DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Drawings
The drawings are objected to because some text from figures 4-6 are blurry or illegible. Corrected drawing sheets in compliance with 37 CFR 1.121(d) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. The figure or figure number of an amended drawing should not be labeled as “amended.” If a drawing figure is to be canceled, the appropriate figure must be removed from the replacement sheet, and where necessary, the remaining figures must be renumbered and appropriate changes made to the brief description of the several views of the drawings for consistency. Additional replacement sheets may be necessary to show the renumbering of the remaining figures. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either “Replacement Sheet” or “New Sheet” pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance.
Claim Objections
Claims 16-20 are objected to because of the following informalities: Claim 15 recites “A non-transitory computer-readable medium containing instructions for classifying assets…” However, claims 16-20 recite “the computer-readable medium of claim …” Instead, claims 16-20 should read “the non-transitory computer-readable medium of claim …” Appropriate correction is required.
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.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea (mental process or math concept) without significantly more.
Claim 1:
Regarding claim 1, in step 1 of the 101-analysis set forth in MPEP 2106, the claim recites
“A method for classifying assets in a facility, the method comprising: receiving, by at least one processor, item point metadata and time series data associated with an asset; applying, by at least one processor, a machine learning neural network to the item point metadata and time series data to determine a similarity between metadata associated with the asset and metadata associated with a known asset type; and determining, by at least one processor, that the asset is of the asset type based on the similarity satisfying a predetermined threshold, wherein the machine learning neural network comprises anchor-based learning trained on a plurality of mined triplets each including an anchor input, a positive input, and a negative input, wherein the machine learning neural network is configured to generate vector representations associated with each of the anchor input, the positive input, and the negative input, and wherein a loss function applied to the vector representations is configured to differentiate between the vector representation associated with the positive input and the vector representation associated with the negative input,” and a method is one of the four statutory categories of invention.
In step 2A prong 1 of the 101-analysis set forth in the MPEP 2106, the examiner has determined that the following limitations recite a process that, under the broadest reasonable interpretation, covers a mental process or math concept but for recitation of generic computer components:
to determine a similarity between metadata associated with the asset and metadata associated with a known asset type, (This recites a mental process, since a person can mentally evaluate and determine a similarity between metadata of the asset and metadata of a known asset type, see MPEP 2106.04(a)(2)(III))
and determining, … that the asset is of the asset type based on the similarity satisfying a predetermined threshold…(This recites a mental process, since a person can mentally evaluate and determine that the asset is an asset type from evaluating that a similarity satisfies a preset threshold, see MPEP 2106.04(a)(2)(III)).
to generate vector representations associated with each of the anchor input, the positive input, and the negative input, (This recites a mental process, since a person can mentally evaluate and generate vector representations that are associated with each anchor input, see MPEP 2106.04(a)(2)(III)).
and wherein a loss function applied to the vector representations is configured to differentiate between the vector representation associated with the positive input and the vector representation associated with the negative input, (This recites a math concept, see specification [0044] state a loss function has math calculation in “A distance (e.g., cosine distance) between positive embedding 434 and anchor embedding 432 is less than a distance (e.g., cosine distance) between negative embedding 436 and anchor embedding 432, such that a loss function can differentiate between positive and negative embeddings 434, 436” – see MPEP 2106.04(a)(2), subsection I),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process or math concept but for the recitation of generic computer components, then it falls within the mental process or math concept grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
In step 2A prong 2 of the 101-analysis set forth in MPEP 2106, the examiner has determined that the following additional elements do not integrate this judicial exception into a practical application:
A method for classifying assets in a facility, the method comprising: receiving, …, item point metadata and time series data associated with an asset; (In step 2A, prong 2, this recites mere data gathering, which is considered insignificant extra-solution activity – see MPEP 2106.05(g)),
by at least one processor… (This recites a generic computer component being used as a tool – see MPEP 2106.05(f)),
applying, …, a machine learning neural network to the item point metadata and time series data, (This recites mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)),
wherein the machine learning neural network comprises anchor-based learning trained on a plurality of mined triplets each including an anchor input, a positive input, and a negative input, wherein the machine learning neural network is configured …, (This recites mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)),
Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is “directed” to an abstract idea.
In step 2B of the 101-analysis set forth in the 2019 PEG, the examiner has determined that the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above, additional elements vi, vii, and viii recite mere instructions to apply the judicial exception using generic computer components or a generic computer component being used as a tool, which are not indicative of significantly more. The additional element v recites mere data gathering, and is considered an insignificant extra-solution activity. In step 2B, this insignificant extra-solution activity is a well understood routine and conventional activity, which includes receiving or transmitting data over a network from court case Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016), – see MPEP 2106.05(d) (II)(i)),
Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Claim 2:
Regarding claim 2, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 2 recites the following additional element:
The method of claim 1, wherein the machine learning neural network comprises: a sentence transformer for transforming the item point metadata; and a time series transformer for transforming the time series data, (In step 2A, prong 2, this is considered mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)), (In step 2B, this is also considered mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)),
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 3:
Regarding claim 3, it is dependent upon claim 2, and thereby incorporates the limitations of, and corresponding analysis applied to claim 2. Further, claim 3 recites the following additional elements:
The method of claim 2, wherein the neural network further comprises: a feedforward neural network … and a layer normalization module, (In step 2A, prong 2, this is considered mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)), (In step 2B, this is also considered mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)),
configured for receiving output from the sentence transformer and the time series transformer; (In step 2A, prong 2, this recites mere data gathering, which is considered insignificant extra-solution activity – see MPEP 2106.05(g),). In step 2B, this insignificant extra-solution activity is well understood routine and conventional activity which includes receiving or transmitting data over a network from court case Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016) – see MPEP 2106.05(d) (II)(i),
… configured for receiving output from the feedforward neural network, (In step 2A, prong 2, this recites mere data gathering, which is considered insignificant extra-solution activity – see MPEP 2106.05(g),). In step 2B, this insignificant extra-solution activity is well understood routine and conventional activity which includes receiving or transmitting data over a network from court case Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016) – see MPEP 2106.05(d) (II)(i),
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 4:
Regarding claim 4, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 4 recites the following additional element:
The method of claim 1, wherein the machine learning neural network comprises zero shot learning, (In step 2A, prong 2, this is considered mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)), (In step 2B, this is also considered mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)),
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 5:
Regarding claim 5, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 5 recites the following abstract idea:
The method of claim 1, wherein the predetermined threshold at least about 30%, (this recites a mental process, since a person can mentally evaluate and set a threshold quantity, see MPEP 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic computer components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 6:
Regarding claim 6, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 6 recites the following additional elements:
The method of claim 1, further comprising: receiving, …, asset name data associated with the asset, (In step 2A, prong 2, this recites mere data receiving, which is considered insignificant extra-solution activity – see MPEP 2106.05(g),). In step 2B, this insignificant extra-solution activity is well understood routine and conventional activity which includes receiving or transmitting data over a network from court case Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016) – see MPEP 2106.05(d) (II)(i),
by at least one processor…, (In step 2A, prong 2, this recites a generic computer component being used as a tool – see MPEP 2106.05(f)), (In step 2B, this also recites a generic computer component being used as a tool – see MPEP 2106.05(f)),
… wherein the neural network comprises an ensemble model, (In step 2A, prong 2, this is considered mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)), (In step 2B, this is also considered mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)),
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 7:
Regarding claim 7, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 7 recites the following abstract idea:
The method of claim 1, further comprising: generating, … a model of the assets in the facility based on the determination of asset type, (this recites a mental process, since a person can mentally evaluate and generate a model of assets after determining by asset type, see MPEP 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic computer components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Further, claim 7 recites the following additional elements:
by at least one processor, (In step 2A, prong 2, this is considered a generic computer component being used as a tool – see MPEP 2106.05(f)), (In step 2B, this also recites a generic computer component being used as a tool – see MPEP 2106.05(f)),
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 8:
Regarding claim 8, in step 1 of the 101-analysis set forth in MPEP 2106, the claim recites
“A computer system for classifying assets in a facility, the computer system comprising: at least one memory having processor-readable instructions stored therein; and at least one processor configured to access the memory and execute the processor-readable instructions, which when executed by the processor configure the processor to perform a plurality of functions, including functions for: receiving item point metadata and time series data associated with an asset; applying a machine learning neural network to the item point metadata and time series data to determine a similarity between metadata associated with the asset and metadata associated with a known asset type; and determining that the asset is of the asset type based on the similarity satisfying a predetermined threshold, wherein the machine learning neural network comprises anchor-based learning trained on a plurality of mined triplets each including an anchor input, a positive input, and a negative input, wherein the machine learning neural network is configured to generate vector representations associated with each of the anchor input, the positive input, and the negative input, and wherein a loss function applied to the vector representations is configured to differentiate between the vector representation associated with the positive input and the vector representation associated with the negative input, and a system or machine is one of the four statutory categories of invention.
In step 2A prong 1 of the 101-analysis set forth in the MPEP 2106, the examiner has determined that the following limitations recite a process that, under the broadest reasonable interpretation, covers a mental process or math concept but for recitation of generic computer components:
to determine a similarity between metadata associated with the asset and metadata associated with a known asset type, (This recites a mental process, since a person can mentally evaluate and determine a similarity between metadata of the asset and metadata of a known asset type, see MPEP 2106.04(a)(2)(III))
and determining that the asset is of the asset type based on the similarity satisfying a predetermined threshold…(This recites a mental process, since a person can mentally evaluate and determine that the asset is an asset type from evaluating that a similarity satisfies a preset threshold, see MPEP 2106.04(a)(2)(III)).
… to generate vector representations associated with each of the anchor input, the positive input, and the negative input, (This recites a mental process, since a person can mentally evaluate and generate vector representations associated with each anchor input, see MPEP 2106.04(a)(2)(III)).
and wherein a loss function applied to the vector representations is configured to differentiate between the vector representation associated with the positive input and the vector representation associated with the negative input, (This recites a math concept, see specification [0044] state a loss function has a math calculations in “A distance (e.g., cosine distance) between positive embedding 434 and anchor embedding 432 is less than a distance (e.g., cosine distance) between negative embedding 436 and anchor embedding 432, such that a loss function can differentiate between positive and negative embeddings 434, 436” – see MPEP 2106.04(a)(2), subsection I),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process or math concept but for the recitation of generic computer components, then it falls within the mental process or math concept grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
In step 2A prong 2 of the 101-analysis set forth in MPEP 2106, the examiner has determined that the following additional elements do not integrate this judicial exception into a practical application:
A computer system for classifying assets in a facility, the computer system comprising: at least one memory having processor-readable instructions stored therein; and at least one processor configured to access the memory and execute the processor-readable instructions, which when executed by the processor configure the processor to perform a plurality of functions, including functions for…( This recites a generic computer component being used as a tool – see MPEP 2106.05(f)),
receiving item point metadata and time series data associated with an asset; (In step 2A, prong 2, this recites mere data gathering, which is considered insignificant extra-solution activity – see MPEP 2106.05(g)),
applying a machine learning neural network to the item point metadata and time series data, (This recites mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)),
wherein the machine learning neural network comprises anchor-based learning trained on a plurality of mined triplets each including an anchor input, a positive input, and a negative input, wherein the machine learning neural network is configured to generate vector representations associated with each of the anchor input, the positive input, and the negative input, (This recites mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)),
Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is “directed” to an abstract idea.
In step 2B of the 101-analysis set forth in the 2019 PEG, the examiner has determined that the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above, additional elements v, vii, and viii recite mere instructions to apply the judicial exception using generic computer components or a generic computer component being used as a tool, which are not indicative of significantly more. The additional element vi recites mere data gathering, and is considered an insignificant extra-solution activity. In step 2B, this insignificant extra-solution activity is a well understood routine and conventional activity, which includes receiving or transmitting data over a network from court case Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016), – see MPEP 2106.05(d) (II)(i)),
Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Claims 9-14:
Regarding claim 8, all of claim 8’s dependent claims follow the deficiencies of their parent claim. Since claims 9-14 recite similar limitations as corresponding claims 2-7 listed above, and are rejected for similar reasons under 35 U.S.C. 101.
Claim 15:
Regarding claim 15, in step 1 of the 101-analysis set forth in MPEP 2106, the claim recites “A non-transitory computer-readable medium containing instructions for classifying assets in a facility, the non-transitory computer-readable medium storing instructions that, when executed by at least one processor, configure the at least one processor to perform: receiving item point metadata and time series data associated with an asset; applying a machine learning neural network to the item point metadata and time series data to determine a similarity between metadata associated with the asset and metadata associated with a known asset type; and determining that the asset is of the asset type based on the similarity satisfying a predetermined threshold, wherein the machine learning neural network comprises anchor-based learning trained on a plurality of mined triplets each including an anchor input, a positive input, and a negative input, wherein the machine learning neural network is configured to generate vector representations associated with each of the anchor input, the positive input, and the negative input, and wherein a loss function applied to the vector representations is configured to differentiate between the vector representation associated with the positive input and the vector representation associated with the negative input, ” and a non-transitory computer-readable medium or machine is one of the four statutory categories of invention.
In step 2A prong 1 of the 101-analysis set forth in the MPEP 2106, the examiner has determined that the following limitations recite a process that, under the broadest reasonable interpretation, covers a mental process or math concept but for recitation of generic computer components:
to determine a similarity between metadata associated with the asset and metadata associated with a known asset type, (This recites a mental process, since a person can mentally evaluate and determine a similarity between metadata of the asset and metadata of a known asset type, see MPEP 2106.04(a)(2)(III))
and determining that the asset is of the asset type based on the similarity satisfying a predetermined threshold…(This recites a mental process, since a person can mentally evaluate and determine that the asset is an asset type from evaluating that a similarity satisfies a preset threshold, see MPEP 2106.04(a)(2)(III)).
… to generate vector representations associated with each of the anchor input, the positive input, and the negative input, (This recites a mental process, since a person can mentally evaluate and generate vector representations associated with each anchor input, see MPEP 2106.04(a)(2)(III)),
and wherein a loss function applied to the vector representations is configured to differentiate between the vector representation associated with the positive input and the vector representation associated with the negative input, (This recites a math concept, see specification [0044] state a loss function has math calculations in “A distance (e.g., cosine distance) between positive embedding 434 and anchor embedding 432 is less than a distance (e.g., cosine distance) between negative embedding 436 and anchor embedding 432, such that a loss function can differentiate between positive and negative embeddings 434, 436” – see MPEP 2106.04(a)(2), subsection I),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process or math concept, but for the recitation of generic computer components, then it falls within the mental process or math concept grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
In step 2A prong 2 of the 101-analysis set forth in MPEP 2106, the examiner has determined that the following additional elements do not integrate this judicial exception into a practical application:
A non-transitory computer-readable medium containing instructions for classifying assets in a facility, the non-transitory computer-readable medium storing instructions that, when executed by at least one processor, configure the at least one processor to perform…(This recites a generic computer component being used as a tool – see MPEP 2106.05(f)),
receiving item point metadata and time series data associated with an asset; (In step 2A, prong 2, this recites mere data gathering, which is considered insignificant extra-solution activity – see MPEP 2106.05(g)),
applying a machine learning neural network to the item point metadata and time series data, (This recites mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)),
wherein the machine learning neural network comprises anchor-based learning trained on a plurality of mined triplets each including an anchor input, a positive input, and a negative input, wherein the machine learning neural network is configured to generate vector representations associated with each of the anchor input, the positive input, and the negative input, (This recites mere instructions to apply an exception using generic computer – see MPEP 2106.05(f)),
Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is “directed” to an abstract idea.
In step 2B of the 101-analysis set forth in the 2019 PEG, the examiner has determined that the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above, additional elements v, vii, and viii recite mere instructions to apply the judicial exception using generic computer components or a generic computer component being used as a tool, which are not indicative of significantly more. The additional element vi recites mere data gathering, and is considered an insignificant extra-solution activity. In step 2B, this insignificant extra-solution activity is a well understood routine and conventional activity, which includes receiving or transmitting data over a network from court case Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016), – see MPEP 2106.05(d) (II)(i)),
Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Claims 16-18, 19-20:
Regarding claim 15, all of claim 15’s dependent claims follow the deficiencies of their parent claim. Since claims 16-18, and 19-20, recite similar limitations as corresponding claims 2-4, and 6-7, respectively listed above, and are rejected for similar reasons under 35 U.S.C. 101.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1, 6, 8, 13, 15, and 19 are rejected under 35 U.S.C. 103 over Li, S. et al., in “Relation inference among sensor time series in smart buildings with metric learning,” published in April 3, 2020, available at https://scholar.archive.org/work/b2yv64ahizb2zojqqtotbohg74/access/wayback/https://aaai.org/ojs/index.php/AAAI/article/download/5900/5756, and https://doi.org/10.1609/aaai.v34i04.5900 , (hereafter, LI), in view of Mishra, S. et al., in “Unified architecture for data-driven metadata tagging of building automation systems,” published on September 15, 2020, available at https://www.sciencedirect.com/science/article/pii/S0926580520309912?ref=pdf_download&fr=RR-2&rr=9f4fb933ba18200c , (hereafter, MISHRA), and further in view of Koh, J., (US PG Pub. No. US20230359830A1), filed May 5, 2023, (hereafter KOH).
Claim 1:
Regarding claim 1, LI teaches “A method for classifying assets in a facility, the method comprising: receiving, by at least one processor, item point metadata and time series data associated with an asset;”
See LI in page 4684, section Related work, describe “there are two different approaches to acquiring relationships between sensors: parsing the sensor metadata (i.e., point names as shown in Fig. 1) and inferring from the sensor time series readings.” Here, LI shows that this method covers evaluating sensor metadata by point names (i.e. item point metadata), and sensor time series readings (i.e. time series data associated with an asset).
Further, see LI in page 4683, in Introduction, describe “a necessary first step in deploying any smart building application would be to obtain the contextual information about the points in the building.” LI further elaborates to obtain context information, which also relates to receiving information regarding the metadata and time series data.
Further, LI also teaches “wherein the machine learning neural network comprises anchor-based learning trained on a plurality of mined triplets each including an anchor input, a positive input, and a negative input, wherein the machine learning neural network is configured to generate vector representations associated with each of the anchor input, the positive input, and the negative input, and wherein a loss function applied to the vector representations is configured to differentiate between the vector representation associated with the positive input and the vector representation associated with the negative input.”
See LI in page 4686, Deep metric learning neural network section, describe “we perform metric learning via a triplet network to capture the relative relatedness among sensors instead. The triplet network is comprised of three identical feed forward networks with shared parameters. In each iteration, a mini-batch of training triplets consisting of an anchor sensor Xa, accompanied with a pair of positive sensor Xp (i.e., a sensor in functional/spatial relation) and negative sensor Xn (i.e., a sensor not in functional/spatial relation) are fed into the triplet network. In one triplet T , Xa and Xp are sampled from the group of related sensors (e.g., those in the same room), while Xn is sampled from the non-related groups. When fed with a triplet, the network outputs the corresponding embedding vectors ya, yp, and yn. The objective of the network is to learn an embedding space such that the anchor sensor is closer to the positive sensor than to the negative sensor.” See figure 3 for details. In the paper, LI describes a neural network that embeds the input tensors into 1-D feature vector, (i.e. relates to vector representation associated with the inputs). From the study, LI also mentions that the method involves evaluating if a network with embedding space contains sensors that are closely related to each other (i.e. positive input), and if sensors are not related (i.e. negative input), and are compared to with an anchor sensor Xa (i.e. anchor input).
PNG
media_image1.png
568
1482
media_image1.png
Greyscale
Further, See LI in page 4686, section Deep metric learning triplet network, describes "embedding vectors ya, yp, and yn. The objective of the network is to learn an embedding space such that the anchor sensor is closer to the positive sensor than to the negative sensor, i.e., dp = ||ya − yp||2 <dn = ||ya − yn||2. We achieve this objective by combining the triplet loss
(Weinberger and Saul 2009) and the angular loss (Wang et al. 2017). Specifically, the triplet loss is defined as [see (3)],"
PNG
media_image2.png
730
768
media_image2.png
Greyscale
Here, LI describes a triplet loss function in item (3), where this function incorporates the positive sensors and negative sensors, and whether the embedding space (i.e. vector representation) can differentiate between the two sensor inputs by dp and dn.
Further, LI teaches “applying, by at least one processor, a machine learning neural network …”
See LI in page 4685, in figure 3, state “Through multiple convolutional layers, the neural network embeds the input tensors into 1-D feature vectors, which encode the relatedness among the sensing time series.” Here, LI explicitly states using a neural network.
However, LI did not teach:
“A method for classifying assets in a facility, the method comprising: receiving, by at least one processor, …”
“applying, by at least one processor, a machine learning neural network to the item point metadata and time series data to determine a similarity between metadata associated with the asset and metadata associated with a known asset type; ”
“and determining, by at least one processor, that the asset is of the asset type based on the similarity satisfying a predetermined threshold,”
In an analogous field, MISHRA teaches “applying, … , a machine learning neural network to the item point metadata and time series data to determine a similarity between metadata associated with the asset and metadata associated with a known asset type;”
See MISHRA in page 2, section 1.2 Literature review, describe "Automatically analyzing time-series data for metadata generation requires the identification and quantification of the unique characteristics of each data stream. In general, machine-learning-based approaches to time-series data analysis can be divided into two steps: (i) characterization of descriptive data features and (ii) clustering or classification of the data based on these features [17]. Methods for building metadata generation based on time-series analysis have explored various techniques for performing these two tasks..." Here, MISHRA describes using a machine learning approach such as clustering or classification to group metadata from time-series information.
Further, see MISHRA in section 1.2, page 3, first two paragraphs, describe a method that "identify the specific rooms in which sensors are located by correlating expected and actual sensor readings with the known HVAC energy input into a room. By identifying a single sensor in a room, this process can be identified with previous work to map all sensor locations in a building. However, this approach requires some a priori knowledge regarding the building's HVAC system, such as the building layout and the mapping of heating/cooling setpoints – calling for increased human intervention...that time-series analysis depends both on the methods for characterizing the data and for classifying or clustering those characterizations. For much of the work done in this field, the classification or clustering techniques are based on classical machine learning algorithms." Note the examiner construes the asset to be any asset information from incoming data record, from the specification in [0059] mention “receiving asset name data associated with the asset,” and the term ‘known asset’ is any asset or facility that is already recorded on profile or identified in general, such as the asset being an Air Handling Unit, Chiller, or other equipment from [0039]. Here, MISHRA shows using machine learning algorithm to perform classification on identifying between expected sensor data (i.e. an asset) and actual sensor reading data (i.e. metadata associated with a known asset type). The term ‘actual’ relates to ‘known’, and this compares one asset with an already existing data about a specific asset, which also relates to MISHRA’s ‘a priori knowledge regarding the building's HVAC system’ indicating known asset. See MISHRA in paragraph 1, section 2.1 for more details.
Later, see MISHRA in section 5.2 Unified Architecture, page 10 describes "UA first clusters points names based on k-mer analysis of raw point names, then compares the applied tags of points within each cluster. The UA computes and applies outlier scores that are then used to identify points in the cluster which have markedly different tags applied. This process assumes that points with high similarity in the raw point names will have highly similar tags, and therefore flags outliers for manual inspection." Here, MISHRA mentions a method called unified architecture called UA that compares tags of points within each cluster and calculating similarities (i.e. determine a similarity between raw point names which relate to metadata ), the method finds points that have different tags to identify any dissimilar points.
Further, MISHRA teaches “and determining, …, that the asset is of the asset type based on the similarity satisfying a predetermined threshold,”
See MISHRA, in page 10 describe "The iterative procedure continues until the maximum cluster similarity dips below a given threshold. Figure 3 shows a similarity matrix for all of the points in a given building before and after clustering. " Here, MISHRA mentions determining if a cluster similarity is below a given threshold relates to repeating a procedure until the similarity goes below a given or predetermined threshold. MISHRA from figure 3 also shows similarity measures of asset data. Note the examiner construes the asset is of the asset type to be any asset information from the data record, from the specification in [0059] mention “receiving asset name data associated with the asset,” and the term ‘asset type’ is any asset or facility that is already identified in the data being analyzed, such as the asset being an Air Handling Unit, Chiller, or other equipment from [0039].
Further, see MISHRA in section 1.2, page 3, first two paragraphs, describe a method that "identify the specific rooms in which sensors are located by correlating expected and actual sensor readings with the known HVAC energy input into a room. By identifying a single sensor in a room, this process can be identified with previous work to map all sensor locations in a building. However, this approach requires some a priori knowledge regarding the building's HVAC system, such as the building layout and the mapping of heating/cooling setpoints” MISHRA shows if the asset is classified into a specific asset type, such as if the asset is an air conditioner, heater, chiller, etc. or other equipment, then this relates to the asset belongs to the asset type (i.e. asset is of the asset type).
PNG
media_image3.png
724
1390
media_image3.png
Greyscale
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of LI and incorporate into the teachings of MISHRA because both references teach classifying assets in a facility by using similarity to compare the metadata and time series data of assets with anchor-based learning.
One of ordinary skill in the art would be motivated to do so because such a method “provides an opportunity for leveraging operator knowledge most efficiently to manually assess only the tags applied to the points flagged by the k-mers block, for enhancing the accuracy of the overall result,” (see MISHRA, page 10, section 5.2 Unified Architecture), and having “ automated application of metadata to BAS objects to efficiently enable the necessary interoperability to achieve GEB goals,” (see MISHRA, page 2, section 1.2 Literature Review).
However, Li in view of MISHRA did not teach:
“A method for classifying assets in a facility, the method comprising: receiving, by at least one processor, …”
“applying, by at least one processor, a machine learning neural network … ”
“and determining, by at least one processor,…”
In an analogous art, KOH teaches “by at least one processor, …”
See KOH in [0009] describe "Also described herein, in certain embodiments, are systems comprising at least one computing device comprising at least one processor, a memory, and instructions executable by the at least one processor to create an application comprising a data retrieving module configured to retrieve a plurality of data sets". Here, KOH describes using at least one processor with a memory, that runs instructions to, in this case, create an application.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of LI along with the secondary reference of MISHRA with the teachings of KOH by using the teachings from LI and MISHRA of classifying assets in a facility by using similarity and anchor-based learning, with KOH’s teaching of a system with a processor and memory.
One of ordinary skill in the art would be motivated to do so because by integrating KOH’s framework into the methods of LI and MISHRA, one with ordinary skill in the art would achieve the goal of providing a method with the “ use of a semantic map can help to improve energy efficiency and the quality of life. A semantic map may provide rich information to downstream optimization and provide actionable insights” (see KOH in [0031]).
Claim 6:
Regarding claim 6, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 1.
Further, KOH teaches “The method of claim 1, further comprising: receiving, by at least one processor, asset name data associated with the asset, …”
See KOH in [0009] describe " systems comprising at least one computing device comprising at least one processor, a memory, and instructions executable by the at least one processor to create an application comprising a data retrieving module configured to retrieve a plurality of data sets". Here, KOH describes using at least one processor with a memory, that runs instructions to, in this case, create an application.
See KOH in [0044-0045] describe " the NLP may process the text and extract the relevant information. In some embodiments, based on at least in part on the data structure of the documentations, one or more types of NLP techniques, or combinations thereof, may be selected to extract data from the documentations. For example, if information regarding the names and functions of different devices or data points in an IoT system is of interest, the system may use named entity recognition (NER) techniques to identify and classify named entities in the text. Alternatively or additionally, if the system is attempting to extract general information regarding the overall structure and content of the documentation (e.g., which pages of the documentation may contain the content of interest), techniques like part-of-speech tagging or sentence parsing may be selected and utilized to analyze the syntax and semantics of the text. In some embodiments, the NLP may extract the names and definitions of different data points (i.e., what information does the IoT device provide, the location of the IoT device, individual pieces of data collected by sensors, etc.)." Here, KOH describes classifying the names of different devices using named entity recognition (NER) techniques to identify and classify named entities in the text and relates to (i.e. asset name data).
Later, see KOH in [0127] describe “When the input data from a number of IoT devices associated with a building is received, the system of present disclosure may identify one or more base classifiers using a gating model. The IoT devices may comprise, for example, assembly line with sensors, actuators, and effectors, lighting systems, HVAC systems with temperature sensors, etc. The gating model may decide which base classifiers' predictions should be considered in calculating the final output based on the quality of the different input data types (e.g., whether the textual metadata is more informative than the time series data) and the similarity of the input data to the base classifiers' training data”. Here, KOH shows that input data from a device is received by the system. The device can include sensors, HVAC systems with temperature sensors, and other assets.
Further, see KOH in [0041-0042] describe "or example, controller manufacturers such as, by way of non-limiting examples, Carrier and Honeywell publish documentation recommending naming conventions of data points. For example, in a Carrier equipment operation manual, sa\_temp means Supply Air Temperature, and the base classifier may infer the semantics of the same name occurring in a target building. In some embodiments, the dictionaries provide an initial set of insights for understanding the semantics of the data, such as the actual usage of the data points in a given building. [0042] In some embodiments, the implementation of the dictionary look-up technique (i.e., construction of the base classifier) can take various forms, such as using natural language processing (NLP) models or sets of regular expressions. Often, it is a simple and error-resistant approach to use exact matching to the dictionary, which involves looking for an exact match between the name of a data point and a term in the dictionary." Here, KOH further specifies the types of assets can include air temperature supplier. KOH shows the sa\_temp example to illustrate an asset name data that is associated with the asset type of air temperature supplier.
Further, KOH teaches “…, wherein the neural network comprises an ensemble model.”
See KOH in [0064] describe "In some embodiments, the ensemble model may be a trainable ML model, such as Bagging algorithms, Boosting algorithms,…"
Further, see KOH describe in [0070] "Next, the process 200 may proceed to operation 204, wherein the system may select one or more base classier. In some embodiments, the system may utilize a gating model to make the selection. In some embodiments, the gating model may comprise a decision tree, neural network-based gating models, or rule-based gating models, etc. In some embodiments, the gating model may decide which base classifiers are most likely to make accurate predictions for a given input, and it can then pass the input on to the selected base classifiers for further processing," Here, KOH describes using an ensemble model as part of the system, and later mentions using a neural network model for a classifier model.
See KOH in [0074] describe "Next, the process 200 may proceed to operation 206, wherein the system may ensemble the one or more gate classifiers. In some embodiments, the system may utilize a pooling model to ensemble the gate classifiers." Here, KOH connects the gating model of a neural network connecting with an ensemble model for processing.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the references of LI and MISHRA, and incorporate with the teachings of KOH by using the teachings from LI and MISHRA of classifying assets in a facility by using similarity and anchor-based learning, with KOH’s teaching of a system with a processor that receives asset name data associated with the asset using a neural network with an ensemble model.
One of ordinary skill in the art would be motivated to do so because by integrating KOH’s framework into the methods of LI and MISHRA, one with ordinary skill in the art would achieve the goal of providing a method with the “ use of a semantic map can help to improve energy efficiency and the quality of life. A semantic map may provide rich information to downstream optimization and provide actionable insights” (see KOH in [0031]).
Claim 8:
Regarding claim 8, KOH further teaches “A computer system for classifying assets in a facility, the computer system comprising: at least one memory having processor-readable instructions stored therein; and at least one processor configured to access the memory and execute the processor-readable instructions, which when executed by the processor configure the processor to perform a plurality of functions, including functions…”
See KOH in [0096] describe "Computer system 100 may provide functionality for the components depicted in FIG. 1 as a result of the processor(s) 101 executing non-transitory, processor-executable instructions embodied in one or more tangible computer-readable storage media, such as memory 103, storage 108," Here, KOH teaches a computer system.
Further, see KOH in [0009] describe "Also described herein, in certain embodiments, are systems comprising at least one computing device comprising at least one processor, a memory, and instructions executable by the at least one processor to create an application comprising a data retrieving module configured to retrieve a plurality of data sets". Here, KOH describes using at least one processor, a memory, that runs instructions to, in this case, create an application.
Regarding claim 8, the claim recites similar limitations as corresponding independent claim 1 and is rejected for similar reasons as claim 1 using similar teachings and rationale.
Claim 13:
Regarding claim 13, LI in view of MISHRA, further in view of KOH teach the limitations in claim 8. Regarding claim 13, the claim recites similar limitations as corresponding claim 6 and is rejected for similar reasons as claim 6 using similar teachings and rationale.
Claim 15:
Regarding claim 15, KOH further teaches “A non-transitory computer-readable medium containing instructions for classifying assets in a facility, the non-transitory computer-readable medium storing instructions that, when executed by at least one processor, configure the at least one processor to perform…”
See KOH in [0096] describe "Computer system 100 may provide functionality for the components depicted in FIG. 1 as a result of the processor(s) 101 executing non-transitory, processor-executable instructions embodied in one or more tangible computer-readable storage media, such as memory 103, storage 108," Here, KOH teaches a computer system.
Later, see KOH describe in [0112] “ the platforms, systems, media, and methods disclosed herein include one or more non-transitory computer readable storage media encoded with a program including instructions executable by the operating system of an optionally networked computing device.” Here, KOH teaches a non-transitory computer readable storage media encoded with a program including instructions for a device.
Further, see KOH in [0009] describe "Also described herein, in certain embodiments, are systems comprising at least one computing device comprising at least one processor, a memory, and instructions executable by the at least one processor to create an application comprising a data retrieving module configured to retrieve a plurality of data sets". Here, KOH describes using at least one processor, a memory, that runs instructions to, in this case, create an application.
Regarding claim 15, the claim recites similar limitations as corresponding independent claim 1 and is rejected for similar reasons as claim 1 using similar teachings and rationale.
Claim 19:
Regarding claim 19, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 15. Regarding claim 19, the claim recites similar limitations as corresponding claim 6 and is rejected for similar reasons as claim 6 using similar teachings and rationale.
Claims 2, 3, 9, 10, 16, and 17 are rejected under 35 U.S.C. 103 over LI in view of MISHRA, further in view of KOH, further in view of Waterworth, D. et al., in “Advancing smart building readiness: automated metadata extraction using neural language processing methods,” published in August 25, 2021, available at https://www.sciencedirect.com/science/article/pii/S2666792421000330 , (hereafter, WATERWORTH), and further in view of Deng, X. et al., in "Toward Smart Multizone HVAC Control by Combining Context-Aware System and Deep Reinforcement Learning," published on November 1, 2022, available at https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9776508&tag=1 , (hereafter, DENG).
Claim 2:
Regarding claim 2, LI in view of MISHRA, further in view of KOH, teach the limitations in claim 1.
However, LI in view of MISHRA, further in view of KOH, did not teach “The method of claim 1, wherein the machine learning neural network comprises: a sentence transformer for transforming the item point metadata; and a time series transformer for transforming the time series data.”
In an analogous field, WATERWORTH teaches “The method of claim 1, wherein the machine learning neural network comprises: a sentence transformer for transforming the item point metadata;”
See WATERWORTH in section 1.1 Background, page 2 describes “The objective of this study is to answer the research question if the use of transformers [23], and pretrained language models such as Bidirectional Encoder Representations from Transformers (BERT) [24] can be used for building metadata inference. Because of the lack of suitable preprocessing tools [25], these models and methods have not, to the best of our knowledge, been evaluated for building metadata inference.”
Further, see WATERWORTH in page 3, section 1.1 describes the model “BERT uses Masked Language Modelling (MLM) and Next Sentence Prediction (NSP) to train the context dependent representation.” Here, WATERWORTH describes using sentence prediction using the BERT transformer to evaluate metadata, showing a sentence transformer. See WATERWORTH in page 6, section 3.3 Pretrained language model for more details.
PNG
media_image4.png
207
565
media_image4.png
Greyscale
See WATERWORTH illustrate in “Fig. 1. The point metadata inference pipeline. Named Entity Recognition is used to extract text spans from the input text. Span labels are often provided by and active learning process. The contextual data in the spans (entities and/or relationships) is extracted from the spans and then linked to a well known schema.” See WATERWORTH in table 1 illustrate the metadata inference on different asset devices.
PNG
media_image5.png
866
886
media_image5.png
Greyscale
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the references of LI, MISHRA, and KOH and incorporate with the teachings of WATERWORTH by using the teachings from LI, MISHRA, and KOH of classifying assets in a facility by using similarity and anchor-based learning, with WATERWORTH’s teaching of using a sentence transformer for transforming the item point metadata.
One of ordinary skill in the art would be motivated to do so because by integrating WATERWORTH’s framework into the methods of LI, MISHRA, and KOH, one with ordinary skill in the art would achieve the goal of providing a method with the “ deployment of smart building applications would be more cost effective” (see WATERWORTH in page 1, section Introduction, 5th paragraph).
However, Li in view of MISHRA, further in view of KOH, and further in view of WATERWORTH, did not teach “and a time series transformer for transforming the time series data,”
In an analogous art, DENG teaches “and a time series transformer for transforming the time series data,”
See DENG in section B. Transformer Encoder, page 21015, paragraphs 1-2 describe “The transformer is a sequence-to-sequence deep neural network model with an encoder–decoder structure proposed by [56]. The transformer is first intended for machine translation and later also obtains significant success in other areas, such as computer vision. For HVAC systems, a transformer can be an efficient model for mining the latent features of the time-serial context data. Knowing the context evolution is key to the continuous control problem for the building thermal environments." Here, DENG shows a transformer for transforming time-serial (i.e. time series) context data for HVAC systems.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the references of LI, MISHRA, KOH, and WATERWORTH, and incorporate with the teachings of DENG by using the teachings from LI, MISHRA, KOH, WATERWORTH of classifying assets in a facility by using similarity and anchor-based learning, with DENG’s teaching of using a time-series transformer for transforming the time series data.
One of ordinary skill in the art would be motivated to do so because by integrating DENG’s framework into the methods of LI, MISHRA, KOH, and WATERWORTH one with ordinary skill in the art would achieve “for HVAC systems, a transformer can be an efficient model for mining the latent features of the time-serial context data,” (see DENG in page 21015, section B.), with “s. Encoder layers provide efficient contextual learning,” (see DENG in page 21016, section Context-Aware Network), and “DRL-based HVAC control algorithms are efficient in saving energy and maintaining thermal comfort,” (see DENG in page 21012, section II. Background and related work, subsection A. Reinforcement Learning).
Claim 3:
Regarding claim 3, LI in view of MISHRA, further in view of KOH, further in view of WATERWORTH, and further in view of DENG, teach the limitations in claim 2.
Further, LI teaches “The method of claim 2, wherein the neural network further comprises: a feedforward neural network configured for receiving output from the sentence transformer and the time series transformer;”
See LI in page 4684, section Related work, describe “there are two different approaches to acquiring relationships between sensors: parsing the sensor metadata (i.e., point names as shown in Fig. 1) and inferring from the sensor time series readings.” Here, LI shows that this method covers evaluating sensor metadata by point names (i.e. item point metadata), and sensor time series readings (i.e. time series data associated with an asset). See LI in page 4684, Introduction for details.
Further, see LI in page 4686, section Deep metric learning triplet network, describes “The triplet network is comprised of three identical feed forward networks with shared parameters. In each iteration, a mini-batch of training triplets consisting of an anchor sensor Xa, accompanied with a pair of positive sensor Xp (i.e., a sensor in functional/spatial relation) and negative sensor Xn (i.e., a sensor not in functional/spatial relation) are fed into the triplet network ...when fed with a triplet, the network outputs the corresponding embedding vectors ya, yp, and yn.” Here, LI describes using a feed forward network that when inputting a triplet, the neural network outputs embedding vectors for sensors in function/spatial relation that relates to text data from metadata, and time-series data. See figure 1 in LI for details. LI in figure 1 also shows the context information can be portrayed as a text label.
PNG
media_image6.png
264
806
media_image6.png
Greyscale
Further, WATERWROTH teaches “the method of claim 2, wherein the neural network further comprises: a feedforward neural network configured for receiving output from the sentence transformer…”
See WATERWORTH in section 1.1 Background, page 2 describes “The objective of this study is to answer the research question if the use of transformers [23], and pretrained language models such as Bidirectional Encoder Representations from Transformers (BERT) [24] can be used for building metadata inference. Because of the lack of suitable preprocessing tools [25], these models and methods have not, to the best of our knowledge, been evaluated for building metadata inference.”
Further, see WATERWORTH in page 3, section 1.1 describes the model “BERT uses Masked Language Modelling (MLM) and Next Sentence Prediction (NSP) to train the context dependent representation.” Here, WATERWORTH describes using sentence prediction using the BERT transformer to evaluate metadata, showing a sentence transformer. See WATERWORTH in page 6, section 3.3 Pretrained language model for more details.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the references of LI, MISHRA, and KOH and incorporate with the teachings of WATERWORTH by using the teachings from LI, MISHRA, and KOH of classifying assets in a facility by using similarity and anchor-based learning, with WATERWORTH’s teaching of using a sentence transformer.
One of ordinary skill in the art would be motivated to do so because by integrating WATERWORTH’s framework into the methods of LI, MISHRA, and KOH, one with ordinary skill in the art would achieve the goal of providing a method with the “ deployment of smart building applications would be more cost effective” (see WATERWORTH in page 1, section Introduction, 5th paragraph).
Further, DENG teaches “for receiving output from the sentence transformer and the time series transformer … and a layer normalization module configured for receiving output from the feedforward neural network.”
See DENG in page 21015, section B. Transformer Encoder, paragraphs 1-2, describe
“The transformer is a sequence-to-sequence deep neural network model with an encoder–decoder structure proposed by [56]. The transformer is first intended for machine translation and later also obtains significant success in other areas, such as computer vision. For HVAC systems, a transformer can be an efficient model for mining the latent features of the time-serial context data. Knowing the context evolution is key to the continuous control problem for the building thermal environments.” Here, DENG shows a transformer is involved in mining time-series data.
The encoder–decoder structure of the transformer is shown in Fig. 2. The encoder is stacked by a group of identical encoder blocks. Each block consists of a self-attention module, a feedforward network (FFN), and residual connection and normalization modules. FFN is composed of 2-layer position wise fully connected neural networks. The residual connection and normalization module is used for the output of the self-attention module and FFN in order to alleviate the gradient vanishing problems of a deep model.” Here, DENG shows using a feedforward network (i.e. feedforward neural network) and a normalization module (i.e. normalizing layer) where the normalization module is involved in being part of a deep learning model that includes a transformer for time series data.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the references of LI, MISHRA, KOH, and WATERWORTH, and incorporate with the teachings of DENG by using the teachings from LI, MISHRA, KOH, WATERWORTH of classifying assets in a facility by using similarity and anchor-based learning, with DENG’s teaching of using a time-series transformer for transforming the time series data.
One of ordinary skill in the art would be motivated to do so because by integrating DENG’s framework into the methods of LI, MISHRA, KOH, and WATERWORTH one with ordinary skill in the art would achieve “for HVAC systems, a transformer can be an efficient model for mining the latent features of the time-serial context data,” (see DENG in page 21015, section B.), with “s. Encoder layers provide efficient contextual learning,” (see DENG in page 21016, section Context-Aware Network), and “DRL-based HVAC control algorithms are efficient in saving energy and maintaining thermal comfort,” (see DENG in page 21012, section II. Background and related work, subsection A. Reinforcement Learning).
Claim 9:
Regarding claim 9, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 8. Regarding claim 9, the claim recites similar limitations as corresponding claim 2 and is rejected for similar reasons as claim 2 using similar teachings and rationale.
Claim 10:
Regarding claim 10, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 9. Regarding claim 10, the claim recites similar limitations as corresponding claim 3 and is rejected for similar reasons as claim 3 using similar teachings and rationale.
Claim 16:
Regarding claim 16, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 15. Regarding claim 16, the claim recites similar limitations as corresponding claim 2 and is rejected for similar reasons as claim 2 using similar teachings and rationale.
Claim 17:
Regarding claim 17, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 16. Regarding claim 17, the claim recites similar limitations as corresponding claim 3 and is rejected for similar reasons as claim 3 using similar teachings and rationale.
Claims 4, 11, and 18 are rejected under 35 U.S.C. 103 over LI in view of MISHRA, further in view of KOH, further in view of Tariq, S. et al, in “Deep-AI soft sensor for sustainable health risk monitoring and control of fine particulate matter at sensor devoid underground spaces: A zero-shot transfer learning approach,” available online on November 8, 2022 and published in January 2023, available at https://www.sciencedirect.com/science/article/pii/S0886779822004837 , (hereafter, TARIQ).
Claim 4:
Regarding claim 4, LI in view of MISHRA, further in view of KOH, teach the limitations in claim 1.
However, LI in view of MISHRA, further in view of KOH, did not teach “The method of claim 1, wherein the machine learning neural network comprises zero shot learning.”
In an analogous art, TARIQ teaches “The method of claim 1, wherein the machine learning neural network comprises zero shot learning,”
See TARIQ in pages 9-10, section 3.6 describes “Additionally, semantic information available in both source and target domain is selected as input for the neural network model. For early warning of platform PM2.5 CIAI levels, the dataset is then processed into sequences using a rolling window with six hours of four variables as inputs and one hour ahead PM2.5 values as output.”
See TARIQ in page 10, section 3.6, paragraph 2, where TARIQ describes “ several machine learning models are employed to capture temporal dynamics of platform PM2.5 in the form of an AI soft sensor for source domain metro stations. These models include LSTM, GRU, CNN, CNN-LSTM, and the proposed attention-aware bidirectional GRU network.”
Here, TARIQ shows using a neural network as a model to evaluate semantic information in soft sensors to evaluate air quality in PM 2.5. See TARIQ in figure 7 for details.
PNG
media_image7.png
774
1246
media_image7.png
Greyscale
Further, see TARIQ in page 6, section 3.2 Zero-shot transfer learning describe “The objective of our study is to develop a zero-shot model that can identify target variables Yt utilizing the predictor variables available in the target domain Xt by the knowledge extracted from X,Z, and y, because the data for target variable in target domain denoted as Yt=[yt1,yt2..,ytN]T∈RN×Dv is not available for training.” Here, TARIQ mentions using information to input into a neural network from pages 9-10, including using zero-shot model, which includes zero-shot learning methods, that can identify target variables cited in page 6.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the references of LI, MISHRA, and KOH and incorporate with the teachings of TARIQ by using the teachings from LI, MISHRA, and KOH of classifying assets in a facility by using similarity and anchor-based learning, with TARIQ’s teaching of using zero shot learning for a neural network.
One of ordinary skill in the art would be motivated to do so because by integrating TARIQ’s framework into the methods of LI, MISHRA, and KOH, one with ordinary skill in the art would achieve results that “ demonstrate the effectiveness of the proposed frame work, where the similarity function helps properly capture the PM 2.5 dynamics at different metro stations. The ZSTL methodology represents a reliable and efficient way to forecast future PM 2.5 concentrations to provide an early health risk warning for metro stations that lack the necessary monitoring equipment to identify PM 2.5 levels” (see TARIQ, page 15, section 4.2 AI-soft sensor transfer on target domain stations).
Claim 11:
Regarding claim 11, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 8. Regarding claim 11, the claim recites similar limitations as corresponding claim 4 and is rejected for similar reasons as claim 4 using similar teachings and rationale.
Claim 18:
Regarding claim 18, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 15. Regarding claim 18, the claim recites similar limitations as corresponding claim 4 and is rejected for similar reasons as claim 4 using similar teachings and rationale.
Claims 5 and 12 are rejected under 35 U.S.C. 103 over LI in view of MISHRA, further in view of KOH, and further in view of WATERWORTH.
Claim 5:
Regarding claim 5, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 1.
However, LI in view of MISHRA, further in view of KOH did not teach “The method of claim 1, wherein the predetermined threshold at least about 30%.”
In an analogous field, WATERWORTH teaches “The method of claim 1, wherein the predetermined threshold at least about 30%.”
See WATERWORTH section 3.4, Tagset classifier, page 6… where WATERWORTH describes "There may be cases where the model produces a low quality prediction where there is no single peak in the distribution. By placing a threshold 𝜖 on the probabilities in 4 we can tune the model to return the empty tagset in this case. This will trade off precision for recall."
Further, see WATERWORTH in figure 6 caption in page 8 note "Fig. 6. Accuracy by building. The shorter (blue) bars show the base accuracy, the longer bars (blue plus yellow) show the accuracy when a threshold of 0.8 is applied to P (yi|x) from Eq. 4." Here, WATERWORTH shows the threshold is set as 𝜖, and the threshold of 0.8 is applied as a value for the threshold for the predictions in figure 6, and 0.8 is equal to 80% for this threshold, and 80% is greater than 30% or at least about 30%.
PNG
media_image8.png
474
1306
media_image8.png
Greyscale
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the references of LI, MISHRA, and KOH and incorporate with the teachings of WATERWORTH by using the teachings from LI, MISHRA, and KOH of classifying assets in a facility by using similarity and anchor-based learning, with WATERWORTH’s teaching of a predetermined threshold be about at least 30%.
One of ordinary skill in the art would be motivated to do so because by integrating WATERWORTH’s framework into the methods of LI, MISHRA, and KOH, one with ordinary skill in the art would achieve the goal of providing a method with the “ deployment of smart building applications would be more cost effective” (see WATERWORTH in page 1, section Introduction, 5th paragraph).
Claim 12:
Regarding claim 12, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 8. Regarding claim 12, the claim recites similar limitations as corresponding claim 5 and is rejected for similar reasons as claim 5 using similar teachings and rationale.
Claims 7, 14, and 20 are rejected under 35 U.S.C. 103 over LI in view of MISHRA, further in view of KOH, and further in view of Vitullo, S., (US PG Pub. No. US20180113482A1), published on April 26, 2018, (hereafter, VITULLO).
Claim 7:
Regarding claim 7, LI in view of MISHRA, further in view of KOH, teach the limitations in claim 1.
However, LI in view of MISHRA, further in view of KOH, did not teach “The method of claim 1, further comprising: generating, by at least one processor, a model of the assets in the facility based on the determination of asset type.”
In an analogous art, VITULLO teaches “The method of claim 1, further comprising: generating, by at least one processor, a model of the assets in the facility based on the determination of asset type.”
See VITULLO discuss in [0054] “For example, central plant 200 may include heaters, chillers, heat recovery chillers, cooling towers, or other types of equipment configured to serve the heating and/or cooling loads of a building or campus” . Here, VITULLO describes that the equipment or assets can be used to serve a building,(i.e. a facility).
See VITULLO in HVAC Component Models, paragraph [0112] describe “Still referring to FIG. 5, predictive modeling system 402 is shown to include several HVAC component models 510. HVAC component models 510 can include any of a variety of predictive models configured to predict the performance of a HVAC component based on a set of input variables (e.g., monitored variables and/or operating points). In some embodiments, each of HVAC component models 510 corresponds to a particular HVAC device, a particular type of HVAC device (e.g., a chiller, a boiler, an actuator, etc.) or a particular model of HVAC device (e.g., chiller model A, chiller model B, etc.) and can be used to predict the performance of the corresponding HVAC device, device type, or model. For example, HVAC component models 510 can include power consumption models that can be used to predict the power consumption of various HVAC devices. In some embodiments, HVAC component models 510 define power consumption as a function of equipment load and/or equipment setpoints. For example, the HVAC component model for a chiller may define the power consumption of the chiller as a function of cold water production and/or chiller load setpoints.” Here, VITULLO mentions that after identifying a HVAC device as a particular type (such as a chiller, boiler, actuator), then models are generated that corresponds to a particular HVAC device, for example, there is a model that just corresponds to chillers, and VITULLO relates to generating a model of assets in the facility based on determining asset type.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the references of LI, MISHRA, and KOH and incorporate with the teachings of VITULLO by using the teachings from LI, MISHRA, and KOH of classifying assets in a facility by using similarity and anchor-based learning, with VITULLO’s teaching of generating a model of the assets in the facility from determining asset type.
One of ordinary skill in the art would be motivated to do so because by integrating VITULLO’s framework into the methods of LI, MISHRA, and KOH, one with ordinary skill in the art would achieve a method "Advantageously, the variance weighting technique can reduce the combined estimate error variance relative to the equal weighting technique by assigning greater weights to more accurate or more certain component model predictions and lesser weights to less accurate or less certain component model predictions" (see VITULLO in [0182]).
Claim 14:
Regarding claim 14, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 8. Regarding claim 14, the claim recites similar limitations as corresponding claim 7 and is rejected for similar reasons as claim 7 using similar teachings and rationale.
Claim 20:
Regarding claim 20, LI in view of MISHRA, further in view of KOH teaches the limitations in claim 15. Regarding claim 20, the claim recites similar limitations as corresponding claim 7 and is rejected for similar reasons as claim 7 using similar teachings and rationale.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to WENWEI ZENG whose telephone number is (571)272-7111. The examiner can normally be reached Monday-Friday, 8am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Usmaan Saeed can be reached at (571) 272-4046. 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.
/WenWei Zeng/Examiner, Art Unit 2146
/USMAAN SAEED/Supervisory Patent Examiner, Art Unit 2146
Prosecution Insights