DETAILED ACTION
1. This office action is in response to the Application No. 19054759 filed on 01/26/2026. Claims 1-18 are presented for examination and are currently pending.
Response to Arguments
2. On page 5 of the remarks, the Applicant argued that “The Examiner asserts that the claims are directed to a mental process, specifically "observing the operations of the base graph layer." Applicant respectfully disagrees. The claimed invention is not merely directed to monitoring or observation. Rather, it is directed to a specific improvement in the operation of a computer system, namely dynamically adapting network architecture during runtime operation through agent-based encoding optimization, generation, and pruning driven by real-time telemetry feedback”.
The above argument is not persuasive because the claims recited are so broad that the details in the specification are not reflected in the claims to be apparent to a person of ordinary skill in the art to recognize that the claimed invention leads to an improvement in technology. If the specification sets forth an improvement in technology, the claim should be evaluated to ensure that the claim itself reflects the disclosed improvement. That is, the claim includes the components or steps of the invention that provide the improvement described in the specification, 2106.04(d)(1). However, the instant claim discloses no details about what observing the operations of the base graph layer is actually doing to lead to an alleged improvement.
On page 6 of the remarks, the Applicant argued that “These limitations cannot practically be performed in the human mind. Collecting and analyzing encoding efficiency and resource utilization metrics across a network of interconnected computational agents requires technological components executing machine-level operations. Instantiating new agents from received encodings and removing agents from a layered network architecture based on detected bottlenecks and utilization patterns are operations performed on computational data structures during system execution, not mental observations. The Examiner's characterization isolates a single phrase while ignoring the claim as a whole …”.
The above argument is not persuasive because the collection and analyzing encoding efficiency and resource utilization metrics, the instantiation of new agents from received encodings and removing agents from a layered network architecture are additional elements as detailed in the 101 rejection. The Applicant needs to argue how the abstract ideas integrate into practical application by using the additional elements to demonstrate that the claims as a whole integrate into practical application and as a result the claims as a whole improve the functioning of the computer or technological field. Citations from the Applicant’s disclosure needs to cited to point out the improvement to technological field or functioning of the computer.
Furthermore, according to 2019 PEG states “the recitation of generic computer components in a claim does not preclude that claim from reciting an abstract idea. For instance, if a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it is still in the mental processes grouping unless the claim limitation cannot practically be performed in the mind.” In this case, the office action has correctly identified processes that can practically be performed in the human mind, but for recitation of generic computer components, and those generic computer components have been found to insignificantly apply the judicial exception, which, according to the 2019 PEG, is not indicative of an inventive concept.
On page 6 of the remarks, the Applicant argued that “Even assuming, arguendo, that the claims involve an abstract idea, they are integrated into a practical application under Step 2A, Prong Two. The claims include specific, meaningful limitations that apply any such idea in a concrete and constrained manner. For example, the claims recite: a base graph layer comprising interconnected computational agents that form the foundation for agent-based processing; telemetry agents that collect and analyze specific operational metrics, including encoding efficiency and resource utilization; dynamically-encoded agents that store encoded operational characteristics; agent generation that instantiates new agents from received encodings when operational metrics indicate processing bottlenecks; and agent pruning that removes agents when resource utilization patterns indicate redundant processing. The Examiner characterized these limitations as "field of use" under MPEP 2106.05(h). This characterization is incorrect. Field-of-use limitations merely state the environment in which an idea is applied. Here, the claim limitations define what the system does and how it operates”.
The above argument is not persuasive because the recited limitations are so broad that under the broadest reasonable interpretation, the limitations like: “a base graph layer comprising interconnected computational agents”, “telemetry agents that collect and analyze specific operational metrics, including encoding efficiency and resource utilization”, “agent generation that instantiates new agents …” and “agent pruning that removes agents …” falls under linking the use of a judicial exception (i.e., “monitoring operation of the base layer”) to a particular technological environment or field of use in the absence of details of how this system argued by the Applicant operates.
Furthermore, If the specification sets forth an improvement in technology, the claim should be evaluated to ensure that the claim itself reflects the disclosed improvement. That is, the claim includes the components or steps of the invention that provide the improvement described in the specification, 2106.04(d)(1). Here, there are no further details given in the claims as to how the agent generation is generated or how the agent pruning removes agents to achieve the improvement claimed by the Applicant.
On page 7 of the remarks, the Applicant argued that “Instantiating new agents from encodings in response to detected bottlenecks, and removing agents in response to utilization patterns, are concrete operational steps that modify the structure and operation of the system itself. Moreover, the telemetry-derived metrics are not mere data gathering or transmission. They are causally linked to structural adaptation of the running system. Applying telemetry-derived operational metrics to trigger agent instantiation and removal constitutes a practical application that improves system operation rather than merely manipulating data. These limitations impose meaningful restrictions that prevent preemption. Other adaptive systems remain outside the claim scope, including systems that scale through static provisioning, systems that adapt through parameter tuning rather than agent generation and pruning, systems that do not instantiate agents from received encodings, or systems that employ fixed agents rather than dynamically-encoded agents. The claims therefore do not preempt all approaches to adaptive network operation. Accordingly, the claims are not directed to a judicial exception”.
The above argument is not persuasive because the claims recited are so broad that the details in the specification are not reflected in the claims to be apparent to a person of ordinary skill in the art to recognize that the claimed invention leads to an improvement in technology. If the specification sets forth an improvement in technology, the claim should be evaluated to ensure that the claim itself reflects the disclosed improvement. That is, the claim includes the components or steps of the invention that provide the improvement described in the specification, 2106.04(d)(1). However, the instant claim discloses no details about how the instantiation of new agents from encodings in response to detected bottlenecks or removing agents in response to utilization patterns is actually achieved to modify the system that leads to an improvement as argued by the Applicant.
On page 7 of the remarks, the Applicant argued that “Even under Step 2B, the claims recite significantly more than any alleged judicial exception. The claim limitations reflect a non-conventional arrangement of technological features operating in an ordered combination. For example: dynamically-encoded agents that store encoded operational characteristics and adapt network operations based on performance objectives; agent generation triggered by telemetry-derived bottleneck detection, where new agents are instantiated from received encodings specifying agent characteristics; and agent pruning triggered by resource utilization patterns indicating redundant processing. These elements operate together such that telemetry analysis directly governs structural modification of the system through agent instantiation and removal. This behavior does not exist absent the claimed architecture”.
The Applicant argued that “the claim limitations reflect a non-conventional arrangement of technological features operating in an ordered combination”. This feature argued by the Applicant does not appear to be claimed. There is no such limitation that is directed to the ordered combination of the claimed invention or is there any limitation that is directed to an agent generation being triggered by the telemetry telemetry-derived bottleneck detection. These limitations are not reflected in the claims. As a result, the claims does not amount to significantly more than judicial exception.
On pages 7-8 of the remarks, the Applicant argued that “The Examiner has not provided any evidence-such as citations to patents, publications, or official notice-establishing that this specific telemetry-driven agent generation and pruning architecture is well-understood, routine, or conventional. Absent such evidence, the rejection cannot be sustained under Step 2B. See Berkheimer V. HP Inc., 881 F.3d 1360 (Fed. Cir. 2018). In summary, the claims are directed to a specific improvement in computer system operation through dynamically-encoded agents that are generated and pruned based on real-time operational metrics. The claims are not abstract. Even if viewed as involving an abstract idea, they are integrated into a practical application through meaningful limitations and recite significantly more. Applicant respectfully requests withdrawal of the § 101 rejection”.
It is noted that regarding Berkheimer, Berkheimer evidence is required for the
limitations identified as IESA and WURC. The only element identified as that is
the “telemetry agents collect and analyze operational metrics …”. The Examiner did show Berkheimer evidence for this element by identifying the court case in MPEP 2106.05(d)(II), example i.
As argued earlier, if the specification sets forth an improvement in technology, the claim should be evaluated to ensure that the claim itself reflects the disclosed improvement. That is, the claim includes the components or steps of the invention that provide the improvement described in the specification, 2106.04(d)(1). Here, there are no further details given in the claims as to how dynamically-encoded agents are generated or how the pruned based on real-time operational metrics achieves the improvement claimed by the Applicant.
Furthermore, it is noted that the limitation “pruned based on real-time operational metrics” is not claimed.
In addition, the 101 rejection is maintained and adjusted to reflect the newly
added limitations.
On pages 8-9 of the remarks, the Applicant argued that “Anticipation requires that each and every element of the claimed invention be explicitly or inherently disclosed in a single reference. Functional similarity, conceptual overlap, or general correspondence is insufficient to establish anticipation. Jain fails to disclose several limitations of independent Claim 1, including: 1. "agent generation that instantiates new dynamically-encoded agents from received encodings when the operational metrics indicate processing bottlenecks" The Examiner cites Jain [0478], which states that "for every new workload having one or more models to be optimized (e.g., compressed), optimization resources (e.g., exploration agents) are spawned from a ground state absent of anything learned from previously spawned agents." This passage teaches the opposite of what Claim 1 requires. The claim recites agent generation that instantiates new agents from received encodings. Jain explicitly teaches spawning agents "from a ground state absent of anything learned from previously spawned agents." An agent spawned from a ground state without prior encodings is not, and cannot be, an agent instantiated from received encodings, as expressly required by Claim 1. Jain's teaching directly contradicts the claimed limitation”.
On page 9 of the remarks, the Applicant argued that “Additionally, the claim requires that agent generation occur when the operational metrics indicate processing bottlenecks. The Examiner has not identified any passage in Jain where agents are generated in response to detected processing bottlenecks derived from operational metrics. Jain's agents are spawned for new workloads, not in response to bottleneck conditions identified during system operation”.
It is noted that since the limitation “agent generation that instantiates new dynamically-encoded agents from received encodings when the operational metrics indicate processing bottlenecks” is a new limitation. Jain has been used to remap the added new limitation.
Jain teaches agent generation (As a result, for every new workload having one or more models to be optimized (e.g., compressed), optimization resources (e.g., exploration agents) are spawned from a ground state absent of anything learned from previously spawned agents [0478]. The Examiner notes “to spawn” is a process creation of a new agent from an existing one) that instantiates new dynamically-encoded agents from received encodings (the FPGA circuitry D400 of the example of FIG. D4 includes interconnections and logic circuitry that may be configured and/or interconnected in different ways after fabrication to instantiate, for example, some or all of the machine readable instructions represented by the flowcharts disclosed herein in their entirety [0152]) when the operational metrics indicate processing bottlenecks (In certain examples, the cloud server ID4_C70 can leverage telemetry and mile marker data from the database ID4_B50 (e.g., implemented as a cloud-accessible database) to identify one or more bottlenecks such as a compute bottleneck, a memory bottleneck, an interconnect bottleneck, a communication bottleneck, etc. [0418]),
On page 9 of the remarks, the Applicant argued that “2. "agent pruning that removes dynamically-encoded agents when resource utilization patterns indicate redundant processing". The Examiner cites Jain [0051] and [0509], which describe pruning operations on neural network models. Specifically, [0509] states that "the agent learns to be more aggressive in pruning as it is the only way to improve performance via memory savings." Jain teaches model pruning, which involves removing weights or connections from a neural network model to compress it. The claim recites agent pruning, which is the removal of dynamically-encoded agents from the network architecture. These are fundamentally different operations. In Jain, the agent performs pruning on a model; in the claimed invention, agents themselves are pruned from the system. Furthermore, the claim requires that agent pruning occur when resource utilization patterns indicate redundant processing. The Examiner has not identified any passage in Jain where agents are removed from the system based on resource utilization patterns indicating redundancy. Jain's pruning is directed to model compression for performance improvement, not removal of redundant agents based on utilization patterns”.
It is noted that since the limitation “agent pruning that removes dynamically-encoded agents when resource utilization patterns indicate redundant processing” is a new limitation. Jain has been used to remap the added new limitation.
Jain teaches agent pruning that removes dynamically-encoded agents when resource utilization patterns indicate redundant processing (For example, an agent could learn that a fully-connected (FC) layer poses more redundancy than convolutional layer (e.g., more FC layers) [0509]).
On pages 9-10 of the remarks, the Applicant argued that “3. "a telemetry layer that monitors operations of the base graph layer, wherein telemetry agents collect and analyze operational metrics including encoding efficiency and resource utilization". The Examiner maps the "layer-wise mixed-precision sparsity policy predictor circuitry ID6_113" to the claimed telemetry layer. However, this circuitry operates within individual agents to determine convergence during iterative optimization. It does not constitute a separate telemetry layer that monitors operations of a base graph layer. The claim requires a telemetry layer as a distinct architectural component that monitors a base graph layer comprising interconnected computational agents. Jain's policy predictor circuitry is internal to each agent and monitors that agent's own iterative convergence. Jain does not disclose telemetry agents that collect and analyze operational metrics, including encoding efficiency and resource utilization, across multiple agents or across a base graph layer of interconnected agents”.
It is noted that since the limitation “wherein telemetry agents collect and analyze operational metrics including encoding efficiency and resource utilization” is a new limitation. Jain has been used to remap the added new limitation.
Jain teaches wherein telemetry agents (The three ID6_113 nodes in FIG. ID6_2 are the telemetry agents) collect and analyze operational metrics (During each iteration, the layer-wise mixed-precision sparsity policy predictor circuitry ID6_113 explores a potential solution and by the end of the iteration, the layer-wise mixed-precision sparsity policy predictor circuitry ID6_113 is converged to an optimal solution [0486]) including encoding efficiency (The agent is the predictor/inferencer that is reused during training. The output of the trained agent is a layer-wise or mixed-precision and/or sparsity policy [0486]; Because the learned embeddings are one/multi-hot encoded, they can be reused (e.g., reused on new target networks) [0512]) and resource utilization (Layer-wise mixed-precision configuration is a technique used to find the optimal configuration for every layer of a trained neural network so inference is accelerated, and accuracy is maintained [0486]);
On page 10 of the remarks, the Applicant argued that “4. "a base graph layer comprising interconnected computational agents" The Examiner maps agents A, B, and C (ID6_602, ID6_604, ID6_606) to the claimed base graph layer. However, Jain describes these as separate agents with independent requirements and different target hardware. For example, agent A has "an object detection workload, a VPU (Int8/4/2/1) target hardware, and a 2x latency improvement" requirement ([0499]). The claim requires a base graph layer comprising interconnected computational agents. The Examiner has not identified any passage in Jain establishing that agents A, B, and C are interconnected, communicate with one another, or form a graph structure. Merely depicting multiple agents in a figure does not establish interconnection, communication, or dataflow between those agents”.
The above argument is not persuasive because FIG. ID6_6 explicitly shows a base graph that shows an arrow connecting Agent A ID6_602 to Agent B ID6_604 to Agent C ID6_606. These citations of Jain reads on the limitation “a base graph layer comprising interconnected computational agents”.
Further, on page 10 of the remarks, the Applicant argued that “5. "dynamically-encoded agents that store encoded operational characteristics" .The Examiner cites Jain |[0511], which describes an "embedding layer in agent architecture to map models under compression to latent space representations." This passage describes embeddings of models being compressed, not agents that store their own encoded operational characteristics. Jain's embeddings represent properties of external models being optimized, not characteristics stored by the agents themselves. Accordingly, Jain fails to disclose this limitation”.
It is noted that since the limitation “dynamically-encoded agents that store encoded operational characteristics” is a new limitation. Jain has been used to remap the added new limitation.
Jain teaches one or more agent layers, wherein each agent layer comprises a plurality of dynamically-encoded agents (An example compressible operation embedder ID6_506 is example structure that employs embedding layers in agent architecture to map models under compression to latent space representations (vectors) [0511]) that store encoded operational characteristics (The experience replay buffer ID6_710 contains a historical policy, reward, feedback from compression environment and hardware evaluator are saved to substantiate the training of the Agent ID6_704 [0532]; In some examples, the example agent ID6_704 is scalable when implemented in a generalized architecture FIG. ID6_5 [0531])
Further, on page 10 of the remarks, the Applicant argued that “In summary, Jain fails to disclose: (1) agent generation from received encodings triggered by bottleneck detection; (2) agent pruning that removes agents from the system based on redundancy patterns; (3) a telemetry layer that monitors a base graph layer; (4) interconnected computational agents forming a base graph layer; and (5) dynamically-encoded agents that store encoded operational characteristics. Accordingly, Jain does not anticipate Claim 1, and the §102 rejection should be withdrawn”.
The arguments above are not persuasive because Jain still teaches the claimed limitations above and anticipates the independent claims 1 and 10. As a result, the anticipation rejections are maintained.
On page 11 of the remarks, the Applicant argued that “As argued above, the independent Claims 1 and 10 are not anticipated by Jain under §102. For the same reasons, Jain does not render obvious these independent claims under §103. Therefore, the dependent claims are patentable at least by virtue of their dependency on independent claims. Applicant therefore respectfully requests that all §103 rejections of dependent claims be withdrawn accordingly”.
The arguments above are not persuasive because Jain still teaches the claimed limitations above and anticipates the independent claims 1 and 10. As a result, the anticipation rejection is maintained. Also, the dependent claims are still obvious over the prior art of record. As a result, the obviousness rejections are maintained.
Furthermore, the Examiner notes that regarding independent claim 10, which is similar to claim 1, the Applicant’s argument are not persuasive for similar reasons argued above regarding claim 1. As a result, the independent claims are not patentable.
In addition, the Examiner notes dependent claims 2-9 and 11-18, which depend directly or indirectly from claims 1 and 10 are not allowable because the Applicant’s argument are not persuasive for similar reasons argued above regarding claim 1.
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.
3. Claims 1-18 are rejected under 35 U.S.C 101 because the claimed invention is directed towards an abstract idea without significantly more.
Step 1
Independent claim 1 is directed to a system, and falls into one of the four statutory categories.
Step 2A, Prong 1
Claim 1 recites the following abstract ideas:
monitors operations of the base graph layer (Mental process directed to
observing the operations of the base graph layer),
Step 2A, Prong 2
implement a layered network architecture (this limitation is directed to mere instruction to apply a judicial exception. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)) comprising:
a base graph layer comprising interconnected computational agents (this limitation is directed to linking the use of a judicial exception to a particular technological environment or field of use. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(h));
a telemetry layer (this limitation is directed to linking the use of a judicial exception to a particular technological environment or field of use. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(h)) that
wherein telemetry agents collect and analyze operational metrics including encoding efficiency and resource utilization (this limitation is directed to insignificant extra-solution activity of data transmission. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(g)); and
one or more agent layers, wherein each agent layer comprises a plurality of dynamically-encoded agents that store encoded operational characteristics and adapt network operations through encoding optimization, agent generation, and agent pruning that removes dynamically-encoded agents when resource utilization patterns indicate redundant processing based on network performance objectives (this limitation is directed to linking the use of a judicial exception to a particular technological environment or field of use. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(h)).
Step 2B, Prong 2
implement a layered network architecture (this limitation is directed to mere instruction to apply a judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f)) comprising:
a base graph layer comprising interconnected computational agents (this limitation is directed to linking the use of a judicial exception to a particular technological environment or field of use. This does not integrate the abstract idea into a practical application This does not amount to significantly more than judicial exception. See MPEP 2106.05(h));
a telemetry layer (this limitation is directed to linking the use of a judicial exception to a particular technological environment or field of use. This does not integrate the abstract idea into a practical application This does not amount to significantly more than judicial exception. See MPEP 2106.05(h)) that
wherein telemetry agents collect and analyze operational metrics including encoding efficiency and resource utilization (this limitation is directed Insignificant extra solution activity of data transmission and it is well understood routine and conventional. This does not amount to significantly more than judicial exception. See MPEP 2106.05(d)(II), example i); and
one or more agent layers, wherein each agent layer comprises a plurality of dynamically-encoded agents that store encoded operational characteristics and adapt network operations through encoding optimization, agent generation, and agent pruning that removes dynamically-encoded agents when resource utilization patterns indicate redundant processing based on network performance objectives (this limitation is directed to linking the use of a judicial exception to a particular technological environment or field of use. This does not integrate the abstract idea into a practical application This does not amount to significantly more than judicial exception. See MPEP 2106.05(h)).
4. Dependent claim 2 is directed to a system, and falls into one of the four statutory categories.
Claim 2 do not recites ant abstract ideas.
Claim 2 recite the following additional elements:
wherein agent encodings comprise dynamic representations of agent operational characteristics (this limitation is directed to linking the use of a judicial exception to a particular technological environment or field of use. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(h))
Claim 2 recite the following additional elements:
wherein agent encodings comprise dynamic representations of agent operational characteristics (this limitation is directed to linking the use of a judicial exception to a particular technological environment or field of use. This does not amount to significantly more than judicial exception. See MPEP 2106.05(h))
5. Dependent claim 3 is directed to a system, and falls into one of the four statutory categories.
Claim 3 recites the following abstract ideas:
wherein the telemetry layer implements continuous monitoring using adaptive kernel functions and topology-aware distance metrics (Mental process directed to observing and making a judgement of monitoring the layer using kernel functions and topology-aware distance metrics).
Claim 3 do not recite any additional elements.
6. Dependent claim 4 is directed to a system, and falls into one of the four statutory categories.
Claim 4 do not recite any abstract ideas.
Claim 4 recites the following additional elements:
network performance objectives comprise encoding costs, transmission costs, latency costs, and performance improvements (this limitation is directed to a particular type or source of data, which is field of use and it does not integrate the abstract idea into a practical application. See MPEP 2106.05(h)).
Claim 4 recites the following additional elements:
network performance objectives comprise encoding costs, transmission costs, latency costs, and performance improvements (this limitation is directed to a particular type or source of data, which is field of use and it does not amount to significantly more than judicial exception. See MPEP 2106.05(h)).
7. Dependent claim 5 is directed to a system, and falls into one of the four statutory categories.
Claim 5 do not recite any abstract ideas.
Claim 5 recites the following additional elements:
wherein agent generation comprises creating new agents from received encodings that specify agent characteristics (this limitation is directed to insignificant extra-solution activity of data gathering. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(g)).
Claim 5 recites the following additional elements:
wherein agent generation comprises creating new agents from received encodings that specify agent characteristics (this limitation is directed Insignificant extra solution activity of data gathering and it is well understood routine and conventional. This does not amount to significantly more than judicial exception. See MPEP 2106.05(d)(II), example i).
8. Dependent claim 6 is directed to a system, and falls into one of the four statutory categories.
Claim 6 do not recite any abstract ideas.
Claim 6 recites the following additional elements:
wherein agent pruning is based on resource utilization patterns and contribution to network objectives (this limitation is directed to linking the use of a judicial exception to a particular technological environment or field of use. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(h)).
Claim 6 recites the following additional elements:
wherein agent pruning is based on resource utilization patterns and contribution to network objectives (this limitation is directed to linking the use of a judicial exception to a particular technological environment or field of use. This does not amount to significantly more than judicial exception. See MPEP 2106.05(h)).
9. Dependent claim 7 is directed to a system, and falls into one of the four statutory categories.
Claim 7 do not recite any abstract ideas.
Claim 7 recites the following additional elements:
wherein the base graph layer implements a latent transformer core for processing encoded information (this limitation is directed to mere instruction to apply a judicial exception. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)).
Claim 7 recites the following additional elements:
wherein the base graph layer implements a latent transformer core for processing encoded information (this limitation is directed to mere instruction to apply a judicial exception. This does not amount to significantly more than judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f)).
10. Dependent claim 8 is directed to a system, and falls into one of the four statutory categories.
Claim 8 do not recite any abstract ideas.
Claim 8 recites the following additional elements:
wherein agent layers implement memory management through short-term and long-term memory systems (this limitation is directed to mere instruction to apply a judicial exception. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)).
Claim 8 recites the following additional elements:
wherein agent layers implement memory management through short-term and long-term memory systems (this limitation is directed to mere instruction to apply a judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f)).
11. Dependent claim 9 is directed to a system, and falls into one of the four statutory categories.
Claim 9 do not recite any abstract ideas.
Claim 9 recites the following additional elements:
wherein the layered network architecture implements error detection and recovery mechanisms during agent generation and pruning operations (this limitation is directed to mere instruction to apply a judicial exception. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)).
Claim 9 recites the following additional elements:
wherein the layered network architecture implements error detection and recovery mechanisms during agent generation and pruning operations (this limitation is directed to mere instruction to apply a judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f)).
12. Independent claim 10 is directed is directed to a method, and falls into one of the four statutory categories.
With regards to claim 10, it is substantially similar to claim 1, and is rejected in the same manner and reasoning applying.
13. Claim 11 is directed is directed to a method, and falls into one of the four statutory categories.
With regards to claim 11, it is substantially similar to claim 2, and is rejected in the same manner and reasoning applying.
14. Claim 12 is directed is directed to a method, and falls into one of the four statutory categories.
With regards to claim 12, it is substantially similar to claim 3, and is rejected in the same manner and reasoning applying.
15. Claim 13 is directed is directed to a method, and falls into one of the four statutory categories.
With regards to claim 13, it is substantially similar to claim 4, and is rejected in the same manner and reasoning applying.
16. Claim 14 is directed is directed to a method, and falls into one of the four statutory categories.
With regards to claim 14, it is substantially similar to claim 5, and is rejected in the same manner and reasoning applying.
17. Claim 15 is directed is directed to a method, and falls into one of the four statutory categories.
With regards to claim 15, it is substantially similar to claim 6, and is rejected in the same manner and reasoning applying.
18. Claim 16 is directed is directed to a method, and falls into one of the four statutory categories.
With regards to claim 16, it is substantially similar to claim 7, and is rejected in the same manner and reasoning applying.
19. Claim 17 is directed is directed to a method, and falls into one of the four statutory categories.
With regards to claim 17, it is substantially similar to claim 8, and is rejected in the same manner and reasoning applying.
20. Claim 18 is directed is directed to a method, and falls into one of the four statutory categories.
With regards to claim 18, it is substantially similar to claim 9, and is rejected in the same manner and reasoning applying.
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.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
21. Claims 1-4, 6, 8-15, 17 and 18 are rejected under 35 U.S.C 102(a)(1) as being anticipated by Jain et al. (US20240007414 PCT filed 06/25/2021)
Regarding claim 1, Jain teaches a computer system comprising a hardware memory (The one or more illustrative data storage devices/disks D110 may be embodied as one or more of any type(s) of physical device(s) configured for short-term or long-term storage of data such as, for example, memory devices, memory, circuitry, memory cards, flash memory, hard disk drives, solid-state drives (SSDs), and/or other data storage devices/disks [0114]),
wherein the computer system is configured to execute software instructions stored on nontransitory machine-readable storage media that: implement a layered network architecture (FIG. ID6_6 is an example of transfer learning techniques disclosed herein that reuse data corresponding to customers with different requirements to create a scalable model compression method for optimal platform specialization [0524]. The Examiner notes FIG. ID6_6 is a layered network architecture) comprising:
base graph layer (agent A ID6_602 → agent B ID6_604 → agent C ID6_606 [0527]) comprising interconnected computational agents (agent A ID6_602, agent B ID6_604 and agent C ID6_606 [0527], FIG. ID6_6; agent A ID6_202, agent B ID6_206, and agent C ID6_208 in FIG. ID6_2; FIG. ID6_6 is the same diagram as FIG. ID6_2 [0525]; The requirements of the agent A ID6_202 include an object detection workload, a VPU (Int8/4/2/1) target hardware, and a 2× latency improvement with ±1% accuracy goal [0499]);
a telemetry layer (Example layer-wise mixed-precision sparsity policy predictor circuitry ID6_113 [0486]; For example, identified features and associated telemetry data can be used to form one or more data sets for training, testing, and/or other validation of an AI model construct (e.g., AI model development) [0408]) that monitors operations of the base graph layer (Example layer-wise mixed-precision sparsity policy predictor circuitry ID6_113 is an example structure that takes the prediction from the agent ID6_110 and checks for changes in delta (e.g., monitoring for diminishing or increasing returns) to decide when to stop iterations when performance is no longer improving [0486]; An example agent A ID6_202 is an example of the agent ID6_110 [0499]; An example agent B ID6_206 is an example of the agent ID6_110 [0500]; An example agent C ID6_210 is an example of the agent ID6_110 [0501]),
wherein telemetry agents (The three ID6_113 nodes in FIG. ID6_2 are the telemetry agents) collect and analyze operational metrics (During each iteration, the layer-wise mixed-precision sparsity policy predictor circuitry ID6_113 explores a potential solution and by the end of the iteration, the layer-wise mixed-precision sparsity policy predictor circuitry ID6_113 is converged to an optimal solution [0486]) including encoding efficiency (The agent is the predictor/inferencer that is reused during training. The output of the trained agent is a layer-wise or mixed-precision and/or sparsity policy [0486]; Because the learned embeddings are one/multi-hot encoded, they can be reused (e.g., reused on new target networks) [0512]) and resource utilization (Layer-wise mixed-precision configuration is a technique used to find the optimal configuration for every layer of a trained neural network so inference is accelerated, and accuracy is maintained [0486]); and
one or more agent layers, wherein each agent layer comprises a plurality of dynamically-encoded agents (An example compressible operation embedder ID6_506 is example structure that employs embedding layers in agent architecture to map models under compression to latent space representations (vectors) [0511]) that store encoded operational characteristics (The experience replay buffer ID6_710 contains a historical policy, reward, feedback from compression environment and hardware evaluator are saved to substantiate the training of the Agent ID6_704 [0532]; In some examples, the example agent ID6_704 is scalable when implemented in a generalized architecture FIG. ID6_5 [0531]) and adapt network operations through encoding optimization (These embeddings will be learned during reinforcement learning operations performed by the agent. Because the learned embeddings are one/multi-hot encoded [0512], [0514]; One/multi-hot encoding can be reused and expanded on any set of optimization targets [0512]), agent generation (As a result, for every new workload having one or more models to be optimized (e.g., compressed), optimization resources (e.g., exploration agents) are spawned from a ground state absent of anything learned from previously spawned agents [0478]. The Examiner notes “to spawn” is a process creation of a new agent from an existing one) that instantiates new dynamically-encoded agents from received encodings (the FPGA circuitry D400 of the example of FIG. D4 includes interconnections and logic circuitry that may be configured and/or interconnected in different ways after fabrication to instantiate, for example, some or all of the machine readable instructions represented by the flowcharts disclosed herein in their entirety [0152]) when the operational metrics indicate processing bottlenecks (In certain examples, the cloud server ID4_C70 can leverage telemetry and mile marker data from the database ID4_B50 (e.g., implemented as a cloud-accessible database) to identify one or more bottlenecks such as a compute bottleneck, a memory bottleneck, an interconnect bottleneck, a communication bottleneck, etc. [0418]), and
agent pruning that removes dynamically-encoded agents when resource utilization patterns indicate redundant processing (For example, an agent could learn that a fully-connected (FC) layer poses more redundancy than convolutional layer (e.g., more FC layers) [0509]) based on network performance objectives (a method of conducting pruning operations in a semiconductor apparatus [0051]; the agent learns to be more aggressive in pruning as it is the only way to improve performance via memory savings [0509]; instant specification discloses “For example, telemetry agents 5620 may identify pruning candidates” (instant specification [0918])).
Regarding claim 2, Jain teaches the computer system of claim 1, Jain teaches wherein agent encodings comprise dynamic representations of agent operational characteristics (An example platform embedder ID6_508 is example structure that embeds layers in agent architecture … Categorical HW attributes such as CPU, VPU, GPU, FPGA, SKU, etc. capability types are also encoded in a similar fashion. These embeddings will be learned during reinforcement learning operations by the agent [0514]; An example static attributer ID6_510 … is a direct representation of attributes of model(s) under compression and the properties of the input target hardware [0516]).
Regarding claim 3, Jain teaches the computer system of claim 1, Jain teaches wherein the telemetry layer implements continuous monitoring using adaptive kernel functions (the processor circuitry to at least one of perform at least one of the first operations, the second operations or the third operations to: monitor, in a first phase, a hardware platform to identify features to train an artificial intelligence model [0442]; example adaptive kernels disclosed herein require approximately 6500 parameters to achieve the same accuracy, thereby enabling substantially reduced memory requirements [0692]) and topology-aware distance metrics (Additionally, examples disclosed herein allow users to scale faster and to dynamically convert their custom network topologies or variants to specialize across hardware platforms [0536]; These deployments may accomplish processing in network layers that may be considered as “near Edge”, “close Edge”, “local Edge”, “middle Edge”, or “far Edge” layers, depending on latency, distance, and timing characteristics [0091]).
Regarding claim 4, Jain teaches the computer system of claim 1, Jain teaches wherein network performance objectives comprise encoding costs, transmission costs, latency costs, and performance improvements (The quantization controller circuitry ID7_102 optimizes the training and inference of deep learning models to reduce (e.g., minimize) costs (e.g., memory costs, CPU costs, bandwidth consumption costs, accuracy tradeoff costs, storage space costs, etc.) [0571]).
Regarding claim 5, Jain teaches the computer system of claim 1, Jain teaches wherein agent generation comprises creating new agents (As a result, for every new workload having one or more models to be optimized (e.g., compressed), optimization resources (e.g., exploration agents) are spawned from a ground state absent of anything learned from previously spawned agents [0478].The Examiner notes “to spawn” is a process creation of a new agent from an existing one) from received encodings that specify agent characteristics (Categorical HW attributes such as CPU, VPU, GPU, FPGA, SKU, etc. capability types are also encoded in a similar fashion. These embeddings will be learned during reinforcement learning operations by the agent. As they are one/multi-hot encoded, the learned embeddings can be reused and expanded on any target hardware [0514]).
Regarding claim 6, Jain teaches the computer system of claim 1, Jain teaches wherein agent pruning is based on resource utilization patterns and contribution to network objectives (For example, the selected resource utilization model may describe a plurality of candidate models which shows the performance in response to the amount of resources utilized [0343]).
Regarding claim 8, Jain teaches the computer system of claim 1, Jain teaches agent layers implement memory management through short-term and long-term memory systems (The one or more illustrative data storage devices/disks D110 may be embodied as one or more of any type(s) of physical device(s) configured for short-term or long-term storage of data [0114]).
Regarding claim 9, Jain teaches the computer system of claim 1, Jain teaches wherein the layered network architecture implements error detection and recovery mechanisms during agent generation and pruning operations (an indication that training is to occur, an indication that new training data is present, errors being detected in the trained model [0773]; The training dataset(s) are used to recover the accuracy degradation caused by compression policies (e.g., quantization, pruning) [0482]).
Regarding claim 10, claim 10 is similar to claim 1. It is rejected in the same manner and reasoning applying.
Regarding claim 11, claim 11 is similar to claim 2. It is rejected in the same manner and reasoning applying.
Regarding claim 12, claim 12 is similar to claim 3. It is rejected in the same manner and reasoning applying.
Regarding claim 13, claim 13 is similar to claim 4. It is rejected in the same manner and reasoning applying.
Regarding claim 14, claim 14 is similar to claim 5. It is rejected in the same manner and reasoning applying.
Regarding claim 15, claim 15 is similar to claim 6. It is rejected in the same manner and reasoning applying.
Regarding claim 17, claim 17 is similar to claim 8. It is rejected in the same manner and reasoning applying.
Regarding claim 18, claim 18 is similar to claim 9. It is rejected in the same manner and reasoning applying.
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.
22. Claims 7 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Jain et al. (US20240007414 PCT filed 06/25/2021) in view of Amirloo et al. ("Latentformer: Multi-agent transformer-based interaction modeling and trajectory prediction." arXiv preprint arXiv:2203.01880 (2022))
Regarding claim 7, Jain teaches the computer system of claim 1, Jain does not explicitly teach wherein the base graph layer implements a latent transformer core for processing encoded information.
Amirloo teaches wherein the base graph layer implements a latent transformer core (Vision Transformer, Fig. 4, pg. 5, left col.; our approach is a conditional variational method that uses past trajectories and map information to form a discrete latent representation (pg. 2, right col., last para.); we use a vision transformer to focus on parts of the map representations that are relevant to each agent, pg. 2, left col., first para.) for processing encoded information (Multi-Resolution Encoding (Fig. 4, pg. 5, left col.); First, we implement a multi-resolution encoding scheme using local and global patches extracted from the map. These patches are then fed into a vision transformer to generate a sequence of learned representations (pg. 3, right col., second para.); Our method uses a self-attention module that encodes the interactions between agents according to their dynamics, pg. 2, left 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 Jain to incorporate the teachings of Amirloo for the benefit a novel transformer-based architecture for trajectory prediction in a multi-agent setting that produces diverse multi-modal predictions (pg. 2, left col., third para.)
Regarding claim 16, claim 16 is similar to claim 7. 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