DETAILED ACTION
1. This office action is in response to the Application No. 19197957 filed on 07/29/2026. Claims 1-20 are presented for examination and are currently pending. Applicant’s arguments have been carefully and respectfully considered.
Response to Arguments
2. The double patenting rejection has been withdrawn in light of the terminal disclaimer filed and approved 07/29/2026.
The Applicant’s argument on pg. 9 that “the amendment requiring that the hierarchical supervisory system coordinate pruning decisions "in real time during operation of the deep learning network" recites a specific technical mechanism by which the architecture of the deep learning network is dynamically modified concurrently with, rather than offline from, execution of the network” and “the hierarchical supervisory system coordinates the pruning decisions based on the utility metrics" generated by the greedy neural system. This creates a specific, claimed technical interrelationship in which architectural modification of the deep learning network is dynamically driven by the output of the competitive bidding process, in real time and during operation of the network. This closed-loop coordination between the pruning function of the hierarchical supervisory system and the utility-based competitive bidding process of the greedy neural system reflects a specific improvement to the technology of adaptively restructuring a deep learning network during runtime” are persuasive because it leads to an improvement in the technological field of a deep learning network. As a result, the 101 is withdrawn.
The Applicants argument regarding the prior art have been considered and the Examiner is withdrawing the rejections in the previous Office action because Applicant’s amendment necessitated new grounds of rejection presented in this Office Action. It is noted that arguments regarding independent claims 1 and 11 have been considered but are moot because new references have now been used to remap the independent claims.
Furthermore, Yuan, Chen, Mesadieu and Abraham which were applied in the previous Office Action are still relevant to the instant dependent claims. As a result, their teachings have been used in this Office Action.
The dependent claims 2-10 and 12-20 which depend directly or indirectly from independent claims 1 and 11 are not patentable because the instant claims are still obvious over the prior art of record.
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.
3. Claims 1 and 11 are rejected under 35 U.S.C. 103 as being unpatentable over Auda et al. (MODULAR NEURAL NETWORKS: A SURVEY, International Journal of Neural Systems, Vol. 9, No. 2 (April, 1999) 129-151) in view of Sikka et al. (US20210012210)
Regarding claim 1, Auda teaches a computer system comprising a hardware memory (For example, implementing an MNN on NCUBE,56 BSP400 neurocomputer...In case of physical implementation of an NN, all nodes and connections are physically implemented in hardware. pg. 133, right column, second to the last para.) wherein the computer system is configured to execute software instructions (Having less connections per node results in decrease of the communication overhead on the buses and efficient use of utilizing the processors. pg. 133, right column, second to the last para.) that:
operate a neural network (A Modular Neural Network (MNN) is a Neural Network (NN) that consists of several modules, each module carrying out one sub-task of the NN's global task, and all modules functionally integrated. A module can be a sub-structure or a learning sub-procedure of the whole network. Pg. 129, left column, first para.))
comprising interconnected nodes arranged in layers (Figure 4 and Table 1 show an experiment we have carried out...by a three-layer Backpropagation network pg. 134, right column, second full para.));
implement a hierarchical supervisory system (an unsupervised
learning stage is introduced before the supervised learning stage in order to represent data in a more advantageous way, pg. 138, right column, third para.) monitoring the neural network through multiple supervisory levels (using a hierarchy of unsupervised NNs for task decomposition. Pg. 144, right column, last full para.); Feature extraction techniques include using an unsupervised delta rule self-organization hidden layer before the supervised layers, pg. 138, right column, third para.)),
wherein the hierarchical supervisory system collects activation data (uses “fuzzy integrals" to combine several NN activations, pg. 144, left column, second para.)),
identifies operation patterns (new training patterns may affect the learning of nearly all existing patterns, pg. 133, left column, second para.),
implements architectural changes (new modular architectures where adding more modules allows, hopefully, unlimited extendibility. Pg. 132, right column, first para.)),
detects network sparsity (decomposing the objective task over smaller, sparsely-connected, and less complex modules decreases the connections-per-node ratio substantially, pg. 136 right column, second para.)),
coordinates pruning decisions in real time during operation of the neural network (pruning during learning, pg. 136, right column, third para.), and
manages resource redistribution (Distributing the representation of multidimensional spaces among multiple networks (modules) can reduce the length and the complexity of the required connections. Pg. 134 left column, first para.));
implement a meta-supervisory system that tracks supervisory behavior patterns (studying the behavior of the hidden units when the inputs and outputs are trained to patterns similar to some input-output information given by the nervous system; studying...using the famous multiple experts MNN, pg. 131 left column, second to the last para.)), and
extracts generalizable principles (Reducing the dimensionality of the input data via unsupervised feature extraction decreases the computational expenses by decreasing the number of weights of the supervised learning NN. Moreover, this will enhance the generalization abilities of the network by decreasing the number of free parameters, pg. 138, right column, third para.));
manage signal transmission pathways providing direct connections between non-adjacent regions of the neural network (columnar structures appear with intracolumnar connections, pg. 131, right column, third para.; The Examiner notes intracolumnar connections occur between neural network nodes in a layer) with signal modification and temporal coordination during transmission (Temporal crosstalk occurs when the network receives conflicting training information over time (iterations), e.g., when the network is forced to learn several dissimilar functions simultaneously. Pg. 137, left column, second para.)); and
implement a greedy neural system (there are greedy algorithms (e.g., gradient search algorithms) which can offer good approximations. Pg. 134 left column last para.)) that selectively processes activation patterns based on utility metrics (The supervised network, which offers better generalization," is used afterwards for selecting the winner sub-net. Pg. 143, right column, second para.; Repeatedly: the first hidden layer is modified for a certain number of iterations. Then, its activations are mapped onto the output layer in order to modify the rest of the weight matrix, pg. 139, second para.)),
wherein the greedy neural system (In most of the MNNs (modular neural networks) with voting modules, the Majority (Plurality) voting scheme is used, i.e., each module “nominates" one class and the class which takes the majority of the votes wins. Therefore, the Majority vote only uses the maximum activation at the output layer (which corresponds to the module's choice). Pg. 144, left column, second para.) comprises a competitive bidding manager (Unsupervised NNs, typically, contain two layers of neurons. One is the input layer, which receives the testing sample vector. The second is the output (competitive) layer, pg. 143, left column, first para.; The objective of a voting scheme in real-life elections is to represent the information available at the different bids in order to reach a “fair" final decision. Pg. 143 right column last para.)) that receives bids from the activation patterns competing for access to limited computational resources (The aim is, also, to reach a representative final decision according to the local opinions (activations) of the modules. Pg. 144, left column, first para.)) and
allocates the limited computational resources to high-utility activation patterns based on the bids (In most of the MNNs with voting modules, the Majority (Plurality) voting scheme is used, i.e., each module “nominates" one class and the class which takes the majority of the votes wins. Therefore, the Majority vote only uses the maximum activation at the output layer (which corresponds to the module's choice) pg. 144, left column, second para.)), and
wherein the hierarchical supervisory system coordinates the pruning decisions based on the utility metrics (Meanwhile, there is no general criteria for choosing the network's size (number of hidden nodes) since the parameters of each application demand different network capabilities...Examples are: determining the network's size by trial-and-error, starting learning with a very large network and then pruning unchanging nodes and/or connections, pg. 136, left column, second para.)).
Auda already teaches a three-layer neural network comprising a hidden layer, but does not explicitly teach a deep neural network having multiple hidden layers; computer system is configured to execute software instructions stored on nontransitory machine-readable storage media that: monitoring the deep learning network through multiple supervisory levels, stores successful modification and pruning patterns.
Sikka teaches a computer system is configured to execute software instructions stored on nontransitory machine-readable storage media that (Computer system, including a desktop computer, a laptop computer, a mobile device [0055]; Some embodiments include a non-transitory computer-readable medium storing program instructions that, when executed by a processor [0212]);
a deep neural network (deep neural networks can have complex network architectures that include numerous layers and intricate connection topologies [0052]);
monitoring the deep learning network through multiple supervisory levels (The AI design application also includes a network analyzer that analyzes the behavior of the neural network at the layer level, neuron level, and weight level in response to test inputs. Abstract; At step 2722, network evaluator 220 receives an expression that relates activation levels of a set of neurons [0161]),
stores successful modification and pruning patterns (network generator 200 could delete a corresponding portion of the model definition [0083]; network generator 200 receives a modification to the network architecture [0093]).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Auda to incorporate the teachings of Sikka for the benefit of analyzing and evaluating how a neural network operates relative to training data (Sikka [0164])
Regarding claim 11, claim 11 is similar to claim 1. It is rejected in the same manner and reasoning applying.
4. Claims 2-4, 6, 12-13 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Auda et al. (MODULAR NEURAL NETWORKS: A SURVEY, International Journal of Neural Systems, Vol. 9, No. 2 (April, 1999) 129-151) in view of Sikka et al. (US20210012210) and further in view of Yuan et al. ("STICKER: An energy-efficient multi-sparsity compatible accelerator for convolutional neural networks in 65-nm CMOS." IEEE Journal of Solid-State Circuits 55.2 (2019): 465-477)
Regarding claim 2, Auda and Sikka teaches the computer system of claim 1, Yuan teaches wherein the hierarchical supervisory system detects network sparsity using thresholds that adapt based on neural network state (The whole sparsity adaptive control/data flow shown in Fig. 5 treats different sparsity situations as different modes. We get activations mode (AM) from the sparsity adaptor and the weights mode (WM) from the offline analysis. Then convolutional operations are processed in nine different modes (three AMs × three WMs). Later, output activations go through the zero detector. They are classified into three different modes based on two thresholds given by users (TH1 and TH2), pg. 468, left col., last para.).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Auda and Sikka to incorporate the teachings of Yuan for the benefit of improving energy
efficiency by making full use of sparsity and the network sparsity can potentially lower storage and computation requirements (Yuan, abstract)
Regarding claim 3, Auda and Sikka teaches the computer system of claim 1, Yuan teaches wherein the hierarchical supervisory system exchanges information about resource availability and network sparsity across the multiple supervisory levels (For software optimization, a data schedule method is adopted to reduce possible collisions in the 2-way set-associative PEs when activations are in sparse mode….The scheduled results will be the input feature maps of the next layer without many collisions. An example is shown in Fig. 13(d). After this schedule flow, Data 2 and Data 3 are switched and all collisions in set0 and set1 are reduced, pg. 472, right col., second to the last para.; The Examiner notes Data 2 and Data 3 are exchanged).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Auda and Sikka to incorporate the teachings of Yuan for the benefit of improving energy
efficiency by making full use of sparsity and the network sparsity can potentially lower storage and computation requirements (Yuan, abstract)
Regarding claim 4, Auda and Sikka teaches the computer system of claim 1, Yuan teaches wherein the meta-supervisory system maintains operational stability of the deep learning network while identifying patterns (Weights after Pruning, Table 1, 467) across implemented pruning decisions (Weight pruning results on Alexnet Convolutional Layers Without Accuracy Loss, Table 1, 467, left col., The Examiner notes without accuracy loss indicates network stability).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Auda and Sikka to incorporate the teachings of Yuan for the benefit of improving energy
efficiency by making full use of sparsity and the network sparsity can potentially lower storage and computation requirements (Yuan, abstract)
Regarding claim 6, Auda and Sikka teaches the computer system of claim 1, Yuan teaches wherein managing the signal transmission pathways includes modifying signal strengths based on observed transmission effectiveness and detected network sparsity (A multi-sparsity control and data flow with an online sparsity adaptor. It can detect the sparsity of activations online and switch the chip state among nine different sparse modes, (pg. 466, left col., second para.); The whole multi-sparsity control and data flow are controlled by instructions … If the instruction number for one task is too large, the chip needs to be frequently blocked to transmit instructions between DRAM and on-chip instruction memory, pg. 469, left col., second to the last para.; The Examiner notes blocking the chip modifies signal strength to enable signal transmission of instructions).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Auda and Sikka to incorporate the teachings of Yuan for the benefit of improving energy
efficiency by making full use of sparsity and the network sparsity can potentially lower storage and computation requirements (Yuan, abstract)
Regarding claim 12, claim 12 is similar to claim 2. It is rejected in the same manner and reasoning applying.
Regarding claim 13, claim 13 is similar to claim 3. It is rejected in the same manner and reasoning applying.
Regarding claim 14, claim 14 is similar to claim 4. It is rejected in the same manner and reasoning applying.
Regarding claim 16, claim 16 is similar to claim 6. It is rejected in the same manner and reasoning applying.
5. Claims 7, 10, 17 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Auda et al. (MODULAR NEURAL NETWORKS: A SURVEY, International Journal of Neural Systems, Vol. 9, No. 2 (April, 1999) 129-151) in view of Sikka et al. (US20210012210) and further in view of Chen et al. ("An autonomous agent for negotiation with multiple communication channels using parametrized deep Q network." Mathematical Biosciences and Engineering 19.8 (2022): 7933-7951)
Regarding claim 7, Auda and Sikka teaches the computer system of claim 1, Chen teaches wherein the greedy neural system further comprises a local utility calculator that assigns value metrics to activation patterns based on novelty, gradient magnitude, or key performance indicators (We used the two evaluation metrics (pg. 7944, last para.); The first one is the average utility made between a volunteer and an agent, and the second one is the degree to winning friends, which is quantified through the questionnaire, pg. 7945 last sentence to pg. 7946 first para.).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Auda and Sikka to incorporate the teachings of Chen for the benefit of a novel deep reinforcement learning technique which generates an efficient strategy that can interact with different opponents, i.e., other negotiation agents or human players (Chen, abstract).
Regarding claim 10, Auda and Sikka teaches the computer system of claim 1, Chen teaches wherein the greedy neural system further comprises a feedback learning mechanism that optimizes utility assessment and intervention strategies based on historical outcomes (After updating the state, the deep reinforcement learning (DRL) policy outputs an action to the proposal generator guiding the optimal proposal ... We evaluate the MCAN agents and other rule-based negotiation agents based on the metrics, pg. 7934, second to the last para.).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Yuan to incorporate the teachings of Chen for the benefit of a novel deep reinforcement learning technique which generates an efficient strategy that can interact with different opponents, i.e., other negotiation agents or human players (Chen, abstract).
Regarding claim 17, claim 17 is similar to claim 7. It is rejected in the same manner and reasoning applying.
Regarding claim 20, claim 20 is similar to claim 10. It is rejected in the same manner and reasoning applying.
6. Claims 8, 9, 18 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Auda et al. (MODULAR NEURAL NETWORKS: A SURVEY, International Journal of Neural Systems, Vol. 9, No. 2 (April, 1999) 129-151) in view of Sikka et al. (US20210012210) and further in view of Mesadieu et al. ("Leveraging deep reinforcement learning technique for intrusion detection in SCADA infrastructure." IEEE Access 12 (2024): 63381-63399, date of publication 18 April 2024).
Regarding claim 8, Auda and Sikka teaches the computer system of claim 1, Auda and Sikka does not explicitly teach the limitations of claim 8.
Mesadieu teaches wherein the greedy neural system further comprises an anomaly detection framework that identifies statistically significant deviations in activation patterns (we propose a deep reinforcement learning (DRL) framework for anomaly detection in the SCADA network (abstract); DRL’s neural network application uses a function estimator to observe data state or labels’ input, as it operates in a set environment, and uses a greedy algorithmic policy to process unlimited network traffic’s history for anomalous and malicious attacks, pg. 63382, left col., second para.) and a response integration subsystem that implements real-time interventions (The results obtained, demonstrates that our DRL model can effectively classifies threats in real time and provides detection and response, pg. 63383, left col., second para.).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Auda and Sikka incorporate the teachings of Mesadieu for the benefit of learning patterns effectively (pg. 63389, Summary) and achieving an accuracy of 99.36% in attack detection which enhances the security of critical infrastructure (Mesadieu, abstract)
Regarding claim 9, Auda and Sikka teaches the computer system of claim 1, Auda and Sikka does not explicitly teach the limitations of claim 9.
Mesadieu teaches wherein the greedy neural system further comprises a local buffer management system that stores valuable activation patterns across multiple time steps (we designed a DRL framework that balances explorative strategies IV-B using epsilon-greedy as presented in (Algorithm 3) (pg. 63389, right col., section c);To address issues stemming from correlated data and non stationary distributions, we introduce a ‘‘ReplayMemory’’ function designed to store experiences and sample transitions (i.e., state−action−reward−next−state, tuples) that the agent encounters during its exploration of the environment. This memory buffer enables the agent to learn from past experiences by randomly sampling previous transitions from the memory during training, pg. 63383, left col., first para.) and
a hierarchical aggregation unit that synthesizes patterns across network regions (We continuously monitor our DQN performance by testing it on additional datasets to ensure its efficacy of real-time threat detection and response, thus simulating the automation of regular updates of new attack patterns and emerging security risks, pg. 63390, right col., last bullet point).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Auda and Sikka to incorporate the teachings of Mesadieu for the benefit of learning patterns effectively (pg. 63389, Summary) and achieving an accuracy of 99.36% in attack detection which enhances the security of critical infrastructure (Mesadieu, abstract)
Regarding claim 18, claim 18 is similar to claim 8. It is rejected in the same manner and reasoning applying.
Regarding claim 19, claim 19 is similar to claim 9. It is rejected in the same manner and reasoning applying.
7. Claims 5 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Auda et al. (MODULAR NEURAL NETWORKS: A SURVEY, International Journal of Neural Systems, Vol. 9, No. 2 (April, 1999) 129-151) in view of Sikka et al. (US20210012210) and further in view of Abraham et al. ("Back to the Future: Reversible Runtime Neural Network Pruning for Safe Autonomous Systems." 2024 Design, Automation & Test in Europe Conference & Exhibition (25 March 2024). IEEE, 2024).
Regarding claim 5, Auda and Sikka teaches the computer system of claim 1, Auda and Sikka does not explicitly teach the limitation of claim 5.
Abraham teaches wherein the hierarchical supervisory system establishes temporary support pathways to enable reversal of architectural changes during pruning (Our approach allows the pruned model to quickly revert to the full model when unsafe behavior is detected, pg. 6, right col, first para.).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Auda and Sikka to incorporate the teachings of Abraham for the benefit of enhancing safety and reliability, and providing seamless reversion to the accurate version of the model, demonstrating its applicability for safe autonomous systems design (Abraham, abstract)
Regarding claim 15, claim 15 is similar to claim 5. It is rejected in the same manner and reasoning applying.
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 extension fee 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 date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MORIAM MOSUNMOLA GODO whose telephone number is (571)272-8670. The examiner can normally be reached Monday-Friday 8am-5pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Michelle T Bechtold can be reached on (571) 431-0762. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/M.G./Examiner, Art Unit 2148
/MICHELLE T BECHTOLD/Supervisory Patent Examiner, Art Unit 2148
Prosecution Insights