DETAILED ACTION
1. This office action is in response to the Application No. 17177391 filed on 10/10/2025. Claims 2-4, 7-9, 13, 15-17, 19 and 20 has been cancelled and claims 1, 5, 6,10 ,11, 12, 14, 18 and 21-32 are presented for examination and are currently pending. Applicant’s arguments have been carefully and respectfully considered.
Notice of Pre-AIA or AIA Status
2. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Continued Examination Under 37 CFR 1.114
3. A request for continued examination under 37 CFR 1.114, including the fee set
forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this
application is eligible for continued examination under 37 CFR 1.114, and the fee set
forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action
has been withdrawn pursuant to 37 CFR 1.114. Applicant’s submission filed on
10/10/2025 has been entered.
Response to Arguments
4. It is noted that the objection to specification has been withdrawn because the Applicant has made the necessary correction required from the last Office Action.
The claim amendment of 10/10/2025 has overcome the 112(b) rejection of 08/26/2025. As a result, the 112(b) rejection has been withdrawn.
Additionally, the claim amendment of 10/10/25 has not overcome the 101 rejection of 08/26/2025. As a result, the 101 rejection is maintained and adjusted to reflect the newly added limitations.
On page 9 of the remarks, the Applicant argued that “Applicant respectfully traverses the §101 rejection. Properly construed, Claim 1 (i) does not recite a mental process (Step 2A, Prong 1); (ii) integrates any alleged abstract idea into a practical application (Step 2A, Prong 2); and (iii) in the alternative, recites significantly more than any alleged exception (Step 2B)”.
On page 9 of the remarks, the Applicant argued that “The Office groups three limitations as abstract ideas/mental steps: (a) computing a cost function, (b) using the cost to determine a new simulated demand curve, and (c) deciding that the curve is an "output" when a goal is reached. That characterization oversimplifies the claim and omits the computer-executed simulator-in-the-loop and neural network operations that anchor the claim”.
On page 10 of the remarks, the Applicant argued that “The USPTO's eligibility guidance explains that a claim is a "mental process" only if it can be practically performed in the human mind. Tasks that cannot practically be performed in the human mind are not mental processes. Running a domain-specific simulator, executing a neural network arranged to mirror physical adjacency, updating tensors via iterative minimization, and convergence checking over curves produced by simulation are computer technical operations beyond pen-and-paper cognition”.
The above argument is not persuasive because according to MPEP 2106.04(a)(2)(III)(C), a claim can still recite a mental process even if they are carried out on a computer component. The computing cost function and using the cost function to determine a simulated demand curve are mathematical concepts as detailed in the Office Action. The determining the new simulated demand curve is a mental process. As a result, the recitation of generic computer components in a claim does not preclude that claim from reciting an abstract idea.
On page 10 of the remarks, the Applicant argued that “ And in the August 4, 2025 USPTO memorandum, the Office reminded examiners-especially in AI/ML arts-that where claims recite operations the human mind is not equipped to perform, they are not "mental process" abstract ideas under Prong 1. ... Accordingly, Claim 1 is directed to a computerized simulator-in-the-loop neural control technique for a physical system, not to mental "observation/evaluation/judgment" The presence of a cost function within this larger technical pipeline does not convert the claim into a mathematical abstraction-just as an embedded equation did not doom eligibility in Diehr. See MPEP §2106 (integrated process analysis)”.
The above argument is not persuasive because according to MPEP 2106.04(a)(2)(III)(C), a claim can still recite a mental process even if they are carried out on a computer component. The using of a cost function are mathematical concepts ads detailed in the Office Action. As a result, the recitation of generic computer components in a claim does not preclude that claim from reciting an abstract idea.
On pages 10-11 of the remarks, the Applicant argued that “Even if the Office were to locate a judicial exception within isolated sub-steps, Claim 1 applies any such concept in a manner that improves technology and controls a machine: Particular machine / technological environment. The Neural Network's non-input layer neurons are located by physical adjacency of locations/zones-a graph topology mapped to the physical plant. That is not a generic "environment"; it is a specific Neural Network architecture constraint tied to the system's geometry, used during simulation-in-the-loop optimization. This is a meaningful limitation that changes how the model computes.
The above argument is not persuasive because the location of the non-input layer neurons as analyzed in the Office action does not integrate the abstract ideas into practical application. The fact that the architecture of the neural network is tied to the system geometry does not exclude it from being directed to a technological field which does not integrate the abstract ideas into practical application.
On pages 11 of the remarks, the Applicant argued that “Simulator-in-the-loop control. The method executes a domain-specific simulator, uses its outputs to iteratively update a demand curve via the Neural Network, and stops upon a goal state-then actuates the physical system using the resulting curve. That is an application that alters the operation of the system (e.g., energy delivery to zones), not mere data "gathering/output." The USPTO's guidance treats such improvements to computer or other technology as practical applications.
The above argument is not persuasive because as analyzed in the Office action, these limitations are instructions to implement the abstract ideas on a computer component. The iterative update of data (i.e., demand curve) are additional elements that does not integrate the abstract ideas into practical application, MPEP 2106.05(f).
On page 11 of the remarks, the Applicant argued that “Human-mind infeasibility. The 2024/2025 guidance updates instruct examiners to weigh whether the recited AI/ML operations are ones a human could perform; where they are not, Prong-2 integration is satisfied by concrete technological application-not a field-of-use statement. Thus, none of the examiner-identified "additional elements" are insignificant extrasolution activity when viewed in context. They implement a control strategy that changes how the physical system itself operates”.
On pages 11-12 of the remarks, the Applicant argued that “With respect to the limitation "Modifying at least a portion of the physical system by the energy represented by the output demand curve" - This is actuation of a physical system (e.g., energy/flow/temperature modifications across zones). Actuation is not "data storage/gathering"; it changes the state of a machine. ... The USPTO's 2019 PEG Examples 38 and 40 confirm that controlling a physical device or process using computed results (e.g., braking systems, circuit simulations) exemplifies such integration. Thus, device control is a hallmark of a practical application, distinguishing eligible technology-improving claims from abstract data processing”.
It is noted that the detailed 101 analysis in this rejection falls in line with the August 4th memo that was issued recently. The August 4th memo did not change the current guidance regarding 101 rejections. Furthermore, the statement that “They implement a control strategy that changes how the physical system itself operates” is a very broad. There is no details as to how the limitations implement a control strategy that changes how the physical system itself operation is performed. As a result, it encompasses insignificant extra solution activity such as mere data gathering. The argument regarding the “modifying at least a portion of the physical system by the energy represented by the output demand curve” still lacks details as to how that modification is performed. As a result, it falls under the insignificant extra solution activity such as mere data gathering.
On pages 12-13 of the remarks, the Applicant argued that “The Office labels multiple limitations "well-understood, routine, conventional (WURC)" without evidence. Under Berkheimer, WURC is a question of fact that requires support (e.g., admissions in the specification, court decisions, publications, or official notice with justification). Conclusory assertions are insufficient. Here, the ordered combination is not WURC: a graph-topology Neural Network expressly mirroring physical adjacency, executed within a simulator-in-the-loop optimization that iteratively minimizes a comfort-cost until a goal state and then actuates energy delivery to zones. This integrated architecture departs from generic "receive-process-display" patterns and reflects the sort of specific ML configuration and simulation that the USPTO's Examples 38-39 treat as eligible and non-mental. The Office Action identifies no evidence that this architecture- or its arrangement-was routine or conventional at the relevant time. If the Office maintains Step 2B, Applicant requests the Examiner (i) identify the precise limitations alleged WURC, and (ii) provide the requisite evidence for each, consistent with the Berkheimer memorandum”.
On page 13 of the remarks, the Applicant argued that “If the rejection is maintained, Applicant requests a limitation-by-limitation mapping identifying (a) the alleged judicial exception category, (b) how any remaining elements are deemed insignificant under Prong 2, and (c) the evidence for any WURC findings, consistent with MPEP §2106, the 2019 PEG/Oct. 2019 Update, the 2024 Guidance Update, and the Aug. 4, 2025 memorandum addressing AI/ML "mental step" analysis”.
The above argument is not persuasive. The limitations receiving a neural network of a physical system, the receiving a desired comfort curve for at least one of the zones within a physical system, receiving a simulated comfort curve as output and modifying at least a portion of the physical system by the energy represented by the output demand curve falls under insignificant extra solution activity such as mere data gathering/output steps, or mere data storage, according to MPEP 2106.5(g) that “Extra-solution activity includes both pre-solution and post-solution activity. An example of pre-solution activity is a step of gathering data for use in a claimed process”.
Furthermore, insignificant extra-solution activity are well understood routine and conventional, according to MPEP 2106.05(d)(II), example i.
It is noted that this Office Action include a detailed claim by claim analysis of 101 rejection and the detailed 101 analysis in the rejection falls in line with the August 4th memo that was issued recently. The August 4th memo did not change the current guidance regarding 101 rejections.
On page 14 of the remarks, the Applicant argued that “Kim does not disclose several critical limitations-individually or in the claimed combination-of Applicant's amended Claim 1. Specifically, Kim lacks: 1. A neural-network topology structurally defined by the physical system, i.e., non-input neurons connected with respect to location represented by the physical system; 2. A comfort-curve-based cost function computed between simulated and desired comfort curves; 3. A simulator-in-the-loop iterative process using that cost to update a simulated demand curve until a goal state is reached; and 4. A modification of the physical system by energy represented by the resulting output demand curve”.
On page 15 of the remarks, the Applicant has argued “Kim's ANNs model building thermal behavior using historical data, but the paper never discloses a network whose non-input neurons correspond to physical locations or are connected according to spatial relationships”.
The argument above is not persuasive because Kim clearly teaches “hidden layers for zone z”, pg. 4213, left col., and each zone corresponds to each physical perimeter of the multi-zone office building in Figure 7a. Furthermore, in Figure 7a, Perimeter 1 of the multi-zone office building is zone 2, Perimeter 2 is zone 3, Perimeter 3 is zone 4…This indicates a hidden layer comprising hidden neurons corresponds to physical perimeter locations in a building.
Furthermore, the Applicant’ s argument regarding “A modification of the physical system by energy represented by the resulting output demand curve” have been considered but are moot because a new secondary reference has been added.
On page 16 of the remarks, the Applicant argued that “The Applicant has argued “The claim requires that "non-input layer neurons are connected with respect to location represented by the physical system"- a structural adjacency rule that mirrors physical geometry (e.g., connections between zones that touch)”.
The argument above is not persuasive because Kim’s “hidden layers for zone z”, pg. 4213, left col., which corresponds to Perimeter 1 = zone 2, Perimeter 2 = zone 3, Perimeter 3 = zone 4…indicate hidden layers which are zones connected to each perimeter location in the multi-zone office building in Figure 7a.
On page 16 of the remarks, the Applicant argued that “In Kim, "thermal zones continuously interact" (p. 4216 L col. 1) merely describe empirical coupling that already exists in training data. "In other words, the Tt data have already captured coupling between zone z and the other zones." There is no disclosure that network connectivity itself is determined by zone adjacency. Kim's ANNs are generic multilayer networks with feedback loops and time-delay inputs, not a topologically-constrained graph as claimed. This limitation is therefore absent”.
The argument above is not persuasive because the claims do not recite “zone adjacency” or “topologically-constrained graph” as argued by the Applicant. It appears Applicant is arguing what is not claimed.
On page 16 of the remarks, the Applicant argued that “Claim 1 recites "computing a cost function using the simulated comfort curve and the desired comfort curve." Kim's objective (Eq. 13, 14-25, p. 4217 L col. 1) "aims at minimizing the economic operating cost of the HVAC system"; equations 14-25 represent electricity cost + penalties for violating temperature bounds and current and time-delayed power inputs. That objective does not compare a simulated comfort curve to a desired comfort curve; it simply ensures temperature constraints are satisfied while minimizing price”.
On pages 16-17 of the remarks, the Applicant argued that “Kim also reports normalized mean-square errors (NMSEs) to evaluate model accuracy (p. 4219 R col. last D. Those NMSEs are offline evaluation metrics, not the cost in the optimization loop. Thus, Kim's "cost" neither represents nor computes a curve-to-curve comfort error as claimed”.
The argument above is not persuasive because the claim 1 does not recite “compare a simulated comfort curve to a desired comfort curve” but recites “computing a cost function using the simulated comfort curve and the desired comfort curve”.
Furthermore, Kim discloses that the DNN performance is evaluated using the weighted sum ec of the NMSEs (normalized mean square errors) i.e., (1)] of Pt , Tzt , EC, and TV for all daily profiles in D (pg. 4219, right col., last para.). According to the instant specification that discloses cost function uses Mean Square Error (MSE); The DNN can serve as a price-and-optimal-demand curve (pg. 4220, left col., first full para.). This citation of Kim reads on the claimed limitation.
On page 17 of the remarks, the Applicant argued that “Claim 1 requires using the comfort-curve cost to determine a new simulated demand curve, then iteratively executing that loop until a goal state is reached and designating the converged curve as the output. Kim performs a single offline optimization of economic cost; iterations are chosen for convenience, not because of any triggering feature of the iteration. In Kim, the "maximum number of iterations for the optimal search was set to 500" (p. 4224 L col. 1). This refers to internal solver iterations-not iterative refinement of a simulated demand curve based on comfort-curve error. No disclosure exists of any "goal state" defined by convergence of comfort- curve distance, nor of the act of "determining that the new simulated demand curve is an output demand curve" upon reaching such a state. These limitations are absent”.
The argument above is not persuasive because Kim teaches that the maximum number of iterations for the optimal search was set to 500 in each simulation run, (pg. 4224, left col., first para.). This indicates a goal of 500 number of iterations was set for the optimal search in each simulation which reads on the limitation “iteratively executing the performing a machine learning process, the computing a cost function, and the using the cost function, until a goal state is reached”.
On page 17 of the remarks, the Applicant argued that “The final step of Claim 1-"modifying at least a portion of the physical system by the energy represented by the output demand curve"-requires real-world actuation. Kim expressly states, "The simulations were performed on a computer with a four-core 3.5 GHz CPU and 16 GB RAM" (p. 4223 R col. 1). No physical HVAC equipment is driven or modified. Observations that "Ttz decreases when Pₜ increases" (p. 4225 L col. 1) simply describe simulated relationships, not real control actions. Therefore, Kim fails to teach the claimed device control step that changes the physical system's state”.
It is noted that the Applicant’s argument has been considered but are moot because a new secondary reference has been added.
Furthermore, it appears the Applicant is arguing what is not claimed because the Applicant argued that “Kim fails to teach the claimed device control step that changes the physical system’s state”. There is no recitation in the claims that “device control step that changes the physical system’s state”.
On page 18 of the remarks, the Applicant argued that “Even, arguendo, if Kim separately mentions "neural network," "comfort range," and "power input," anticipation requires that the same combination and arrangement be disclosed. Kim's process (offline economic optimization using generic ANNs) differs fundamentally from Applicant's claimed simulator-in-the-loop comfort-optimization with spatially-mapped neural topology and physical actuation”.
The argument above is not persuasive because the Applicant is arguing what is not claimed. The claimed invention do not recite “simulator-in-the-loop comfort-optimization with spatially-mapped neural topology and physical actuation”.
On page 18 of the remarks, the Applicant argued that “Kim omits at least the following elements: 1. Neural-network topology connected with respect to physical location; 2. Computation of a comfort-curve-based cost; 3. Simulator-in-the-loop iterative updating to a goal state; and 4. Modification/actuation of the physical system according to the output demand curve. Because these limitations are not disclosed-individually or in the required combination-Claim 1 is not anticipated by Kim. Applicant, therefore, respectfully requests withdrawal of the § 102 rejection for claim 1. Independent Claims 11 and 18 are also allowable for similar reasons. The dependent claims each depend from an allowable base claim and are each allowable for at least that reason as well as for their own independent reasons. Therefore, Applicant respectfully requests that the § 102 rejection be withdrawn”.
It is noted that the Applicant’s argument has been considered but are moot because a new secondary reference has been added.
Independent claim 1 has now been remapped with Kim in view of Ozonat. Claims 11 and 18 are rejected with the same rationale as 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.
5. Claims 1, 5, 6,10 ,11, 12, 14, 18 and 21-32 are rejected under 35 U.S.C 101 because the claimed invention are directed to an abstract idea without significantly more.
Step 1
Independent claim 1 is directed to a method, and falls into one of the four statutory categories.
Step 2A, Prong 1
Claim 1 recites the following abstract ideas:
computing a cost function using the simulated comfort curve and the desired comfort curve (Mathematical concepts directed to computing a cost function);
using the cost function to determine a new simulated demand curve (Mathematical concepts directed to using a cost function to determine a demand curve. This step is performed by evaluating the cost function and making a judgement to determine the demand curve);
determining that the new simulated demand curve is an output demand curve upon the goal state being reached (Mental process directed to determining a demand curve. This step can be performed by evaluating the simulated demand curve and make a judgement that it’s an output).
Step 2A, Prong 2
Claim 1 recites the following additional elements:
receiving a neural network of a physical system (This limitation is directed to insignificant extra-solution activity mere data gathering. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(g))),
the physical system represented by non-input layer neurons representing locations in the physical system, at least some of the locations being zones the non-input layer neurons arranged such that the non-input layer neurons are connected with respect to location represented by the physical system (This limitation is directed to merely describing the technological environment. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(h));
receiving a desired comfort curve for at least one of the zones within the physical system (This limitation is directed to mere data gathering. This limitation does not integrate the abstract idea into practical application insignificant extra-solution activity of data transmission. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(g));
performing a machine learning process to run the neural network using a simulated demand curve as input (this limitation is directed to 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)) and
receiving a simulated comfort curve as output (this limitation is directed to insignificant extra-solution activity of mere data gathering. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(g));
iteratively executing the performing, computing, and using steps until a goal state is reached (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)); and
modifying at least a portion of the physical system by the energy represented by the output demand curve (this limitation encompasses mere insignificant extra solution activity, such as modifying mere data storage, or mere data gathering/output steps. This limitation does not integrate the abstract idea into practical application. See MPEP 2106.05(g)).
Step 2B
Claim 1 recites the following additional elements:
receiving a neural network of a physical system (This limitation is directed to insignificant extra-solution activity mere 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),
the physical system represented by non-input layer neurons representing locations in the physical system, at least some of the locations being zones the non-input layer neurons arranged such that the non-input layer neurons are connected when the physical system locations touch (This limitation is directed to merely describing the technological environment. This does not amount to significantly more than judicial exception. See MPEP 2106.05(h));
receiving a desired comfort curve for at least one of the zones within the physical system (this limitation is directed insignificant extra-solution activity of receiving data and it is a well understood routine and conventional activity. This limitation does not amount to significantly more. See MPEP 2106.05(d)(II), example i);
performing a machine learning process to run the neural network using a simulated demand curve as input (this limitation is directed to mere instruction to apply a judicial exception. This limitation does not amount to significantly more, see MPEP 2106.05(f)) and
receiving a simulated comfort curve as output (this limitation is directed to insignificant extra-solution activity of data transmission and it is a well understood routine and conventional activity. This limitation does not amount to significantly more. See MPEP 2106.05(d)(II), example i);
iteratively executing the performing, computing, and using steps until a goal state is reached (this limitation is directed to mere instruction to apply a judicial exception. This limitation does not amount to significantly more. see MPEP 2106.05(f));
modifying at least a portion of the physical system by the energy represented by the output demand curve (this limitation encompasses mere insignificant extra solution activity, such as modifying mere data storage, or mere data gathering/output steps. Steps of data storage or data gathering/output are similar to steps of storing and retrieving data in memory, and receiving/transmitting data over a network, which was found by the courts to be well understood, routine, and conventional activity (MPEP 2106.05(d)(II)). This limitation does not amount to significantly more.
6. Dependent claim 5 is directed to a method, and falls into one of the four statutory categories.
Claim 5 do not recite any abstract ideas.
Claim 5 recite the following additional elements:
wherein the goal state comprises the cost function being minimized, the neural network running for a specific time, or the neural network running a specific number of iterations (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into practical application. see MPEP 2106.05(h)).
Claim 5 recite the following additional elements:
wherein the goal state comprises the cost function being minimized, the neural network running for a specific time, or the neural network running a specific number of iterations (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not amount to significantly more, see MPEP 2106.05(h)).
7. Dependent claim 6 is directed to a method, and falls into one of the four statutory categories.
Claim 6 do not recite any abstract ideas.
Claim 6 recite the following additional elements:
wherein the neural network comprises at least one neuron with multiple activation functions (these limitations are directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into practical application. See MPEP 2106.05(h))).
Claim 6 recite the following additional elements:
wherein the neural network comprises at least one neuron with multiple activation functions (these limitations are directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not amount to significantly more than judicial exception. See MPEP 2106.05(h)).
8. Dependent claim 10 is directed to a method, and falls into one of the four statutory categories.
Claim 10 do not recite any abstract ideas.
Claim 10 recites the following additional elements:
wherein any neuron at least one neuron not in an output layer of the neural network may be an output data (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into practical application. see MPEP 2106.05(h)).
Claim 10 recites the following additional elements:
wherein any neuron at least one neuron not in an output layer of the neural network may be an output neuron (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not amount to significantly more, see MPEP 2106.05(h)).
9. Independent claim 11 is directed to a system, and falls into one of the four statutory categories.
With regards to claim 11, it is substantially similar to claim 1, and is rejected in
the same manner and reasoning applying.
Claim 11 further recites “a demand curve creation system, the system comprising: a processor, a memory in operable communication with the processor, and demand curve creation code residing in memory which comprises:” this limitations is directed to using a computer as a tool to perform the abstract idea. This limitation does not integrate the abstract idea into a practical application and does not amount to significantly more. See MPEP 2106.05(f)
10. Dependent claim 12 is directed to a system, and falls into one of the four statutory categories.
Claim 12 do not recite any abstract ideas.
Claim 12 recite the following additional elements:
wherein the machine learning process comprises using backpropagation that computes a cost function gradient for values in the neural network, and then uses an optimizer to update the simulated demand curve (this limitation is directed to mere instructions to apply an exception. This limitation does not integrate the abstract idea into practical application. see MPEP 2106.05(f)).
Claim 12 recite the following additional elements:
wherein the machine learning process comprises using backpropagation that computes a cost function gradient for values in the neural network, and then uses an optimizer to update the simulated demand curve (this limitation is directed to mere instructions to apply an exception. This limitation does not amount to significantly more, see MPEP 2106.05(f)).
11. Dependent claim 14 is directed to a system, and falls into one of the four statutory categories.
Claim 14 recite the following abstract ideas:
optimizer uses stochastic gradient descent or mini-batch gradient descent to minimize the cost function (Mathematical concepts directed to optimization of a neural network using gradient descent which is directed to use of mathematical algorithm).
Claim 14 do not recite any additional elements.
12. Independent claim 18 is directed to a machine, and falls into one of the four statutory categories.
With regards to claim 18, it is substantially similar to claim 1, and is rejected in
the same manner and reasoning applying.
Claim 18 further recites “a non-transitory computer-readable storage medium configured with executable instructions to perform a method for creation of a demand curve upon receipt of a comfort curve, the method comprising:” this limitations is directed to using a computer as a tool to perform the abstract idea. This limitation does not integrate the abstract idea into a practical application and does not amount to significantly more. See MPEP 2106.05(f)
13. Dependent claim 21 is directed to a method, and falls into one of the four statutory categories.
With regards to claim 21, it is substantially similar to claim 10, and is rejected in
the same manner and reasoning applying.
14. Dependent claim 22 is directed to a method, and falls into one of the four statutory categories.
With regards to claim 22, it is substantially similar to claim 10, and is rejected in
the same manner and reasoning applying.
15. Dependent claim 23 is directed to a method, and falls into one of the four statutory categories.
Claim 23 do not recite any abstract ideas.
Claim 23 recites the following additional elements:
wherein at least some of the non-input layer neurons represent walls within the
physical system (This limitation is directed to the description of the architecture of a neural network. This is generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into practical application. see MPEP 2106.05(h))
Claim 23 recites the following additional elements:
wherein at least some of the non-input layer neurons represent zones within the
physical system (This limitation is directed to the description of the architecture of a neural network. This is generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not amount to significantly more, see MPEP 2106.05(h))
16. Dependent claim 24 is directed to a system, and falls into one of the four statutory categories.
With regards to claim 24, it is substantially similar to claim 10, and is rejected in
the same manner and reasoning applying.
17. Dependent claim 25 is directed to a system, and falls into one of the four statutory categories.
With regards to claim 25, it is substantially similar to claim 10, and is rejected in
the same manner and reasoning applying.
18. Dependent claim 26 is directed to a machine, and falls into one of the four statutory categories.
With regards to claim 26, it is substantially similar to claim 10, and is rejected in
the same manner and reasoning applying.
19. Dependent claim 27 is directed to a machine, and falls into one of the four statutory categories.
With regards to claim 27, it is substantially similar to claim 10, and is rejected in
the same manner and reasoning applying.
20. Dependent claim 28 is directed to a machine, and falls into one of the four statutory categories.
Claim 28 recites the following abstract ideas:
wherein at least one neuron has at least two equations in an activation function
within the at least one neuron (Mathematical concepts directed to having two activation
functions in a neuron).
Claim 28 do not recite any additional elements.
21. Dependent claim 29 is directed to a method, and falls into one of the four statutory categories.
Claim 28 recites the following abstract ideas:
wherein at least one neuron has an activation function, the activation function
comprising an equation that models state moving through space (Mathematical
concepts directed to having an activation functions in a neuron that has an equation).
Claim 28 do not recite any additional elements.
22. Dependent claim 30 is directed to a system, and falls into one of the four statutory categories.
With regards to claim 30, it is substantially similar to claim 29, and is rejected in
the same manner and reasoning applying.
23. Dependent claim 31 is directed to a machine, and falls into one of the four statutory categories.
With regards to claim 31, it is substantially similar to claim 29, and is rejected in
the same manner and reasoning applying.
24. Dependent claim 32 is directed to a system, and falls into one of the four statutory categories.
Claim 32 do not recite any abstract ideas.
Claim 32 recites the following additional elements:
wherein the locations in the physical system comprise walls, windows, floors, or ceilings (This limitation is directed to the description of the architecture of a neural network. This is generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into practical application. see MPEP 2106.05(h)).
Claim 32 recites the following additional elements:
wherein the locations in the physical system comprise walls, windows, floors, or ceilings (This limitation is directed to the description of the architecture of a neural network. This is generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not amount to significantly more. see MPEP 2106.05(h)).
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.
25. Claims 20, 22, 24, 25, 27,29,31, 32, 34, 36, 38 and 39 are rejected under 35 U.S.C. 103 as being unpatentable over Kim ("A supervised-learning-based strategy for optimal demand response of an HVAC system." arXiv preprint arXiv:1904.13304 (2019), Manuscript received August 1, 2019; revised February 11, 2020 and March 29, 2020; accepted April 5, 2020. Date of publication April 8, 2020) in view of Ozonat et al. (US20200351171 filed 04/30/2019)
Regarding claim 1, Kim teaches a method of determining a demand curve (The SLAMP (supervised learning-aided meta-prediction) method is developed to directly schedule the optimal power inputs of the HVAC system for the electricity prices and building thermal conditions during the next 24 h. This significantly reduces computation time and produces a price-and-optimal-demand curve, pg. 4214, right col., second bullet point) implemented by one or more computers (The simulations were performed on a computer with a four-core 3.5-GHz CPU and 16 G of RAM, pg. 4223, right col., first para.)
comprising: receiving a neural network of a physical system (The multi-zone building was simulated using historical weather data, (pg. 4220, left col., last para.); The corresponding e(Tz, Tz) are high, implying that the ANNs successfully reflect the complicated thermal dynamics of the multi-zone building, pg. 4221, left col., first para. The Examiner notes the ANN (Artificial Neural Network) is a neural network of a building),
the physical system represented by non-input layer neurons representing locations in the physical system (HJz, HKz, HLz sets of the indices of neurons in the Jth, Kth, and Lth hidden layers for zone z; bj(k)z, oz bias values for the jth (or kth) hidden neuron and the output neuron for zone z, pg. 4213, left col.,. The Examiner notes hidden layer is a non-input layer, and the hidden layers has hidden neurons for zone z representing locations in building in Fig. 7a, left col., pg. 4220),
at least some of the locations being zones (Specifically, the building has six thermal zones, including an attic that is unoccupied, pg. 4220, left col., second to the last para.; Fig. 7a, pg. 4220),
the non-input layer neurons arranged such that the non-input layer neurons are connected (Specifically, an ANN with feedback loops, time-delayed inputs, multiple hidden layers, and different activation functions is implemented to estimate the temperature variations within each zone when the power input to the HVAC system changes, pg. 4214, left col., last para.) with respect to location represented by the physical system (Note that ANN training is achieved using the historical operating data of a multi-zone building, in which the thermal zones continuously interact. In other words, the Tzt data have already captured coupling between zone z and the other zones (pg. 4216, left col., first para.). The Examiner notes that physical locations touch in Fig. 7a);
receiving a desired comfort curve for at least one of the zones within the physical system (It is assumed that the occupants in zones z = 1, 3, and 4 feel comfortable for Tz t ,min = 20◦C ≤ Tz t ≤ Tz t ,max = 25◦C during 8 h ≤ t ≤ 19 h, and those in zones z = 2 and 5 are satisfied when Tz t is between 19◦C and 23◦C, pg. 4220, left col., second to the last para. The Examiner notes data of temperature in Celsius versus time can be plotted as curve for instance see pg. 4220, Fig. 8b);
performing a machine learning process to run the neural network using a simulated demand curve as input (The DNN can serve as a price-and-optimal-demand curve (pg. 4224, left col., first full para.); ANN receives output from DNN as input (Fig. 6c, pg. 4219, right col., first para.); Similarly, for Case 2, the time was defined as that required to determine the optimal Pt after replacing (5)–(25) with the DNN via Algorithm 1. ... The simulations were performed on a computer with a four-core 3.5-GHz CPU and 16 G of RAM, pg. 4223, right col., first para. The Examiner notes ANN (artificial neural network) receives a demand curve from DNN as input) and
receiving a simulated comfort curve as output (At each time t, α = 0 is assigned to the test score sct,γ ,d,c for the case where the output Tzt of the trained ANN for Pt + γ · P is smaller than Tzt for Pt + (γ − 1)· P, (pg. 4225, left col., first para.); output of ANN is Tzt , see Fig. 2, which means indoor temperature Tzt of each zone z at time t. Furthermore, Tzt as a function of time t is a curve, see page 4220, Fig. 8b);
computing a cost function using the simulated comfort curve and the desired comfort curve (The DNN performance is evaluated using the weighted sum ec of the NMSEs [i.e., (1)] of Pt , Tz t , EC, and TV for all daily profiles in D (pg. 4219, right col., last para.); The DNN can serve as a price-and-optimal-demand curve, pg. 4220, left col., first full para. The Examiner notes NMSEs means normalized mean square errors and instant specification discloses cost function uses Mean Square Error (MSE));
using the cost function to determine a new simulated demand curve (The objective function (13) aims at minimizing the operating cost of the HVAC system: i.e., the 24-h sum of CEt multiplied by Pt . The remaining terms represent the penalties incurred when Tzt goes above Tzt ,max by TztH or below Tzt ,min by …TztL, as shown in (14) and (15), respectively, pg. 4218, left col., first para. The Examiner notes cost function is an example of an objective function);
iteratively executing the performing a machine learning process, the computing a cost function, and the using the cost function, until a goal state is reached (The maximum number of iterations for the optimal search was set to 500 in each simulation run, pg. 4224, left col., first para.);
determining that the new simulated demand curve is an output demand curve upon the goal state being reached (Note that ANN training is achieved using the historical operating data of a multi-zone building, in which the thermal zones continuously interact. In other words, the Tzt data have already captured coupling between zone z and the other zones. Therefore, the ANN-based model for each zone z enables accurate estimation of Tzt for Pt and Et [i.e., Tz t = f (t, Pt , Et )] while reflecting interactions among all thermal zones. The ANN model for zone z can also be implemented by including the temperatures of other zones in the input variables [e.g., Tz=1 t = f (t, Pt , Et , Tz=1 t ), pg. 4216, left col., first para.); and
Kim does not explicitly teach modifying at least a portion of the physical system by energy represented by the output demand curve.
Ozonat teaches receiving a neural network of a physical system (As described above, when the nodes in the neural network correspond to … sensors [0062]; These nodes are intended to indicate that the neural network 600 can include any number of nodes corresponding to any number of … sensors, etc. [0067]; The sensors can include environmental sensors and on-device sensors. The environmental sensors can be provided throughout (e.g., internally and externally) the data center 100 or facility 100 f and provide information relative to their positioning or location therein [0031]),
the physical system represented by non-input layer neurons representing locations in the physical system (The model is run using the input values, and resulting values Y1 to Y4 are output. The outputs can be a vector including the model's predictions of the parameter values and/or the state of the parameter (e.g., an anomaly value), which can be a hidden state in the nodes S5 to S8 of the layer L+1 [0074]. The Examiner notes layer L+1 in Fig. 6 is a non-input layer and the nodes represent locations of sensors)
the non-input layer neurons arranged such that the non-input layer neurons are connected with respect to location represented by the physical system (FIG. 6 illustrates an exemplary neural network 600 corresponding to or encoded from the data center graph 500 of FIG. 5. [0056]; The on-device sensors are associated with (e.g., physically attached thereto) specific systems or sub-systems and provide information relative to those systems or sub-systems [0031])
modifying at least a portion of the physical system by energy (In Fig. 7, output of Neural Network 600 which represents the physical system is sent back as input into the neural network 600; A data center can control or manage its overall IT load (e.g., power consumption) using a job scheduler to achieve facility power consumption profiles [0044]) represented by the output demand curve (The outputs Y1 to Y4 can be the values of the parameters fws_water_supply_temp_sp, it_load … The outputs of the neural network 600, which can be the predicted values of the parameters (or systems, sub-systems, etc.) can be fed to an objective function 750, which can output an objective function result value. The Examiner notes load outputs Y1 to Y4 is a power profile).
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 Kim to incorporate the teachings of Ozonat for the benefit of optimizing the data centers, improving efficiencies, minimizing downtimes, and reducing power consumption (Ozonat [0001])
Regarding claim 5, Kim and Ozonat teaches the method of claim 1, Kim teaches wherein the goal state comprises the cost function being minimized, the neural network running for a specific time, or the neural network running a specific number of iterations (The objective function (13) aims at minimizing the operating cost of the HVAC system: i.e., the 24-h sum of CEt multiplied by Pt , pg. 4218, left col., first para. Examiner notes cost function is an example of an objective function).
Regarding claim 6, Kim and Ozonat teaches the method of claim 5, Kim teaches wherein the neural network comprises at least one neuron with multiple activation functions (Moreover, G is the number of hidden layers, and U is the set of Ug (the number of neurons in the gth hidden layer). Similarly, F is the set of Fg (the activation functions in the gth hidden layer) (pg. 4216, left col., last para.); This paper proposes a new SL-based strategy for optimal DR of an HVAC system in a multi-zone commercial building. Specifically, an ANN with feedback loops, time-delayed inputs, multiple hidden layers, and different activation functions is implemented to estimate the temperature variations within each zone when the power input to the HVAC system changes (pg. 4214, left col., last para.)).
Regarding claim 10, Kim and Ozonat teaches the method of claim 1, Kim teaches wherein at least one neuron not in an output layer of the neural network outputs data (In detail, the output njzt of the hidden neuron j ∈ H1z in the first hidden layer can be estimated as:
PNG
media_image1.png
76
542
media_image1.png
Greyscale
(pg. 4217, left col., first para.)).
Regarding claim 11, claim 11 is similar to claim 1. It is rejected in the same manner and reasoning applying. Further, Kim teaches a demand curve creation system, the system comprising: a processor, a memory in operable communication with the processor, and demand curve creation code residing in memory which comprises (Similarly, for Case 2, the time was defined as that required to determine the optimal Pt after replacing (5)–(25) with the DNN via Algorithm 1 …The simulations were performed on a computer with a four-core 3.5-GHz CPU and 16 G of RAM, pg. 4223, right col., first para. The Examiner notes algorithm 1 which is the code resides in the RAM (i.e, memory) of the computer):
Regarding claim 18, claim 18 is similar to claim 1. It is rejected in the same manner and reasoning applying. Further, Kim teaches a non-transitory computer-readable storage medium configured with executable instructions to perform a method for creation of a demand curve upon receipt of a comfort curve, the method comprising (The simulations were performed on a computer with a four-core 3.5-GHz CPU and 16 G of RAM, pg. 4223, right col., first para. The Examiner notes that RAM includes computer-readable storage medium that executes instructions to perform the method):
Regarding claim 21, Kim and Ozonat teaches the method of claim 1, Kim teaches wherein at least one neuron not in an input layer of the neural network accepts input (In addition, the output neuron produces:
PNG
media_image2.png
88
566
media_image2.png
Greyscale
The active function FO of the output layer is chosen to be the linear identity function, assuming that the multiple hidden layers enable the ANN to reflect the building thermal dynamics correctly, pg. 4217, right col., second to the last para. The Examiner notes that output neuron accepts data form hidden neuron which is not the output layer).
Regarding claim 22, claim 22 is similar to claim 10. It is rejected in the same manner and reasoning applying.
Regarding claim 23, Kim and Ozonat teaches the method of claim 1, Kim teaches wherein at least some of the non-input layer neurons (HJz, HKz, HLz sets of the indices of neurons in the Jth, Kth, and Lth hidden layers for zone z; bj(k)z, oz bias values for the jth (or kth) hidden neuron and the output neuron for zone z, pg. 4213, left col.,. The Examiner notes hidden layer is a non-input layer, and the hidden layers has hidden neurons for zone z representing locations in building in Fig. 7a, left col., pg. 4220) represent walls within the physical system (Specifically, the building has six thermal zones, including an attic that is unoccupied, pg. 4220, left col., second to the last para.; Fig. 7a, pg. 4220. The Examiner notes that zone z= 2, z=3, z=4 and z=5 are walls of the building.).
Regarding claim 24, claim 24 is similar to claim 21. It is rejected in the same manner and reasoning applying.
Regarding claim 25, claim 25 is similar to claim 22. It is rejected in the same manner and reasoning applying.
Regarding claim 26, claim 26 is similar to claim 22. It is rejected in the same manner and reasoning applying.
Regarding claim 27, claim 27 is similar to claim 21. It is rejected in the same manner and reasoning applying.
Regarding claim 28, Kim and Ozonat teaches the non-transitory computer-readable storage medium of claim 18, Kim teaches wherein at least one neuron has at least two equations in an activation function within the at least one neuron (Similarly, the output nkzt of the hidden neuron k ∈ HKz is calculated using the output mjzt of the activation function for the hidden neuron j ∈ HJz for J = K − 1, as:
PNG
media_image3.png
76
570
media_image3.png
Greyscale
In (3), mjzt can be further expressed using njzt, as shown in (4-a) and (4-b), when the activation function FJ is implemented using the sigmoid function or the ReLU, respectively, as:
PNG
media_image4.png
132
544
media_image4.png
Greyscale
pg. 4217, left col., first para.).
Regarding claim 29, Kim teaches the method of claim 1, Kim teaches wherein at least one neuron has an activation function (Similarly, the output nkzt of the hidden neuron k ∈ HKz is calculated using the output mjzt of the activation function for the hidden neuron j ∈ HJz for J = K − 1, as:
PNG
media_image3.png
76
570
media_image3.png
Greyscale
pg. 4217, left col., first para.),
the activation function comprising an equation that models state moving through space (In (3), mjzt can be further expressed using njzt , as shown in (4-a) and (4-b), when the activation function FJ is implemented using the sigmoid function or the ReLU, respectively, as:
PNG
media_image4.png
132
544
media_image4.png
Greyscale
pg. 4217, left col., first para. The Examiner notes that sigmoid function or ReLU function is used to model zones of the building which implies modelling a moving space).
Regarding claim 30, claim 30 is similar to claim 29. It is rejected in the same manner and reasoning applying.
Regarding claim 31, claim 31 is similar to claim 29. It is rejected in the same manner and reasoning applying.
Regarding claim 32, Kim and Ozonat teaches the system of claim 11, Kim teaches wherein the locations in the physical system comprise walls, windows, floors, or ceilings (Fig. 7(a) shows a building which comprises walls, windows, floors and ceilings. The building being a physical system).
26. Claims 12 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Kim ("A supervised-learning-based strategy for optimal demand response of an HVAC system." arXiv preprint arXiv:1904.13304 (2019), Manuscript received August 1, 2019; revised February 11, 2020 and March 29, 2020; accepted April 5, 2020. Date of publication April 8, 2020) in view of Ozonat et al. (US20200351171 filed 04/30/2019) and further in view of Sohn et al. (US20190360711)
Regarding claim 12, Kim and Ozonat teaches the demand curve creation system of claim 11, they do not explicitly teach wherein the machine learning process comprises using backpropagation that computes a cost function gradient for values in the neural network, and then uses an optimizer to update the simulated demand curve.
Sohn teaches wherein the machine learning process comprises using backpropagation that computes a cost function gradient for values in the neural network, and then uses an optimizer to update the simulated demand curve (As an error backprogates through layers, it becomes too small to recover the output in forward propagation [0065]; and updating the optimal power using the gradient descent algorithm via niter iterations S453 [0094]).
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 Kim and Ozonat to incorporate the teachings of Sohn for the benefit of generating a zone-based temperature prediction model by training an artificial neural network based on a plurality of first training data and minimizing a value of a loss function associated with a difference between a sequence of predetermined target temperatures and a sequence of predicted temperatures predicted based on the zone-based temperature prediction model (Sohn, abstract)
Regarding claim 14, Kim, Ozonat and Sohn teaches the demand curve creation system of claim 12, Sohn teaches wherein the optimizer uses stochastic gradient descent or mini-batch gradient descent to minimize the cost function (In one embodiment of the present disclosure, the applicant takes stochastic gradient descent (SGD) method to find the optimal network parameter θ* that minimizes the loss as follows: [0068]).
The same motivation to combine dependent claim 12 applies here.
Conclusion
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 8:00am-5:00pm 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