Prong 2.
Step 2B:
There are no additional elements considered under Step 2B.
Claim 6:
As per dependent claim 6, the claim depends upon claim 1. The analysis of claim 1 is incorporated herein.
Step 2A, Prong 1:
The claim recites in part:
wherein the training dataset is filtered by avoiding recording or storing the one or more recorded input samples at the one or more source recording devices (mental process; As drafted and under its broadest reasonable interpretation, this limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, this limitation encompasses performing an evaluation to perform filtering)
Step 2A, Prong 2:
There are no additional elements considered under Step 2A, Prong 2.
Step 2B:
There are no additional elements considered under Step 2B.
Claim 7:
As per dependent claim 7, the claim depends upon claim 1. The analysis of claim 1 is incorporated herein.
Step 2A, Prong 1:
The claim recites in part:
wherein the training dataset is filtered after recording by the one or more source recording devices by deleting the one or more recorded input samples at a receiver prior to encoding the training dataset into the input layer (mental process; As drafted and under its broadest reasonable interpretation, this limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, this limitation encompasses performing an evaluation to perform filtering, wherein the filtering includes removing the data from consideration)
Step 2A, Prong 2:
There are no additional elements considered under Step 2A, Prong 2.
Step 2B:
There are no additional elements considered under Step 2B.
Claim 8:
As per dependent claim 8, the claim depends upon claim 1. The analysis of claim 1 is incorporated herein.
Step 2A, Prong 1:
The claim is directed toward the abstract idea identified with respect to claim 1.
Step 2A, Prong 2:
The judicial exception is not integrated into a practical application.
The claim recites the additional element:
transmitting a signal indicating an error or poor quality training data to a device along a transmission path from the one or more source recording devices to a neural network training device
These limitations also amount to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data output (see MPEP 2106.05(g)). The courts have similarly found limitations directed to displaying a result, recited at a high level of generality, to be well-understood, routine, and conventional. See (MPEP 2106.05(d)(II), "presenting offers and gathering statistics.", “determining an estimated outcome and setting a price”).
Accordingly, at Step 2A, prong two, the additional elements individually or in combination do no integrate the judicial exception into a practical application.
Step 2B:
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception.
The claim recites the additional element:
transmitting a signal indicating an error or poor quality training data to a device along a transmission path from the one or more source recording devices to a neural network training device
These limitations also amount to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data output (see MPEP 2106.05(g)). The courts have similarly found limitations directed to displaying a result, recited at a high level of generality, to be well-understood, routine, and conventional. See (MPEP 2106.05(d)(II), "presenting offers and gathering statistics.", “determining an estimated outcome and setting a price”).
Accordingly, at Step 2B the additional elements individually or in combination do not amount to significantly more than the judicial exception.
Claim 9:
As per dependent claim 9, the claim depends upon claim 1. The analysis of claim 1 is incorporated herein.
Step 2A, Prong 1:
The claim recites in part:
marking a visualization of the training dataset to illustrate the excluded recorded input samples encoded by the input map to the eliminated nodes of the sparsified neural network (mental process; As drafted and under its broadest reasonable interpretation, this limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, this limitation encompasses performing an evaluation of a visualization of the training dataset and marking, with the aid of pencil and paper, to illustrated excluded recorded input samples)
Step 2A, Prong 2:
There are no additional elements considered under Step 2A, Prong 2.
Step 2B:
There are no additional elements considered under Step 2B.
Claim 10:
As per dependent claim 10, the claim depends upon claim 1. The analysis of claim 1 is incorporated herein.
Step 2A, Prong 1:
The claim is directed toward the abstract idea identified with respect to claim 1.
Step 2A, Prong 2:
The judicial exception is not integrated into a practical application.
The claim recites the additional element:
wherein the neural network is a convolutional neural network, the nodes represent channels of neurons, and each channel of neurons in the input layer is connected by a convolutional filter of weights to a channel of neurons in one or more different hidden layers
The use of the neural network is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Accordingly, at Step 2A, prong two, the additional elements individually or in combination do no integrate the judicial exception into a practical application.
Step 2B:
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception.
The claim recites the additional element:
wherein the neural network is a convolutional neural network, the nodes represent channels of neurons, and each channel of neurons in the input layer is connected by a convolutional filter of weights to a channel of neurons in one or more different hidden layers
The use of the neural network is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Accordingly, at Step 2B the additional elements individually or in combination do not amount to significantly more than the judicial exception.
Claim 11:
As per dependent claim 11, the claim depends upon claim 1. The analysis of claim 1 is incorporated herein.
Step 2A, Prong 1:
The claim is directed toward the abstract idea identified with respect to claim 1.
Step 2A, Prong 2:
The judicial exception is not integrated into a practical application.
The claim recites the additional element:
storing each of a plurality of nodes of the neural network with an association to a unique index, the unique index uniquely identifying the node, wherein only nodes with non-zero weights are stored that are not eliminated and nodes with zero weights are not stored that represent eliminated nodes
These limitations also amount to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data storage (see MPEP 2106.05(g)). The courts have similarly found limitations directed to displaying a result, recited at a high level of generality, to be well-understood, routine, and conventional. See (MPEP 2106.05(d)(II), "storing and retrieving information in memory”).
Accordingly, at Step 2A, prong two, the additional elements individually or in combination do no integrate the judicial exception into a practical application.
Step 2B:
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception.
The claim recites the additional element:
storing each of a plurality of nodes of the neural network with an association to a unique index, the unique index uniquely identifying the node, wherein only nodes with non-zero weights are stored that are not eliminated and nodes with zero weights are not stored that represent eliminated nodes
These limitations also amount to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data storage (see MPEP 2106.05(g)). The courts have similarly found limitations directed to displaying a result, recited at a high level of generality, to be well-understood, routine, and conventional. See (MPEP 2106.05(d)(II), "storing and retrieving information in memory”).
Accordingly, at Step 2B the additional elements individually or in combination do not amount to significantly more than the judicial exception.
Claim 12:
With respect to claim 12, the claim recites the limitations substantially similar to those in claim 1. The analysis of claim 1 is incorporated herein by reference.
Step 2A, Prong 1:
The claim recites the abstract idea identified with respect to claim 1.
Step 2A, Prong 2:
The judicial exception is not integrated into a practical application.
The claim recites the additional element:
one or more memories configured to store
The element is recited at a high-level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer component (See MPEP 2106.05(f)).
Accordingly, at Step 2A, prong two, the additional elements individually or in combination do no integrate the judicial exception into a practical application.
Step 2B:
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception.
The claim recites the additional element:
one or more memories configured to store
The element is recited at a high-level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer component (See MPEP 2106.05(f)).
Accordingly, at Step 2B the additional elements individually or in combination do not amount to significantly more than the judicial exception.
Claims 13-15:
With respect to claims 13-15, the claims recite the limitations substantially similar to those in claims 2 and 6-7, respectively. The analysis of claims 2 and 6-7 are incorporated herein by reference.
Claim 16:
With respect to claim 16, the claim recites the limitations substantially similar to those in claim 1. The analysis of claim 1 is incorporated herein by reference.
Step 2A, Prong 1:
The claim recites the abstract idea identified with respect to claim 1.
Step 2A, Prong 2:
The judicial exception is not integrated into a practical application.
The claim recites the additional element:
a non-transitory computer-readable storage medium having instructions stored thereon, which when executed, cause one or more processors to
The element is recited at a high-level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer component (See MPEP 2106.05(f)).
Accordingly, at Step 2A, prong two, the additional elements individually or in combination do no integrate the judicial exception into a practical application.
Step 2B:
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception.
The claim recites the additional element:
a non-transitory computer-readable storage medium having instructions stored thereon, which when executed, cause one or more processors to
The element is recited at a high-level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer component (See MPEP 2106.05(f)).
Accordingly, at Step 2B the additional elements individually or in combination do not amount to significantly more than the judicial exception.
Claims 17-20:
With respect to claims 17-20, the claims recite the limitations substantially similar to those in claims 2-5, respectively. The analysis of claims 2-5 are incorporated herein by reference.
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.
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.
Claims 1, 3-5, 7-8, 10-12, 15-16, and 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over Yan et al. (US 10832139, patented 10 November 2020, hereafter Yan) and further in view of Liu et al. (US 2021/0026446, published 28 January 2021, hereafter Liu), and further in view of Gong et al. (WO 2022/265875, published 22 December 2022, hereafter Gong).
As per independent claim 1, Yan discloses a method for filtering a neural network training dataset, the method comprising:
receiving a neural network comprising a plurality of layers, each layer comprising a plurality of nodes, each node connected by a plurality of weights to a respective plurality of nodes in one or more different layers of the plurality of layers (Figure 2, item 220; Figure 5; column 7, lines 37-46: Here, a dense deep neural network (DNN) is received that includes a plurality of nodes connected to other nodes in one or more different layers by a plurality of weights)
receiving a training dataset comprising initial instances of a plurality of recorded input samples from one or more source recording devices (Figure 2, item 210; column 5, lines 31-45: Here, a training data set is received with correct input-output pairs for a DNN to measure and improve the predictive performance of the DNN on a specific task of a particular loss function)
encoding the initial instances of the plurality of recorded input samples from the training dataset to a plurality of respective nodes in an input layer of the plurality of layers of the neural network (Figure 3; column 5, line 46- column 6, line 12: Here, the training data and loss function and the pre-trained DNN are provided to iteratively sparsify the DNN through activation compressor (Figure 2). This includes alternating between gradient evaluation and parameter update until a termination criteria. This occurs by augmenting the loss function with activation regularization, a function that measures the activation sparsification level of the DNN with respect to the training dataset (column 3, lines 11-14))
sparsifying the neural network by eliminating one or more nodes in the input layer of the neural network during a training phase (Figure 5; column 7, line 49- column 8, line 39: Here, a dense DNN is sparsified by removing one or more nodes in the input layer)
Yan fails to specifically disclose:
encoding the initial instances of the plurality of recorded input samples, according to an input map
filtering the training dataset to exclude subsequent instances of one or more of the recorded input samples from one or more of the source recording devices encoded by the input map to the eliminated nodes of the sparsified neural network
encoding subsequent instances of the plurality of recorded input samples of the filtered training dataset to remaining nodes not eliminated in the input layer of the sparsified neural network to train the neural network in a subsequent training phase or to predict an output of the neural network in a prediction phase
However, Liu, which is analogous to the claimed invention because it is directed toward filtering training data, discloses:
filtering the training dataset to exclude subsequent instances of one or more of the recorded input samples from one or more of the source recording devices (Figures 1 and 4; paragraphs 0015-0016: Here, noise removal is performed (filtering) on the first training dataset to generate a third training dataset by deleting, training data. This is performed by training the neural network using the first training dataset, calculating a first loss value associated with each of the items of training data using the neural network, deleting items based on the on the greatest loss value)
encoding subsequent instances of the plurality of recorded input samples of the filtered training dataset to remaining nodes not eliminated to train the neural network in a subsequent training phase or to predict an output of the neural network in a prediction phase (Figure 4; paragraph 0096: Here, the neural network is pruned (sparsified) and trained based upon the third training dataset)
It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Liu with Yan, with a reasonable expectation of success, as it would have allowed for removing noise from a dataset to improve the subsequent training of models (Liu: paragraph 0015).
Additionally, Gong, which is analogous to the claimed invention because it is directed toward filtering data, discloses:
encoding the initial instances of the plurality of recorded input samples, according to an input map (paragraph 0090: Here, during a training process, the weights or parameters of the neural network ate tuned to approximate the ground truth data, thereby learning a mapping of data items to the neural network)
encoded by the input map to the eliminated nodes of the sparsified neural network (paragraph 0090 and 0098: Here, some of the input data is low-quality data (paragraph 0090) and model tuning may include deleting nodes of the trained neural network that do not affect output)
It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Gong with Yan-Liu, with a reasonable expectation of success, as it would have allowed for deleting noisy data associated with low-quality video data corresponding to nodes of the neural network (Gong: paragraph 0098).
As per dependent claim 3, Yan, Liu, and Gong disclose the limitations similar to those in claim 1, and the same rejection is incorporated herein. Liu discloses the training dataset is filtered to exclude subsequent instances in the eliminated nodes devices (Figures 1 and 4; paragraphs 0015-0016: Here, noise removal is performed (filtering) on the first training dataset to generate a third training dataset by deleting, training data. This is performed by training the neural network using the first training dataset, calculating a first loss value associated with each of the items of training data using the neural network, deleting items based on the on the greatest loss value)
It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Liu with Yan, with a reasonable expectation of success, as it would have allowed for removing noise from a dataset to improve the subsequent training of models (Liu: paragraph 0015).
Further, Gong discloses:
wherein the sparsifying occurs in the input layer (paragraph 0098: Here, model tuning is used to delete nodes) but not in hidden layers in the neural network (paragraph 0090: Here, the hidden layers are used as learnable feature extractors)
instances directly encoded according to the input map (paragraph 0090)
It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Gong with Yan-Liu, with a reasonable expectation of success, as it would have allowed for deleting noisy data associated with low-quality video data corresponding to nodes of the neural network (Gong: paragraph 0098).
As per dependent claim 4, Yan, Liu, and Gong disclose the limitations similar to those in claim 1, and the same rejection is incorporated herein. Yan discloses sparsifying layers in the neural network (Figure 5; column 7, line 49- column 8, line 39)
wherein the training data is filtered to exclude subsequent instances encoded in a root input layer node that is not eliminated, which has a plurality of layer nodes branching therefrom that are eliminated or have cumulative unsatisfactory strength of connection (Figure 3; column 5, line 46- column 6, line 12: Here, the training data and loss function and the pre-trained DNN are provided to iteratively sparsify the DNN through activation compressor (Figure 2). This includes alternating between gradient evaluation and parameter update until a termination criteria. This occurs by augmenting the loss function with activation regularization, a function that measures the activation sparsification level of the DNN with respect to the training dataset (column 3, lines 11-14). Additionally, a dense DNN is sparsified by removing one or more nodes in the input layer (Figure 5; column 7, line 49- column 8, line 39). This includes eliminating (removing) nodes branching from a root)
Yan fails to specifically disclose a hidden layer. However, Gong discloses a hidden layer (paragraph 0090: Here, the hidden layers are used as learnable feature extractors). It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Gong with Yan-Liu, with a reasonable expectation of success, as it would have allowed for deleting noisy data associated with low-quality video data corresponding to nodes of the neural network (Gong: paragraph 0098).
As per dependent claim 5, Yan, Liu, and Gong disclose the limitations similar to those in claim 4, and the same rejection is incorporated herein. Gong discloses wherein the cumulative strength of connection of the plurality of branching hidden layer nodes is a weighted sum of neuron weights or channel filters along paths originating at the root node layer and connecting the hidden layer nodes branching therefore (paragraph 0090: Here, a connection from an input to a neuron is determined by summing the products of all pairs of inputs and their associated weights).
Yan fails to specifically disclose a hidden layer. However, Gong discloses a hidden layer (paragraph 0090: Here, the hidden layers are used as learnable feature extractors). It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Gong with Yan-Liu, with a reasonable expectation of success, as it would have allowed for deleting noisy data associated with low-quality video data corresponding to nodes of the neural network (Gong: paragraph 0098).
As per dependent claim 7, Yan, Liu, and Gong disclose the limitations similar to those in claim 1, and the same rejection is incorporated herein. Liu discloses wherein the training dataset is filtered after recording by the one or more source recording devices (paragraph 0041: Here, a gaze tracking apparatus predicts position information based upon receiving a plurality of recorded face image frames) by deleting the one or more recorded input samples at a receiver prior to encoding the training dataset into the input layer (Figures 1 and 4; paragraphs 0015-0016: Here, noise removal is performed (filtering) on the first training dataset to generate a third training dataset by deleting, training data. This is performed by training the neural network using the first training dataset, calculating a first loss value associated with each of the items of training data using the neural network, deleting items based on the on the greatest loss value)
It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Liu with Yan, with a reasonable expectation of success, as it would have allowed for removing noise from a dataset to improve the subsequent training of models (Liu: paragraph 0015).
As per dependent claim 8, Yan, Liu, and Gong disclose the limitations similar to those in claim 1, and the same rejection is incorporated herein. Yang discloses transmitting data to a device along a transmission path from the one or more source devices to a neural network training device (Figure 2, item 210; column 5, lines 31-45: Here, a training data set is received with correct input-output pairs for a DNN to measure and improve the predictive performance of the DNN on a specific task of a particular loss function. Additionally, a device at which the neural network is trained is a device “along a transmission path from the one or more source devices to a neural network training device”).
Further, Gong discloses transmitting a signal indicating an error or poor quality training data to a device along a transmission path from the one or more source devices to a neural network training device (paragraph 0090: Here, ground truth data distinguishes between low (poor) quality video training data and high quality video training data). It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Gong with Yan-Liu-Gong, with a reasonable expectation of success, as it would have allowed of identifying the quality of the ground truth data used for training (Gong: paragraph 0090).
As per dependent claim 10, Yan, Liu, and Gong disclose the limitations similar to those in claim 1, and the same rejection is incorporated herein Yan discloses a deep neural network, the nodes represent channels of neurons, and each channel of neurons in the input layer is connected by a weight to a channel of neurons in one or more different layers (Figure 5; column 7, line 49- column 8, line 39).
Yan fails to specifically disclose a convolutional neural network, and each channel of neurons in the input layer is connected by a convolutional filter of weights to a channel of neurons in one or more different hidden layers.
However, Gong discloses a convolutional neural network and each channel of neurons in the input layer is connected by a convolutional filter of weights to a channel of neurons in one or more different hidden layers (paragraph 0090: Here, a convolutional neural network (CNN) includes an input layer, a number of hidden layers, and an output layer. Additionally, a plurality of weights are associated with the connection between neurons, and in some cases, the weight may be the summation of all pairs of inputs and their associated weights). It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Gong with Yan-Liu, with a reasonable expectation of success, as it would have allowed for deleting noisy data associated with low-quality video data corresponding to nodes of the neural network (Gong: paragraph 0098).
As per dependent claim 11, Yan, Liu, and Gong disclose the limitations similar to those in claim 1, and the same rejection is incorporated herein. Yan discloses storing each of a plurality of the nodes of the neural network with an association to a unique index, the unique index uniquely identifying the node, wherein only nodes with non-zero weights are stored that are not eliminated and nodes with zero weights are not stored the at represent eliminated nodes (Figure 5; column 7, line 49- column 8, line 39: Here, sparse nodes include an index-value pair. Only nodes that remain (not eliminated) after sparsification are included in the sparse list of index-value pairs. Nodes that have been eliminated during sparsification are not contained in the list).
With respect to claim 12, the claim recites the limitations substantially similar to those in claim 1. The rejection of claim 1 is incorporated herein by reference.
Additionally, Yan discloses a system for efficiently storing a sparse neural network, the system comprising one or more memories configure to store (column 9, lines 18-22).
With respect to claim 15, the claim recites the limitations substantially similar to those in claim 7. The rejection of claim 7 is incorporated herein by reference.
With respect to claim 16, the claim recites the limitations substantially similar to those in claim 1. The rejection of claim 1 is incorporated herein by reference.
Additionally, Yan discloses a non-transitory computer-readable storage medium having instructions stored thereon, which when executed, cause one or more processors to perform operations (column 9, lines 40-52).
With respect to claims 18-20, the claim recites the limitations substantially similar to those in claims 3-5, respectively. The rejection of claims 3-5 are incorporated herein by reference.
Claims 2, 13, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Yan, Liu, and Gong and further in view of Warlock et al. (US 9128783, patented 8 September 2015, hereafter Warlock).
As per dependent claim 2, Yan, Liu, and Gong disclose the limitations similar to those in claim 1, and the same rejection is incorporated herein. Yan fails to specifically disclose:
adjusting recording parameters of one or more of the source recording devices to modify one or more of the excluded recorded input samples encoded in the eliminated nodes
reactivating the eliminated nodes by encoding the modified recording input samples therein for further training or prediction
However, Warlock, which is analogous to the claimed invention because it is directed toward manually editing parameters, discloses:
adjusting recording parameters of one or more of the source recording devices to modify one or more of the excluded recorded input samples encoded in the data (Figure 1C; column 3, lines 15-42: Here, a user modifies a recording parameter associated with a model execution (column 1, lines 41-57) of the one or more recorded parameters.)
reactivating the data by encoding the modified recording input samples therein for further training or prediction (Figure 5A; column 3, lines 15-42: Here, based upon the modified parameters, the model is re-executed with the modified parameter value)
It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Warlock with Yan-Liu-Gong, with a reasonable expectation of success, as it would have allowed for editing data and re-executing based upon the edited data (Warlock: column 3, lines 15-42).
With respect to claims 13 and 17, the claim recites the limitations substantially similar to those in claim 2. The rejection of claim 2 is incorporated herein by reference.
Claims 6 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Yan, Liu, and Gong and further in view of Kliger et al. (US 12525240, filed 17 April 2023, hereafter Kliger).
As per dependent claim 6, Yan, Liu, and Gong disclose the limitations similar to those in claim 1, and the same rejection is incorporated herein. Yan fails to disclose wherein the training dataset is filtered by avoiding recording or storing the one or more recorded input samples at one or more source recording device.
However, Kliger, which is analogous to the claimed invention because it is directed toward filtering data, discloses wherein the training dataset is filtered by avoiding recording or storing the one or more recorded input samples at one or more source recording device (column 17, line 38- column 18, line 3: Here, the communication device captures signal data and performs filtering to remove interfering signals before storing the data). It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Kliger with Yan-Liu-Gong, with a reasonable expectation of success, as it would have allowed for filtering data prior to storing, thereby avoiding storing interfering signals (Kliger: column 17, line 38- column 18, line 3).
With respect to claim 14, the claim recites the limitations substantially similar to those in claim 6. The rejection of claim 6 is incorporated herein by reference.
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Yan, Liu, and Gong and further in view of Rossi et al. (US 2022/0300836, published 22 September 2022, hereafter Rossi).
As per dependent claim 9, Yan, Liu, and Gong disclose the limitations similar to those in claim 1, and the same rejection is incorporated herein. Yan fails to specifically disclose marking a visualization of the training dataset to illustrate the excluded recorded input samples encoded by the input map to the eliminated nodes of the sparsified neural network.
However, Rossi, which is analogous to the claimed invention because it is directed toward generating a visualization, discloses marking a visualization of the training dataset to illustrate the excluded recorded input samples encoded by the input map to the eliminated nodes of the sparsified neural network (paragraph 0039: Here, an attribute feature-extraction module generates the sparse configuration attribute set. This visualization of the sparse configuration illustrates the excluded recorded input samples by eliminating them from the visualization). It would have been obvious to one of ordinary skill in the art at the time of the applicant’s effective filing date to have combined Rossi with Yan-Liu-Gong, with a reasonable expectation of success, as it would have allowed for generating a visualization of the sparsified data (Rossi: paragraph 0039).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Choudhury et al. (US 2021/0174175): Discloses pruning nodes of a neural network (paragraph 0026)
Pandaya et al. (US 11675878): Discloses pruning a convolutional neural network nodes using a mapping (column 21, line 60- column 22, line 31)
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KYLE R STORK whose telephone number is (571)272-4130. The examiner can normally be reached 8am - 2pm; 4pm - 6pm.
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, Omar Fernandez Rivas can be reached at 571/272-2589. 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.
/KYLE R STORK/Primary Examiner, Art Unit 2128
Prosecution Insights