DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
Claim Rejections - 35 USC § 101
Applicants’ amendments and remarks filed on 07/20/2026 have been fully considered but they are not persuasive
Applicant argues that claim 1, is patent eligible because “Amended claim 1 recites limitations that cannot be practically performed in the human mind. Amended claim 1 requires optimizing memory usage of an artificial neural network (ANN) graph including layers and tensors by determining, for each layer, a tensor working set defined using liveness analysis to provide tensors that need to be simultaneously allocated or existing in memory while processing the layer; determining whether a working set exceeds a predetermined memory threshold; identifying the working set and layer responsible for the excess memory usage; and pruning the identified layer. These operations require analysis of computer data structures representing an ANN graph, tensors, layer-by-layer memory consumption, tensor liveness, simultaneous memory allocation, threshold-based memory usage, and pruning of a layer of the ANN graph. Such steps are not practically performable in the human mind, at least because they require tracking and analyzing computer memory allocation/liveness relationships for tensors in an ANN graph, comparable to Example 37's processor-based tracking of memory allocation.
Accordingly, amended claim 1 does not merely invoke a generic computer to automate a mental comparison. Rather, the claim recites computer-specific operations directed to determining and reducing memory usage in an ANN graph based on tensor liveness and layer-specific working sets.”
Examiner respectfully disagrees. Defining a tensor working set as using liveness analysis to provide tensors processing memory in a layer does not take tensors out of being considered a generic computer component. Applicant refers to Example 37 to support their argument and examiner respectfully disagrees. The use of liveness analysis here aligns closer to Claim 1’s 2A-Prong 1 analysis in Example 37 which is labeled as a mental process being done with a generic computer component, and there are no additional limitations present in claim 1 that would lead to an improvement over prior systems unlike in Example 37 which points to an improvement over prior systems.
Claim Rejections - 35 USC § 112
Applicants’ amendments filed on 07/20/2026 have been considered and overcomes the 112(B) on Claims 9,10, and 13 rejections
Applicant’s arguments, filed 07/20/2026, with respect to Claim 3 have been fully considered and are persuasive. The 112(B) rejection of Claim 3 has been withdrawn.
Applicants’ arguments filed 07/20/2026 have been fully considered but they are not persuasive.
For claim 7 applicant argues that the term maximum is not a relative term for similar reasons as claim 3 because “The Specification expressly states that "peak memory usage may be understood as a maximum usage of memory of the artificial neural network." Specification, page 4, 11. 10-11.”
Examiner respectfully disagrees. It is still unclear what maximum usage is when it comes to the artificial neural network as it is with claim 3. Applicant points towards their specification that expressly states that “peak memory usage may be understood as a maximum usage of memory of the artificial neural network”, this is not a definite definition of what maximum usage is when it comes to claim 7, making it relative term.
Claim Rejections - 35 USC § 102
Applicants’ amendments to the claims have been considered and overcomes the 102 rejection on Claim 1, 14, and 15
Claim Rejections - 35 USC § 103
Applicant’s arguments with respect to claim(s) 1-15 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
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 7 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.
The term “maximum” in claim 7 is a relative term which renders the claim indefinite. The term “maximum” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention.
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.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim’s 1-15 are rejected as being directed to an abstract idea without significantly more.
Step 2A prong 1: for each of the plurality of layers, determining a tensor working set, wherein: the tensor working set includes tensors that consume memory with respect to the respective layer, and the tensor working set is defined using liveness analysis to provide tensors that need to be simultaneously allocated or existing in the memory while processing the layer; (Determining a tensor consuming memory is a mental process that can be done with the aid of pen and paper, with a working set being tensors existing in memory while processing the layer a person can observe tensors and see if it is consuming memory) ,
determining whether that at least one working set of the plurality of working sets requires memory usage above a pre-determined threshold; (Determining and identifying a working set of tensors consuming memory above a threshold is a mental process aid of pen and paper)
and in response to a determination that at least one working set of the plurality of working sets requires memory usage above the pre-determined threshold, threshold: identifying a working set of the plurality of working sets which requires memory usage above the pre-determined threshold; (Identifying working sets that requires memory usage above a threshold is a mental process that can be done with the aid of pen and paper)
identifying at least one layer responsible for the memory usage above the pre-determined threshold in the identified working set (The process of identifying a layer responsible for memory usage above a threshold in the working set is a mental process that can be done with the aid of pen and paper).
Step 2A prong 2: The additional limitations pruning the identified at least one layer is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – examiners note: high level recitation of pruning a layer
Step 2B: The additional elements, taken either alone or in combination with other limitations of the claim, do not amount to significantly more than the abstract idea itself. The additional limitation pruning the identified at least one layer is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – examiners note: high level recitation of pruning a layer
Regarding Claim 2
Step 1: A process as above.
Step 2A Prong 1: Recites the abstract ideas of Claim 1
Step 2A Prong 2: The additional limitation wherein the identifying and pruning are repeated until every working set of the plurality of working sets requires memory below the pre- determined threshold. is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – examiners note: high level recitation of repeated pruning below a threshold
Step 2B: The additional elements, taken either alone or in combination with other limitations of the claim, do not amount to significantly more than the abstract idea itself. The additional limitation of wherein the identifying and pruning are repeated until every working set of the plurality of working sets requires memory below the pre- determined threshold. is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – examiners note: high level recitation of repeated pruning below a threshold
Regarding Claim 3
Step 1: A process as above
Step 2A Prong 1: wherein in each step of identifying, the working set which requires a highest amount of memory is identified is a mental process that can be done with the aid of pen and paper, a person can observe a working set to identify what requires the highest amount of memory.
Step 2A Prong 2 & Step 2B: There are no additional elements recited so the claim does
not provide a practical application and is not considered to be significantly more. As such, the claim is patent ineligible.
Regarding Claim 4
Step 1: A process as above
Step 2A Prong 1: The claim recites the abstract idea claim 1
Step 2A Prong 2: The additional limitations of wherein each layer includes a respective plurality of channels is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. See MPEP §§ 2106.04(d), 2106.05(h).
and pruning the at least one identified layer includes reducing a number of channels of the at least one identified layer is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – Examiner’s note: high level recitation of pruning channels identified the layer
Step 2B: The additional elements, taken either alone or in combination with other limitations of the claim, do not amount to significantly more than the abstract idea itself. The additional limitations of each layer includes a respective plurality of channels is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. See MPEP §§ 2106.04(d), 2106.05(h).
and pruning the at least one identified layer includes reducing a number of channels of the at least one identified layer.is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – Examiner’s note: high level recitation of pruning channels identified the layer
Regarding Claim 5
Step 1: A process as above
Step 2A Prong 1: The claim recites the abstract idea claim 1
Step 2A Prong 2: The additional limitation of wherein pruning the at least one identified layer includes removing the at least one identified layer is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – examiners note: high level recitation of pruning a layer
Step 2B: The additional elements, taken either alone or in combination with other limitations of the claim, do not amount to significantly more than the abstract idea itself. The additional limitation of wherein pruning the at least one identified layer includes removing the at least one identified layer is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – examiners note: high level recitation of pruning a layer
Regarding Claim 6
Step 1: A process as above.
Step 2A Prong 1: wherein the working set of the plurality of working sets which requires maximum memory usage is determined based on an architecture of the artificial neural network graph. is a mental process that can be done with the aid of pen and paper, a person can determine memory usage with the aid of a graph
Step 2A Prong 2: The additional limitations of wherein the working set of the plurality of working sets which requires maximum memory usage is determined based on an architecture of the artificial neural network graph. is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. See MPEP §§ 2106.04(d), 2106.05(g) – examiners note: using an artificial neural network graph to determine maximum memory usage amounts to nothing more than mere data gathering
Step 2B: The additional elements, taken either alone or in combination with other limitations of the claim, do not amount to significantly more than the abstract idea itself. The additional limitation of wherein the working set of the plurality of working sets which requires maximum memory usage is determined based on an architecture of the artificial neural network graph is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. See MPEP §§ 2106.04(d), 2106.05(g) – examiners note: using an artificial neural network graph to determine maximum memory usage amounts to nothing more than mere data gathering
Regarding Claim 7
Step 1: A process as above.
Step 2A Prong 1: determining an intermediate representation of the artificial neural network graph wherein the working set of the plurality of working sets which requires maximum memory usage is determined based on the intermediate representation is a mental process that can be done with the aid of pen and paper a human can observe the intermediate representation and determine what set is consuming the maximum memory usage.
Step 2A Prong 2 & Step 2B: There are no additional elements recited so the claim does
not provide a practical application and is not considered to be significantly more. As such, the claim is patent ineligible.
Regarding Claim 8
Step 1: A process as above.
Step 2A Prong 1: The claim recites the ideas of claim 1
Step 2A Prong 2: The additional limitations wherein once every working set of the plurality of working sets requires memory below the pre-determined threshold, the artificial neural network graph after pruning is re-trained from scratch or fine-tuned from a previous training is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – examiners note: high level recitation of retraining a model after pruning
Step 2B: The additional elements, taken either alone or in combination with other limitations of the claim, do not amount to significantly more than the abstract idea itself. The additional limitation of wherein once every working set of the plurality of working sets requires memory below the pre-determined threshold, the artificial neural network graph after pruning is re-trained from scratch or fine-tuned from a previous training is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – examiners note: high level recitation of retraining a model after pruning
Regarding Claim 9
Step 1: A process as above.
Step 2A Prong 1: The claim recites the abstract ideas of claim 1.
Step 2A Prong 2: The additional limitation wherein: the at least one identified layer is pruned based on an importance metric, and the importance metric is provided by user input. is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – examiners note: high level recitation using a user inputted metric to identify a layer
Step 2B: The additional elements, taken either alone or in combination with other limitations of the claim, do not amount to significantly more than the abstract idea itself. The additional limitation of wherein: the at least one identified layer is pruned based on an importance metric, and the importance metric is provided by user input is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2) – examiners note: high level recitation using a user inputted metric to identify a layer
Regarding Claim 10
Step 1: A process as above.
Step 2A Prong 1 wherein: the importance metric is evaluated based on representative test data is a mental process that can be done with the aid of pen and paper a person can evaluate a metric with data.
Step 2A Prong 2: The additional limitation and the computer implemented method further comprises training the artificial neural network graph before evaluating the importance metrics. are an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception and reciting only the idea of a solution or outcome, i.e., the claim fails to recite details of how a solution to a problem is accomplished because it is unclear how the “AI system” or “machine learning” is used nor the specification makes it clear how these actions are performed. Thus, these additional elements are recited in a manner that represent no more than mere instructions to apply the judicial exceptions on a computer. See MPEP § 2106.05(f) and § 2106.04(d) – examiners note: high level recitation of training a neural network graph before evaluating a metric
Step 2B: The additional elements, taken either alone or in combination with other limitations of the claim, do not amount to significantly more than the abstract idea itself. The additional limitation and the computer implemented method further comprises training the artificial neural network graph before evaluating the importance metrics. are an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception and reciting only the idea of a solution or outcome, i.e., the claim fails to recite details of how a solution to a problem is accomplished because it is unclear how the “AI system” or “machine learning” is used nor the specification makes it clear how these actions are performed. Thus, these additional elements are recited in a manner that represent no more than mere instructions to apply the judicial exceptions on a computer. See MPEP § 2106.05(f) and § 2106.04(d) – examiners note: high level recitation of training a neural network graph before evaluating a metric
Regarding Claim 11
Step 1: A process as above.
Step 2A Prong 1: generating a report including at least one of a layer summary report, a tensor summary report, or a working set summary report is a mental process that can be done with the aid of pen and paper, a person can make a report on a layer summary, a tensor summary, or a working set summary.
Step 2A Prong 2 & Step 2B: There are no additional elements recited so the claim does
not provide a practical application and is not considered to be significantly more. As such, the claim is patent ineligible.
Regarding Claim 12
Step 1: A process as above.
Step 2A Prong 1: The claim recites the ideas of claim 1.
Step 2A Prong 2: The additional limitation wherein at least one of the artificial neural network or the pre-determined threshold are provided by user input. is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. See MPEP §§ 2106.04(d), 2106.05(g) – examiners note: providing an input for a threshold amounts to nothing more than selecting a particular data source
Step 2B: The additional elements, taken either alone or in combination with other limitations of the claim, do not amount to significantly more than the abstract idea itself. The additional limitation wherein at least one of the artificial neural network or the pre-determined threshold are provided by user input is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. See MPEP §§ 2106.04(d), 2106.05(g) – examiners note: providing an input for a threshold amounts to nothing more than selecting a particular data source
Regarding Claim 13
Step 1: A process as above.
Step 2A Prong 1: The claim recites the ideas of claim 1.
Step 2A Prong 2: The additional limitation wherein: the artificial neural network graph is to be deployed on a resource-constrained embedded system after pruning is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. See MPEP §§ 2106.04(d), 2106.05(h).
And the embedded system is a mobile computing device, a mobile phone, a tablet computing device, an automotive compute platform, or an edge device. is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. See MPEP §§ 2106.04(d), 2106.05(h).
Step 2B: The additional elements, taken either alone or in combination with other limitations of the claim, do not amount to significantly more than the abstract idea itself. The additional limitation wherein: the artificial neural network graph is to be deployed on a resource-constrained embedded system after pruning is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. See MPEP §§ 2106.04(d), 2106.05(h).
And the embedded system is a mobile computing device, a mobile phone, a tablet computing device, an automotive compute platform, or an edge device. is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. See MPEP §§ 2106.04(d), 2106.05(h).
Regarding Claim 14:
Step 1: A process as above.
Step 2A Prong 1: See the analysis of Claim 1.
Step 2A Prong 2: See the analysis of Claim 1
Step 2B: See the analysis of Claim 1
Regarding Claim 15:
Step 1: A process as above.
Step 2A Prong 1: See the analysis of Claim 1.
Step 2A Prong 2: See the analysis of Claim 1
Step 2B: See the analysis of Claim 1
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.
Claims 1, 11, 14-15 are rejected under USC 103 as being unpatentable over Mateev et al (US 20200160182) (“Mateev”), in view of Wang et al (NPL: SuperNeurons: Dynamic GPU Memory Management for Training Deep Neural Networks)
Regarding Claim 1, Mateev teaches determining whether that at least one working set of the plurality of working sets requires memory usage above a pre-determined threshold ([0102] teaches kernels computation not exceeded a cache memory threshold which functions as a pre-determined threshold)
and in response to a determination that at least one working set of the plurality of working sets requires memory usage above the pre-determined threshold, threshold: identifying a working set of the plurality of working sets which requires memory usage above the pre-determined threshold ([0161] Teaches a conditional statement where when the kernels are large they would be pruned to fit below the memory threshold);
identifying at least one layer responsible for the memory usage above the pre-determined threshold in the identified working set; and pruning the identified at least one layer ([0161] teaches the limitation of identifying layers with the identification of kernels above the memory threshold and those kernels go through sparsification which is another word for pruning.)
Mateev does not teach for each of the plurality of layers, determining a tensor working set, wherein: the tensor working set includes tensors that consume memory with respect to the respective layer, and the tensor working set is defined using liveness analysis to provide tensors that need to be simultaneously allocated or existing in the memory while processing the layer;
However, Wang does teach for each of the plurality of layers, determining a tensor working set, wherein: the tensor working set includes tensors that consume memory with respect to the respective layer, and the tensor working set is defined using liveness analysis to provide tensors that need to be simultaneously allocated or existing in the memory while processing the layer; ([Page 2 Column 1: Lines 1-2) teaches a plurality of layer. (Page 2 Column 1: Lines 9-14) teaches a plurality of tensors). [Page 2 Colum 1: 36-38] teaches using liveness analysis for large tensor allocations/deallocations which teaches this limitation)
Wang and Mateev are analogous art because they both center managing memory usage in a neural network
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Mateev with the liveness analysis of Wang. Doing so would lead to deeper neural network training ([Wang-Abstract])
Regarding Claim 11 Wang and Mateev teaches all the limitations of Claim 1.
Wang also teaches generating a report including at least one of a layer summary report, a tensor summary report, or a working set summary report ([Figure 8] teaches a memory usage breakdown of the different layer types, which is being interpreted as a layer summary report and [Figure 10] teaches an evaluation graph of liveness analysis, which is being interpreted as a working set summary report teaching this limitation
Regarding Claim 14, see the analysis of Claim 1
Regarding Claim 15, see the analysis of Claim 1
Claims 2, 3, and 5 are rejected under USC 103 as being unpatentable over Mateev et al (US 20200160182) (“Mateev”), in view of Wang et al (NPL: SuperNeurons: Dynamic GPU Memory Management for Training Deep Neural Networks), and Ji et al (US 20180232640) (“Ji”).
Regarding Claim 2, Mateev and Wang teaches all the limitations of Claim 1
Mateev does not teach wherein the identifying and pruning are repeated until every working set of the plurality of working sets requires memory below the pre- determined threshold.
However, Ji teaches wherein the identifying and pruning are repeated until every working set of the plurality of working sets requires memory below the pre- determined threshold ([0051] teaches iteration-based pruning below a threshold)
Mateev, Wang, and Ji are analogous art because they all deal with neural networks in relation to a threshold
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Mateev with the liveness analysis of Wang and the repeated pruning of Ji. Doing so would lead to lower computational loads and lower power consumption ([Ji-0003])
Regarding Claim 3, Mateev, Wang and Ji teaches all the limitations of Claim 2
Wang also teaches wherein in each step of identifying, the working set which requires a highest amount of memory is identified ([4.1.1] teaches liveness analysis affecting memory usages and live tensors counts, it can be reasonably interpreted that for the liveness analysis to optimize memory usage that it is identifying the tensor with the highest amount of memory which teaches this limitation)
Regarding Claim 5, Mateev, Wang, and Ji teaches all the limitations of Claim 1
Ji also teaches wherein pruning the at least one identified layer includes removing the at least one identified layer ([Abstract] teaches pruning a layer in a neural network. The pruning process have you remove components of a neural network so pruning a layer is being interpreted as removing the layer.)
Claim 4 and 8 are rejected under USC 103 as being unpatentable over Mateev et al (US 20200160182) (“Mateev”), in view of Wang et al (NPL: SuperNeurons: Dynamic GPU Memory Management for Training Deep Neural Networks), and Zhou et al (NPL: Accelerating Large Scale Real-Time GNN Inference using Channel Pruning (“Zhou”)
Mateev and Wang teaches all the limitations of Claim 1
Mateev does not teach wherein each layer includes a respective plurality of channels
and pruning the at least one identified layer includes reducing a number of channels of the at least one identified layer
However, Zhou does teach wherein each layer includes a respective plurality of channels ([Page 2 Column 2: Lines 4-6] teaches input channels for GNN layers)
and pruning the at least one identified layer includes reducing a number of channels of the at least one identified layer ([Page 2 Column 2: Lines 4-6] teaches pruning input channels for GNN layers).
Mateev, Wang, and Zhou are analogous art because they all deal with neural networks and memory usage.
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Mateev with the liveness analysis of Wang and the retraining of Zhou. Doing so would lead to a more efficient GNN ([Zhou-Abstract])
Regarding Claim 8, Mateev and Wang teaches all the limitations of Claim 1
Wang does teach wherein once every working set of the plurality of working sets requires memory below the pre-determined threshold ([Page 4 Column 2: Lines 15 to 18] teaches liveness analysis being used to reduce the baseline peak memory by 50% by recycling tensors. The baseline peak memory is being interpreted as a pre-determined threshold)
Mateev does not teach the artificial neural network graph after pruning is re-trained from scratch or fine-tuned from a previous training
However, Zhou does teach wherein once every working set of the plurality of working sets requires memory below the pre-determined threshold, the artificial neural network graph after pruning is re-trained from scratch or fine-tuned from a previous training ([4.1] Zhou teaches retraining of the pruned methods, and the pruning is being executed on a GNN which teaches limitation)
Claims 6 and 7 are rejected under USC 103 as being unpatentable over Mateev et al (US 20200160182) (“Mateev”), in view of Wang et al (NPL: SuperNeurons: Dynamic GPU Memory Management for Training Deep Neural Networks), and Fu et al (NPL: Automatic Generation of High-Performance Inference Kernels for Graph Neural Networks on Multi-Core Systems) (“Fu”).
Regarding Claim 6, Mateev and Wang teaches all the limitations of Claim 1
Mateev does not teach wherein the working set of the plurality of working sets which requires maximum memory usage is determined based on an architecture of the artificial neural network graph.
However, Fu does teach wherein the working set of the plurality of working sets which requires maximum memory usage is determined based on an architecture of the artificial neural network graph ([4.2 Memory Usage] teaches measuring maximum memory usage of a neural network graph).
Mateev, Wang, and Fu are analogous art because they all deal with neural networks and memory usage.
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Mateev with the liveness analysis of Wang and the memory usage tracking of Fu. Doing so would improve the GNN efficiency ([Fu-4.2])
Regarding Claim 7, Mateev and Wang teaches all the limitations of Claim 1
Mateev does not teach determining an intermediate representation of the artificial neural network graph ([Abstract] teaches an intermediate representation being an artificial neural network graph)
wherein the working set of the plurality of working sets which requires maximum memory usage is determined based on the intermediate representation ([4.2] teaches using an intermediate representation to measure maximum memory usage)
Claim 9 and 10 are rejected under USC 103 as being unpatentable over Mateev et al (US 20200160182) (“Mateev”), in view of Wang et al (NPL: SuperNeurons: Dynamic GPU Memory Management for Training Deep Neural Networks), and LI et al (US 20230084203 A1) (“Li”).
Regarding Claim 9, Mateev and Wang teaches all the limitations of Claim 1
Mateev does not teach wherein: the at least one identified layer is pruned based on an importance metric, and the importance metric is provided by user input.
However, Li does teach wherein: the at least one identified layer is pruned based on an importance metric, and the importance metric is provided by user input ([0081] teaches pruning based off of a given number of iterations, it can be reasonably interpreted that a given number of iterations is a user input. The iterations here is being interpreted as an importance metric. The pruning is being done on a neural network’s channel, those channels are linked to layers teaching the at least one identified layer part of the limitation teaching this limitation)
Mateev, Wang, And Li are analogous art because they all deal with neural networks and memory usage.
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Mateev with the liveness analysis of Wang and the iteration pruning of Li. Doing so would result in a more efficient neural network ([Li-0005])
Regarding Claim 10, Mateev Wang and Li teaches all the limitations of Claim 9
Li also teaches wherein: the importance metric is evaluated based on representative test data ([0074] teaches a models performance being evaluated based on the validation data. The validation data here is being interpreted as test data. The pruning is done by a given number of iterations teaching the importance metric [0081] teaching this limitation)
and the computer implemented method further comprises training the artificial neural network graph before evaluating the importance metrics ([0081] teaches a methodology with pruning with a neural network graph that includes a training process which leads to the pruned neural network graph being evaluated based off the validation data [0074] teaching this limitation)
Claim 12 is rejected under USC 103 as being unpatentable over Mateev et al (US 20200160182) (“Mateev”), in view of Wang et al (NPL: SuperNeurons: Dynamic GPU Memory Management for Training Deep Neural Networks) and Lou et al (US 12488249) (“Lou”)
Regarding Claim 12, Mateev and Wang teaches all the limitations of Claim 1
Mateev does not teach wherein at least one of the artificial neural network or the pre-determined threshold are provided by user input.
However, Lou does teach wherein at least one of the artificial neural network or the pre-determined threshold are provided by user input. ([Claim 10] teaches an accuracy threshold based on a user input).
Mateev, Wang and Lou are analogous art because they all deal with neural networks in relation to a threshold
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Mateev with the liveness analysis of Wang and the user input for threshold. Doing so would lead to faster and more accurate private neural network inference ([Lou-Description of Art])
Claim 13 is rejected under USC 103 as being unpatentable over Mateev et al (US 20200160182) (“Mateev”), in view of Wang et al (NPL: SuperNeurons: Dynamic GPU Memory Management for Training Deep Neural Networks) and view of Hao et al (CN111260049A) (“Hao”).
Regarding Claim 13, Mateev and Wang teaches all the limitations of Claim 1
Mateev does not teach wherein: the artificial neural network graph is to be deployed on a resource-constrained embedded system after pruning
However, Hao does teach wherein: the artificial neural network graph is to be deployed on a resource-constrained embedded system after pruning; And the embedded system is a mobile computing device, a mobile phone, a tablet computing device, an automotive compute platform, or an edge device. ([Detailed Description] Teaches a pruned neural network model being put on a domestic embedded system. The pruned neural network model is being interpreted as a neural network graph and the use of a domestic embedded system here would cover all the mentioned embedded system devices)
Mateev, Wang and Hao are analogous art because they all deal with neural networks in relation to a threshold
It would have been obvious to a person skilled in the art before the effective filling date of the claimed invention to combine Mateev with the liveness analysis of Wang and the neural network graph deployment of Hao. Doing so would improve neural network model performance on embedded systems ([Hao-Abstract])
Conclusion
The prior art made of record and not relied upon is considered to applicant’s disclosure
Chiu et al US 20200082268 A1 (2020-03-12) (Abstract “An electronic apparatus and a compression method for an artificial neural network are provided. The compression method is adapted for the artificial neural network with a plurality of convolution layers. The compression method includes: setting a first pruning layer for coupling the first pruning layer to Lth convolution layer, where the first pruning layer has a plurality of first weighting values and each of the first weighting values corresponds to each of a plurality of channels of the Lth convolution layer; tuning the first weighting values, selecting a part of the channels of the Lth convolution layer to be at least one first redundancy channel according to the first weighting values, and generating a compressed Lth convolution layer by deleting the at least one first redundancy channel; and removing the first pruning layer, and generating a first compressed artificial neural network.”).
Rhodes et al US 20220051103 A1 (2022-02-17) (Abstract “An apparatus is provided to compress CNN models using a combination of filter pruning and tensor decomposition. For example, the apparatus accesses a trained CNN that includes convolutional tensors. The apparatus prunes the filters of a convolutional tensor to generate a sparse tensor.”)
Watson et al US 20210357817 A1 (2021-11-18) (Claims 1 “ pruning the reduced input MLM such that a first input of a plurality of inputs of the original MLM is not included as an input of the reduced input MLM; determining that a size of the reduced input MLM exceeds a memory threshold of a target device; and based on the determination that the size of the reduced input MLM is greater than the memory threshold of the target device, prune the reduced input MLM to remove a second input of the plurality of inputs of the original MLM included in the reduced input MLM to generate a second reduced MLM;”).
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 URIAH V MOORE whose telephone number is (571)384-8341. The examiner can normally be reached Monday-Friday 8am-5pm.
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, Mariela Reyes can be reached at (571)270-1006. 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.
/URIAH VENDELL MOORE/Examiner, Art Unit 2142 /Mariela Reyes/Supervisory Patent Examiner, Art Unit 2142
