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 .
This action is in response to amendments filed March 11th, 2026. The status of the claims is as follows. Claims 2, 4, 13 and 15 are amended. Claims 1-20 are currently pending.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1, 5-7; 12, 16-18; 20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Gasse et al. (“Exact combinatorial optimization with graph convolutional neural networks” [2019], disclosed in IDS, hereinafter “Gasse”)
Regarding Claim 1,
Gasse discloses A computer-implemented method for training a neural network for solving combinatorial optimization problems, comprising: performing a first stage training of the neural network with a super-model for solving a first mixed integer linear program (MILP) instance using a source dataset, (Gasse [Abstract]; “In this work we propose to address the above challenges by using graph convolutional neural networks. More precisely, we focus on variable selection, also known as the branching problem, which lies at the core of the B&B paradigm yet is still not well theoretically understood [41], and adopt an imitation learning strategy to learn a fast approximation of strong branching, a high-quality but expensive branching rule. While such an idea is not new [30; 4; 24], we propose to address the learning problem in a novel way, through two contributions. First, we propose to encode the branching policies into a graph convolutional neural network (GCNN), which allows us to exploit the natural bipartite graph representation of MILP problems, thereby reducing the amount of manual feature engineering. Second, we approximate strong branching decisions by using behavioral cloning with a cross-entropy loss, a less difficult task than predicting strong branching scores [4] or rankings [30; 24]. We evaluate our approach on four classes of NP-hard problems, namely set covering, combinatorial auction, capacitated facility location and maximum independent set”
Gasse [Section 4.3];
PNG
media_image1.png
54
537
media_image1.png
Greyscale
PNG
media_image2.png
178
541
media_image2.png
Greyscale
Wherein the first pass constituting constraint embeddings updated based on the variable embeddings, constraint embeddings, and edge embeddings read on a first phase)
the neural network receiving a bipartite graph representation of an MILP sample of the MILP instance as input, the bipartite graph consisting of a group of variable nodes, a group of constraint nodes, and edges between nodes in the group of variable nodes and the group of constraint nodes, (Gasse [Abstract]; “In this work we propose to address the above challenges by using graph convolutional neural networks. More precisely, we focus on variable selection, also known as the branching problem, which lies at the core of the B&B paradigm yet is still not well theoretically understood [41], and adopt an imitation learning strategy to learn a fast approximation of strong branching, a high-quality but expensive branching rule. While such an idea is not new [30; 4; 24], we propose to address the learning problem in a novel way, through two contributions. First, we propose to encode the branching policies into a graph convolutional neural network (GCNN), which allows us to exploit the natural bipartite graph representation of MILP problems, thereby reducing the amount of manual feature engineering. Second, we approximate strong branching decisions by using behavioral cloning with a cross-entropy loss, a less difficult task than predicting strong branching scores [4] or rankings [30; 24]. We evaluate our approach on four classes of NP-hard problems, namely set covering, combinatorial auction, capacitated facility location and maximum independent set”)
the neural network comprising one or more normalization layers (Gasse [Section 4.3];
PNG
media_image3.png
45
535
media_image3.png
Greyscale
PNG
media_image4.png
174
538
media_image4.png
Greyscale
Wherein the graph convolution layers comprising a masked softmax activation reads on one or more normalization layer for learning variables)
the source dataset comprising MILP samples for a second MILP instance different from the first MILP instance; (Gasse [Section 4.1];
PNG
media_image5.png
60
547
media_image5.png
Greyscale
Wherein the collection of training instances of interest reads on a training dataset comprising a plurality of instances and their associated MILP samples)
and performing a second stage training of the neural network, the second stage training comprising adapting the super-model with a target dataset different from the source dataset, the target dataset comprising MILP samples for the first MILP instance (Gasse [Section 4.3];
PNG
media_image1.png
54
537
media_image1.png
Greyscale
PNG
media_image2.png
178
541
media_image2.png
Greyscale
Wherein the second pass constituting variable embeddings updated based on the variable embeddings, constraint embeddings, and edge embeddings for all of j iterations read on a second phase done with a target dataset different from the source dataset (I iterations))
wherein only the normalization layers of the neural network are updated during the adapting (Gasse [Section 4.3];
PNG
media_image3.png
45
535
media_image3.png
Greyscale
PNG
media_image4.png
174
538
media_image4.png
Greyscale
Wherein the graph convolution layers comprising a masked softmax activation reads on only normalization layers being updated during the adaptation passes)
Regarding Claim 5,
Gasse teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Gasse further discloses wherein the neural network comprises a graph convolutional neural network (GCN) (Gasse [Section 4.3];
PNG
media_image6.png
221
546
media_image6.png
Greyscale
PNG
media_image6.png
221
546
media_image6.png
Greyscale
)
Regarding Claim 6,
Gasse teaches the method of Claim 5 (and thus the rejection of Claim 5 is incorporated). Gasse further discloses wherein the GCN performs a single graph convolution in the form of two interleaved half-convolutions, the graph convolution being performed by two successive convolution passes, one half-convolution from variable to constraints performed by a first convolution layer and the other half-convolution from constraints to variables performed by a second convolution layer, wherein the two successive convolution passes are accordance with equation (2):
PNG
media_image7.png
77
470
media_image7.png
Greyscale
for all i∈C,j∈V{1,…,n}, where fC, fV, gC and gV are 2-layer perceptrons with rectified linear activation unit (RELU) activation functions, wherein an affine transformation x←(x-β)/σ is applied immediately after each of the first and second convolution layers by respective first and second normalization layers which comprise the one or more normalization layers (Gasse [Section 4.3];
PNG
media_image3.png
45
535
media_image3.png
Greyscale
PNG
media_image4.png
174
538
media_image4.png
Greyscale
)
Regarding Claim 7,
Gasse teaches the method of Claim 6 (and thus the rejection of Claim 6 is incorporated). Gasse further discloses wherein the first and second convolution layers are unnormalized convolution layers (Gasse [Section 4.3];
PNG
media_image8.png
173
540
media_image8.png
Greyscale
)
Claims 12, 16-18 recite a computing device comprising one or more processors configured to execute the exact method of Claims 1, 5-7 respectively. Thus, Claims 12, 16-18 are rejected for reasons set forth in the rejection of Claims 1, 5-7 respectively.
Claim 20 recites a non-transitory machine-readable medium having stored executable instructions to cause one or more processors to execute the exact method of Claim 1. Thus, Claim 20 is rejected for reasons set forth in the rejection of Claim 1 respectively.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 2-4; 13-15 are rejected under 35 U.S.C. 103 as being unpatentable over Gasse et al. (“Exact combinatorial optimization with graph convolutional neural networks” [2019], disclosed in IDS, hereinafter “Gasse”) in view of Nair et al. (“Solving mixed integer programs using neural networks” [2021], disclosed in IDS, hereinafter “Nair”).
Regarding Claim 2,
Gasse teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Gasse already discloses wherein the first stage training comprises: (1-i) receiving a MILP sample of the first MILP instance; (1-ii) generating a representation vector based on the MILP sample of the first MILP instance; (1-iii) selecting one or more variables for the MILP sample from the representation vector (Gasse [Section 3.1];
PNG
media_image9.png
241
545
media_image9.png
Greyscale
Wherein the received MILP is broken down into a generated representation matrix based on the MILP sample of the instance, thus reading on receiving an MILP sample and generating a representation vector based on received sample
Gasse [Section 4.2]; “ We encode the state st of the B&B process at time t as a bipartite graph with node and edge features (G, C, E, V), described in Figure 2 (Left). On one side of the graph are nodes corresponding to the constraints in the MILP, one per row in the current node’s LP relaxation, with C ∈ R m×c their feature matrix. On the other side are nodes corresponding to the variables in the MILP, one per LP column, with V ∈ R n×d their feature matrix. An edge (i, j) ∈ E connects a constraint node i and a variable node j if the latter is involved in the former, that is if Aij 6= 0, and E ∈ R m×n×e represents the (sparse) tensor of edge features. Note that under mere restrictions in the B&B solver (namely, by enabling cuts only at the root node), the graph structure is the same for all LPs in the B&B tree, which reduces the cost of feature extraction. The exact features attached to the graph are described in the supplementary materials. We note that this is really only a subset of the solver state, which technically turns the process into a partially-observable Markov decision process [6], but also that excellent variable selection policies such as strong branching are able to do well despite relying only on a subset of the solver state as well.” wherein the variable selection policies reliant on nodes derived from the feature matrix reads on selecting variables for the MILP sample from the representation vector)
Gasse does not explicitly disclose but Nair discloses (1-iv) determining a classification from the one or more selected variables; (1-v) determining a loss based on the determined classification and a predetermined classification in the source dataset; (1-vi) updating one or more parameters of the neural network based on the determined loss; and (1-vii) repeating steps (1-i) to (1-vi) until the loss is below a threshold (Nair [Section 6.2];
PNG
media_image10.png
582
657
media_image10.png
Greyscale
Wherein the classification of selected variables under labels {0,1} and comparison of such classifications xd against predetermined classifications yd in the loss function reads on determining a classification and related loss; wherein the loss function training the model parameters such as θ reads on updating parameters of the neural network based on the determined loss; wherein the repetition of training until loss is determined to be minimized reads on repeating steps (1-i) to (1-vi) until loss is below some minimal threshold).
It would have been obvious to modify Gasse’s method of training a neural network through selected variables to use Nair’s method of classifying the selected variables and determining a loss function for comparison against a threshold to determine the continuance of model training. One would have been motivated to do so in order to “warm-start SCIP to find high quality solutions in much shorter time” (Nair [Section 6.2 Paragraph 3]).
The combination of Gasse/Nair thus inherently discloses wherein the adapting comprises: (2-i) receiving an MILP sample of the second MILP instance; (2-ii) generating a representation vector based of the MILP sample of the second MILP instance; (2-iii) selecting one or more variables for the MILP sample from the representation vector; (2-iv) determining a classification from the one or more selected variables; (2-v) determining a loss based on the determined classification and a predetermined classification in the target dataset; (2-vi) updating one or more normalization parameters of the one or more normalization layers of the neural network based on the determined loss; and (2-vii) repeating steps (i) to (vi) until the loss is below a threshold (Gasse [Section 4.3];
PNG
media_image3.png
45
535
media_image3.png
Greyscale
PNG
media_image4.png
174
538
media_image4.png
Greyscale
Gasse [Section 3.1];
PNG
media_image9.png
241
545
media_image9.png
Greyscale
Nair [Section 6.2];
PNG
media_image10.png
582
657
media_image10.png
Greyscale
Wherein the disclosed steps (1-i) to (1-vii) performed for the second successive pass reads on steps (2-i) to (2-vii) for the adaptation second phase variable pass)
Regarding Claim 3,
The combination of Gasse/Nair teaches the method of Claim 2 (and thus the rejection of Claim 2 is incorporated). The combination already discloses wherein the loss is determined in accordance with equation (1):
PNG
media_image11.png
47
148
media_image11.png
Greyscale
where the L is the loss, y is a ground truth classification of the respective source or target dataset, and y^ is the determined classification of the neural network. (Nair [Section 6.2];
PNG
media_image10.png
582
657
media_image10.png
Greyscale
Wherein the determined loss comprises the negative summation of the product result of the ground truth classification yd and the log value of the determined classification xd)
Regarding Claim 4,
Gasse teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Gasse already discloses wherein the first stage training comprises: (1-i) receiving an MILP sample of the first MILP instance; (1-ii) generating a representation vector based on the MILP sample of the first MILP instance; (1-iii) selecting one or more variables for the MILP sample from the representation vector; (Gasse [Section 3.1];
PNG
media_image9.png
241
545
media_image9.png
Greyscale
Wherein the received MILP is broken down into a generated representation matrix based on the MILP sample of the instance, thus reading on receiving an MILP sample and generating a representation vector based on received sample
Gasse [Section 4.2]; “ We encode the state st of the B&B process at time t as a bipartite graph with node and edge features (G, C, E, V), described in Figure 2 (Left). On one side of the graph are nodes corresponding to the constraints in the MILP, one per row in the current node’s LP relaxation, with C ∈ R m×c their feature matrix. On the other side are nodes corresponding to the variables in the MILP, one per LP column, with V ∈ R n×d their feature matrix. An edge (i, j) ∈ E connects a constraint node i and a variable node j if the latter is involved in the former, that is if Aij 6= 0, and E ∈ R m×n×e represents the (sparse) tensor of edge features. Note that under mere restrictions in the B&B solver (namely, by enabling cuts only at the root node), the graph structure is the same for all LPs in the B&B tree, which reduces the cost of feature extraction. The exact features attached to the graph are described in the supplementary materials. We note that this is really only a subset of the solver state, which technically turns the process into a partially-observable Markov decision process [6], but also that excellent variable selection policies such as strong branching are able to do well despite relying only on a subset of the solver state as well.” wherein the variable selection policies reliant on nodes derived from the feature matrix reads on selecting variables for the MILP sample from the representation vector)
Gasse does not explicitly disclose but Nair discloses (1-iv) determining a loss based on the one or more selected variables and one or more predetermined selected variables in the source dataset; (1-v) updating one or more parameters of the neural network based on the determined loss; and (1-vi) repeating steps (1-i) to (1-v) until the loss is below a threshold (Nair [Section 6.2];
PNG
media_image10.png
582
657
media_image10.png
Greyscale
Wherein the classification of selected variables under labels {0,1} and comparison of such classifications xd against predetermined classifications yd in the loss function reads on determining a loss based, at least in part, on the selected variables and predetermined selected variables’ values; wherein the loss function training the model parameters such as θ reads on updating parameters of the neural network based on the determined loss; wherein the repetition of training until loss is determined to be minimized reads on repeating steps (1-i) to (1-vi) until loss is below some minimal threshold).
It would have been obvious to modify Gasse’s method of training a neural network through selected variables to use Nair’s method of determining a loss function for comparison against a threshold to determine the continuance of model training. One would have been motivated to do so in order to “warm-start SCIP to find high quality solutions in much shorter time” (Nair [Section 6.2 Paragraph 3]).
The combination of Gasse/Nair thus inherently discloses wherein the adapting comprises: (2-i) receiving an MILP sample of the second MILP instance; (2-ii) generating a representation vector based of the MILP sample of the second MILP instance; (2-iii) selecting one or more variables for the MILP sample from the representation vector; (2-iv) determining a loss based on the one or more selected variables and one or more predetermined selected variables in the target dataset; (2-v) updating one or more parameters of the neural network based on the determined loss; and (2-vi) repeating steps (i) to (v) until the loss is below a threshold (Gasse [Section 4.3];
PNG
media_image3.png
45
535
media_image3.png
Greyscale
PNG
media_image4.png
174
538
media_image4.png
Greyscale
Gasse [Section 3.1];
PNG
media_image9.png
241
545
media_image9.png
Greyscale
Nair [Section 6.2];
PNG
media_image10.png
582
657
media_image10.png
Greyscale
Wherein the disclosed steps (1-i) to (1-v) performed for the second successive pass reads on steps (2-i) to (2-v) for the adaptation second phase variable pass)
Claims 13-15 recite a computing device comprising one or more processors configured to execute the exact method of Claims 2-4 respectively. Thus, Claims 13-15 are rejected for reasons set forth in the rejection of Claims 2-4 respectively.
Claims 8-10 are rejected under 35 U.S.C. 103 as being unpatentable over Gasse et al. (“Exact combinatorial optimization with graph convolutional neural networks” [2019], disclosed in IDS, hereinafter “Gasse”) in view of Robert et al. (GB2586868A, hereinafter “Robert”).
Regarding Claim 8,
Gasse teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Gasse does not explicitly disclose but Robert discloses wherein the one or more selected variables or classification determined based on the one or more selected variables are applied to a system for associated with cellular networks, telecommunication networks, scheduling, renewable energy, aviation dispatching, or artificial intelligence and/or cloud retarget allocation (Robert [Page 1 Line 18]; “Many of the control and optimisation problems encountered in wireless networks can be viewed as combinatorial optimisation problems in which various parameters need to be adjusted to maximise some combination of key performance indicators (KPI). The parameters to be adjusted may include cell transmitted power levels, antenna tilt angles, handover thresholds, admission control thresholds, beamforming configuration and scheduler parameters. Often these optimisation problems are NP-hard and prohibitively expensive to solve, and in practice it is common to use relatively simple heuristics to search for good sub-optimal solutions.”
Robert [Page 4 Line 27]; “The network optimisation may comprise coverage and capacity optimisation (e.g. transmission power optimisation / antenna tilt optimisation). The at least one metric may be estimated using an environment model for said network environment. The at least one respective metric, for a given UE, may comprise at least one of: a cell association for that UE; a signal-to-interference-plus-noise ratio (SINR) for that UE; and a throughput for that UE” wherein selected variables are applied in regards to a signal-to-interference-plus-noise ratio metric, thus reading on selected variables determined so as to connect mobile equipment in a way that minimizes interference)
It would have been obvious to use Gasse’s method of variable selection for Robert’s application in telecommunications. One would have been motivated to do so in order “to adjust network parameters so as to maximise a metric related to the throughput experienced by the users .. [and thus] focus on transmission power optimisation” (Robert [Page 2 Line 20])
Regarding Claim 9,
Gasse teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Gasse does not explicitly disclose but Robert discloses wherein the one or more selected variables or classification determined based on the one or more selected variables are applied to distribute available frequencies across antennas in a cellular network so as to connect mobile equipment and that interference between the antennas in the cellular network is minimized (Robert [Page 1 Line 18]; “Many of the control and optimisation problems encountered in wireless networks can be viewed as combinatorial optimisation problems in which various parameters need to be adjusted to maximise some combination of key performance indicators (KPI). The parameters to be adjusted may include cell transmitted power levels, antenna tilt angles, handover thresholds, admission control thresholds, beamforming configuration and scheduler parameters. Often these optimisation problems are NP-hard and prohibitively expensive to solve, and in practice it is common to use relatively simple heuristics to search for good sub-optimal solutions.”
Robert [Page 4 Line 27]; “The network optimisation may comprise coverage and capacity optimisation (e.g. transmission power optimisation / antenna tilt optimisation). The at least one metric may be estimated using an environment model for said network environment. The at least one respective metric, for a given UE, may comprise at least one of: a cell association for that UE; a signal-to-interference-plus-noise ratio (SINR) for that UE; and a throughput for that UE” wherein selected variables are applied in regards to a signal-to-interference-plus-noise ratio metric, thus reading on selected variables determined so as to connect mobile equipment in a way that minimizes interference)
It would have been obvious to use Gasse’s method of variable selection for application in Robert’s telecommunication interference minimization. One would have been motivated to do so in order “to adjust network parameters so as to maximise a metric related to the throughput experienced by the users .. [and thus] focus on transmission power optimisation” (Robert [Page 2 Line 20])
Regarding Claim 10,
Gasse teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Gasse does not explicitly disclose but Robert discloses wherein the one or more selected variables or label or classification determined based on the one or more selected variables are applied to a determine network lines of a telecommunication network so that so that a predefined set of communication requirements are met and a total cost of the telecommunication network is minimized (Robert [Page 1 Line 18]; “Many of the control and optimisation problems encountered in wireless networks can be viewed as combinatorial optimisation problems in which various parameters need to be adjusted to maximise some combination of key performance indicators (KPI). The parameters to be adjusted may include cell transmitted power levels, antenna tilt angles, handover thresholds, admission control thresholds, beamforming configuration and scheduler parameters. Often these optimisation problems are NP-hard and prohibitively expensive to solve, and in practice it is common to use relatively simple heuristics to search for good sub-optimal solutions.”
Robert [Page 4 Line 27]; “The network optimisation may comprise coverage and capacity optimisation (e.g. transmission power optimisation / antenna tilt optimisation). The at least one metric may be estimated using an environment model for said network environment. The at least one respective metric, for a given UE, may comprise at least one of: a cell association for that UE; a signal-to-interference-plus-noise ratio (SINR) for that UE; and a throughput for that UE” wherein selected variables are applied in regards to a signal-to-interference-plus-noise ratio metric, thus reading on selected variables determined so as to meet a predefined set of communication requirement metrics for minimizing the total cost of the communication network (by maximizing performance of metrics including but not limited to interference reduction))
It would have been obvious to use Gasse’s method of variable selection for application in Robert’s telecommunication network cost minimization. One would have been motivated to do so in order “to adjust network parameters so as to maximise a metric related to the throughput experienced by the users .. [and thus] focus on transmission power optimisation” (Robert [Page 2 Line 20]).
Claim 11 is rejected under 35 U.S.C. 103 as being unpatentable over Gasse et al. (“Exact combinatorial optimization with graph convolutional neural networks” [2019], disclosed in IDS, hereinafter “Gasse”) in view of Hubbs et al. (CA3116855A1, hereinafter “Hubbs”).
Regarding Claim 11,
Gasse teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Gasse does not explicitly disclose but Hubbs discloses wherein the one or more selected variables or label or classification determined based on the one or more selected variables are applied to cost efficient deep learning job allocation (CE-DLA), wherein energy consumption of deep learning clusters is minimized while maintaining an overall system performance within an acceptable threshold (Hubbs [Abstract]; “ Methods and apparatus for scheduling production at a production facility are provided. A model of a production facility utilizing one or more input materials to produce products that satisfy product requests can be determined. Each product request can specify a requested product to be available at a requested time. Policy and value neural networks can be determined for the production facility. The policy neural network can represent production actions to be scheduled at the production facility and the value neural network can represent benefits of products produced at the production facility. The policy and value neural networks can use the model of the production facility during training for generating a schedule of the production actions at the production facility that satisfy the product requests over an interval of time and relates to penalties due to late production of the requested products” wherein scheduling of production tasks through neural network policies for satisfaction of production constraints read on selected variables used for cost efficient deep learning job allocation
Hubbs [0172]; “The MILP model can generate a schedule for 2H time periods to provide better end-state conditions, where H is the number of days in the unchangeable planning horizon; in this example, H = 7. Then, the schedule is passed to a model of the production facility to execute. The model of the production facility is stepped forward one time step and the results are fed back into the MILP model to generate a new schedule over the 2H planning horizon“
Hubbs [0063]; “Additionally, the configuration of cluster routers 206 can be based on the data communication requirements of server devices 202 and data storage 204, the latency and throughput of the local cluster network 208, the latency, throughput, and cost of communication link 210, and/or other factors that can contribute to the cost, speed, fault- tolerance, resiliency, efficiency and/or other design goals of the system architecture.”
Hubbs [0039]; “ During scheduling and planning, these questions can be asked and answered with respect to minimize cost, maximize profit, minimize malcespan (i.e., a time difference between starting and finishing
product production), and/or one or more other metrics” wherein the configuration of cluster routes are based on metrics built to minimize cost and malcespan)
It would have been obvious to use Gasse’s method of variable selection for application in Hubbs’ deep learning job allocation. One would have been motivated to do so for the purpose of “generating a schedule of the production actions at the production facility that satisfy the product requests over an interval of time and relates to penalties” (Hubbs [Abstract]) thus satisfying constraints of the optimization problem while minimizing penalties
Claim 19 is rejected under 35 U.S.C. 103 as being unpatentable over Gasse et al. (“Exact combinatorial optimization with graph convolutional neural networks” [2019], disclosed in IDS, hereinafter “Gasse”) in view of Robert et al. (GB2586868A, hereinafter “Robert”) and further in view of Hubbs et al. (CA3116855A1, hereinafter “Hubbs”).
Gasse teaches the computing device of Claim 12 (and thus the rejection of Claim 12 is incorporated). Gasse does not disclose but Robert discloses wherein the one or more selected variables or classification determined based on the one or more selected variables are applied to a system for associated with cellular networks, telecommunication networks, scheduling, renewable energy, aviation dispatching, or artificial intelligence and/or cloud retarget allocation (Robert [Page 1 Line 18]; “Many of the control and optimisation problems encountered in wireless networks can be viewed as combinatorial optimisation problems in which various parameters need to be adjusted to maximise some combination of key performance indicators (KPI). The parameters to be adjusted may include cell transmitted power levels, antenna tilt angles, handover thresholds, admission control thresholds, beamforming configuration and scheduler parameters. Often these optimisation problems are NP-hard and prohibitively expensive to solve, and in practice it is common to use relatively simple heuristics to search for good sub-optimal solutions.”
Robert [Page 4 Line 27]; “The network optimisation may comprise coverage and capacity optimisation (e.g. transmission power optimisation / antenna tilt optimisation). The at least one metric may be estimated using an environment model for said network environment. The at least one respective metric, for a given UE, may comprise at least one of: a cell association for that UE; a signal-to-interference-plus-noise ratio (SINR) for that UE; and a throughput for that UE” wherein selected variables are applied in regards to a signal-to-interference-plus-noise ratio metric, thus reading on selected variables determined so as to connect mobile equipment in a way that minimizes interference)
wherein the one or more selected variables or classification determined based on the one or more selected variables are applied to distribute available frequencies across antennas in a cellular network so as to connect mobile equipment and that interference between the antennas in the cellular network is minimized (Robert [Page 1 Line 18]; “Many of the control and optimisation problems encountered in wireless networks can be viewed as combinatorial optimisation problems in which various parameters need to be adjusted to maximise some combination of key performance indicators (KPI). The parameters to be adjusted may include cell transmitted power levels, antenna tilt angles, handover thresholds, admission control thresholds, beamforming configuration and scheduler parameters. Often these optimisation problems are NP-hard and prohibitively expensive to solve, and in practice it is common to use relatively simple heuristics to search for good sub-optimal solutions.”
Robert [Page 4 Line 27]; “The network optimisation may comprise coverage and capacity optimisation (e.g. transmission power optimisation / antenna tilt optimisation). The at least one metric may be estimated using an environment model for said network environment. The at least one respective metric, for a given UE, may comprise at least one of: a cell association for that UE; a signal-to-interference-plus-noise ratio (SINR) for that UE; and a throughput for that UE” wherein selected variables are applied in regards to a signal-to-interference-plus-noise ratio metric, thus reading on selected variables determined so as to connect mobile equipment in a way that minimizes interference)
wherein the one or more selected variables or label or classification determined based on the one or more selected variables are applied to a determine network lines of a telecommunication network so that so that a predefined set of communication requirements are met and a total cost of the telecommunication network is minimized (Robert [Page 1 Line 18]; “Many of the control and optimisation problems encountered in wireless networks can be viewed as combinatorial optimisation problems in which various parameters need to be adjusted to maximise some combination of key performance indicators (KPI). The parameters to be adjusted may include cell transmitted power levels, antenna tilt angles, handover thresholds, admission control thresholds, beamforming configuration and scheduler parameters. Often these optimisation problems are NP-hard and prohibitively expensive to solve, and in practice it is common to use relatively simple heuristics to search for good sub-optimal solutions.”
Robert [Page 4 Line 27]; “The network optimisation may comprise coverage and capacity optimisation (e.g. transmission power optimisation / antenna tilt optimisation). The at least one metric may be estimated using an environment model for said network environment. The at least one respective metric, for a given UE, may comprise at least one of: a cell association for that UE; a signal-to-interference-plus-noise ratio (SINR) for that UE; and a throughput for that UE” wherein selected variables are applied in regards to a signal-to-interference-plus-noise ratio metric, thus reading on selected variables determined so as to meet a predefined set of communication requirement metrics for minimizing the total cost of the communication network (by maximizing performance of metrics including but not limited to interference reduction))
It would have been obvious to use Gasse’s method of variable selection for application in Robert’s telecommunication network cost minimization. One would have been motivated to do so in order “to adjust network parameters so as to maximise a metric related to the throughput experienced by the users … [and thus] focus on transmission power optimisation” (Robert [Page 2 Line 20]).
Gasse/Robert does not explicitly disclose but Hubbs discloses wherein the one or more selected variables or label or classification determined based on the one or more selected variables are applied to cost efficient deep learning job allocation (CE-DLA), wherein energy consumption of deep learning clusters is minimized while maintaining an overall system performance within an acceptable threshold (Hubbs [Abstract]; “ Methods and apparatus for scheduling production at a production facility are provided. A model of a production facility utilizing one or more input materials to produce products that satisfy product requests can be determined. Each product request can specify a requested product to be available at a requested time. Policy and value neural networks can be determined for the production facility. The policy neural network can represent production actions to be scheduled at the production facility and the value neural network can represent benefits of products produced at the production facility. The policy and value neural networks can use the model of the production facility during training for generating a schedule of the production actions at the production facility that satisfy the product requests over an interval of time and relates to penalties due to late production of the requested products” wherein scheduling of production tasks through neural network policies for satisfaction of production constraints read on selected variables used for cost efficient deep learning job allocation
Hubbs [0172]; “The MILP model can generate a schedule for 2H time periods to provide better end-state conditions, where H is the number of days in the unchangeable planning horizon; in this example, H = 7. Then, the schedule is passed to a model of the production facility to execute. The model of the production facility is stepped forward one time step and the results are fed back into the MILP model to generate a new schedule over the 2H planning horizon“
Hubbs [0063]; “Additionally, the configuration of cluster routers 206 can be based on the data communication requirements of server devices 202 and data storage 204, the latency and throughput of the local cluster network 208, the latency, throughput, and cost of communication link 210, and/or other factors that can contribute to the cost, speed, fault- tolerance, resiliency, efficiency and/or other design goals of the system architecture.”
Hubbs [0039]; “ During scheduling and planning, these questions can be asked and answered with respect to minimize cost, maximize profit, minimize malcespan (i.e., a time difference between starting and finishing product production), and/or one or more other metrics” wherein the configuration of cluster routes are based on metrics built to minimize cost and malcespan)
It would have been obvious to use Gasse/Robert’s method of variable selection in telecommunication network cost minimization for application in Hubbs’ deep learning job allocation. One would have been motivated to do so for the purpose of “generating a schedule of the production actions at the production facility that satisfy the product requests over an interval of time and relates to penalties” (Hubbs [Abstract]) thus satisfying constraints of the optimization problem while minimizing penalties.
Response to Arguments
The Examiner acknowledges the Applicant’s amendments to Claims 2, 4, 13 and 15.
Applicant’s arguments filed March 11th, 2026, traversing the objection to claims 4 and 15 have been fully considered and are fully persuasive.
Applicant’s arguments filed March 11th, 2026, traversing the rejection of claims 1, 5-7, 12, 16-18, and 20 under 35 U.S.C. § 102(a)(1) and claims 2-4, 8-11, 13-15, and 19 under 35 U.S.C. § 103 have been fully considered, but are not fully persuasive.
Applicant recites, on Pages 11-12 of the Remarks, that Gasse fails to disclose “the source dataset comprising MILP samples for a sample MILP instance different from the first MILP instance … the target dataset comprising MILP samples for the first MILP instance”. Instead, applicant argues that examiner cannot equate “collection of training instances of interest” to disclose a first MILP instance belonging to a target dataset as well as a second MILP instance belonging to a source dataset.
Examiner respectfully disagrees. Gasse’s disclosed “collection of training instances of interest” can be attributed to both a first target dataset as well as a source dataset. Notably, Gasse obtains its instances for training as well as validation by performing a training-validation split on its collection of instances. Specifically, it is elaborated further in Gasse [Section I] that “For each benchmark problem, namely, set covering, combinatorial auction, capacitated facility location and maximum independent set, we generate 10,000 random instances for training, 2,000 for validation, and 3x20 for testing (20 easy instances, 20 medium instances, and 20 hard instances). In order to obtain our datasets of state-action pairs {(si,ai)} for training and validation, we pick an instance from the corresponding set (training or validation), solve it with SCIP, each time with a new random seed, and record new node states and strong branching decision during the branch-and-bound process. We continue processing new instances by sampling with replacement, until the desired number of node samples is reached, that is, 100,000 samples for training, and 20,000 for validation”. As such, the disclosed 10,000 initial random instances for training and 2,000 random instances for validation to generated associated node sample being “datasets of state-action pairs {{si,ai)} for training and validation thus reads on the collection comprising a disclosed source dataset and target dataset, both comprising MILP samples for their respective sample MILP instances.
Applicant recites, on Pages 12-13 of the Remarks, that Gasse fails to disclose domain adaptation by only updating normalization layers based on a separate target dataset. Specifically, that Gasse teaches away from only updating normalization layers based on a separate target dataset due to its disclosed prenorm layers.
Examiner respectfully disagrees. Examiner interprets “normalization layers” as any layer whose principal function is normalization. A masked softmax is disclosed by Gasse to normalize activations into a probability distribution, thus usage of the masked softmax as an activation function for a layer reads on such layers attaining the functionality and interpretability of “normalization layers”. Gasse’s masked softmax is performed across a plurality of variable nodes outputted by a final 2-layer perceptron. As such, the set of values participating in the normalization changes, reading on a changing normalization mechanism. Since such a set of values representative of the “mask” in the masked softmax-based layers controls the normalization being performed, updating the mask is equivalent to updating the normalization layer’s behavior. Such updates are interpretable as “only updating normalization layers” since the step of updating only the mask associated with normalization thus reads on at least some portion of the model where only the normalization layers of the neural network are updated. Notably, the currently recited claim language is broadly interpreted to say that at some point during the adapting process (“during the adapting”), there must be some instance of normalization layers being selectively updated (“only the normalization layers of the neural network are updated”). As such, Gasse’s disclosure of masked-softmax activation based layers of the neural network (normalization layers) being updated by changing the set of masking participating values (only normalization layers updated) is interpretable as only the normalization layers of the neural network are updated during the adapting.
The rejection of Claim 1 under 35 U.S.C. § 102(a)(1) has been maintained. Similarly, the rejection of Claims 12 and 20 under 35 U.S.C. § 102(a)(1) have been maintained.
The rejection of Claims 5-7 under 35 U.S.C. § 102(a)(1), which depend directly or indirectly from Claim 1, have been maintained.
The rejection of Claims 2-4, 8-11 under 35 U.S.C. § 103, which depend directly or indirectly from Claim 1, have been maintained.
The rejection of Claims 16-18 under 35 U.S.C. § 102(a)(1), which depend directly or indirectly from Claim 12, have been maintained.
The rejection of Claims 13-15, 19 under 35 U.S.C. § 103, which depend directly or indirectly from Claim 1, have been maintained.
Conclusion
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 JONATHAN J KIM whose telephone number is (571)272-0523. The examiner can normally be reached 8-6.
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, Matt Ell can be reached on (571) 270-3264. 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.
/JONATHAN J KIM/Examiner, Art Unit 2141
/MATTHEW ELL/Supervisory Patent Examiner, Art Unit 2141
Prosecution Insights