DETAILED ACTION
This Office action is responsive to the Request for Continued Examination (RCE) filed under 37 CFR §1.53(d) for the instant application on November 6, 2025. The Applicants have properly set forth the RCE, which has been entered into the application, and an examination on the merits follows herewith.
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
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, 6-8, 13-15, and 20-30 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea (i.e. a mental process) without significantly more. As described in MPEP § 2106, the analyses as to whether a claim qualifies as eligible subject matter under 35 U.S.C. § 101 includes the following determinations:
(1) Whether the claim is to a statutory category, i.e. to a process, machine, manufacture or composition of matter (“Step 1”) – see MPEP §§ 2106, subsection III, and 2106.03
(2) If the claim is to a statutory category, whether the claim recites any judicial exceptions, including certain groupings of abstract ideas (i.e., mathematical concepts, certain methods of organizing human activity, or mental processes) (“Step 2A, Prong One”) – see MPEP §§ 2106, subsection III, and 2106.04
(3) If the claim recites a judicial exception, whether the claim recites additional elements that integrate the judicial exception into a practical application (“Step 2A, Prong Two”) – see MPEP §§ 2106, subsection III, and 2106.04
(4) If the claim does not recite additional elements that integrate the judicial exception into a practical application, whether the claim recites additional elements that amount to significantly more than the judicial exception (“Step 2B”) – see MPEP §§ 2106, subsection III, and 2106.05
Claim 1
Regarding “Step 1,” independent claim 1 is to a statutory category as claim 1 is directed to a method, i.e. a process.
Accordingly, the analysis proceeds to “Step 2A, Prong One” to determine if the claim recites a judicial exception. In this case, claim 1 recites mathematical concepts and mental processes and thus recites a judicial exception. In particular, “determining residuals…via subtracting a time averaged mean from a measurement…” is considered a recitation of a mathematical concept. The following are considered recitations of mental processes: “wherein in response to one of more of the residuals exceeding a zero value and exceeding a threshold value an action is performed;” “generating a forecast of one or more future conditions for the physical environment…using one or more forecasting…models, the one or more forecasting…models incorporating depth gradients of the multiple sensors and the residuals to generate the forecast;” and “…classifying the forecast as an anomalous forecast….”
Further regarding the claimed mental processes, it is noted that “’the mental processes’ abstract idea grouping in particular is defined as concepts performed in the human mind, and examples of mental processes include observations, evaluations, judgements, and opinions.” MPEP § 2106.04(a)(2), subsection III. “If a claim recites a limitation that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper, the limitation falls within the mental processes grouping, and the claims recites an abstract idea. MPEP § 2106.04(a)(2), subsection III,B (citations omitted). In this case, the limitations reciting “wherein in response to one of more of the residuals exceeding a zero value and exceeding a threshold value an action is performed” and “…classifying the forecast as an anomalous forecast…” are indicative of judgements that can practically be performed in the human mind. The generation of a forecast of one or more future conditions using one or more forecasting models that incorporate sensor depth gradients and residuals, when given its broadest reasonable interpretation, can be considered an evaluation that can practically be performed in the human mind.
Because claim 1 recites a judicial exception (i.e. a mathematical concept, mental processes), the analysis proceeds to “Step2A, Prong Two.” But here the claim does not recite additional elements that integrate the judicial exception into a practical application. In particular, in addition to the above-noted mental processes and mathematical concept, claim 1 recites that the forecast of one or more future conditions is generated “based on the graph neural network using one or more forecasting machine learning models.” Claim 1 also recites, “incorporating, into a graph neural network, data received from a group of multiple sensors distributed in a physical environment, the data relating to physical and environmental conditions of the physical environment” and that the residuals are determined “via the multiple sensors.” In addition, claims 1 recites “presenting an alert indicating the anomalous forecast” in response to classifying the forecast as an anomalous forecast. However, “incorporating….data received from a group of multiple sensors distributed in a physical environment, the data relating to physical and environmental conditions of the physical environment,” wherein the residuals are determined “via the multiple sensors,” and “presenting an alert indicating the anomalous forecast” are indicative of insignificant extra-solution activity, i.e. mere data gathering, and is insufficient to integrate the abstract idea into a practical application. See MPEP § 2106.5(g). That this data is incorporated into a graph neural network, and that the forecast of one or more future conditions is generated “based on the graph neural network using one or more forecasting machine learning models” represent no more than mere instructions to apply the judicial exception on a computer, and thus also do not integrate the judicial exception into a practical application. See MPEP § 2106.05(f). Particularly, the claim omits any detail as to how the graph neural network is involved in generating the forecast of one or more future conditions; the graph neural network appears to be invoked merely as a tool for performing the judicial exception. As such, the graph neural network is tantamount to a generic computer.
Accordingly, as claim 1 does not recite additional elements that integrate the judicial exception into a practical application, the analysis proceeds to “Step 2B” to determine whether the claims recite additional elements that amount to significantly more than the judicial exception. However, in this case, the claim does not. As noted above, in addition to the above-noted mental processes and mathematical concept, claim 1 recites “incorporating….data received from a group of multiple sensors distributed in a physical environment, the data relating to physical and environmental conditions of the physical environment,” wherein the residuals are determined “via the multiple sensors,” and “presenting an alert indicating the anomalous forecast.” As further noted above, these limitations are indicative of insignificant extra-solution activity, i.e. mere data gathering. See MPEP § 2106.5(g). Such data gathering is also well-understood, routine and conventional. See, e.g., Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014). That the forecast of one or more future conditions is generated “based on the graph neural network using one or more forecasting machine learning models” represents mere instructions to apply the abstract idea on a generic computer, as is described above, and thus does not amount to significantly more than the judicial exception.
Consequently, claim 1 recites an abstract idea but does not include additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea. As a result, and for the reasons described above, claim 1 is rejected as being patent ineligible under 35 U.S.C. § 101.
Claim 6
Claim 6 recites “wherein the one or more forecasting machine learning models comprise an edge-based prediction model that generates a model representation of the physical and environmental conditions.” However, when given its broadest reasonable interpretation, “wherein the one or more forecasting…models comprise [a]…prediction model that generates a model representation of the physical and environmental conditions” is considered a characteristic of a mental process, i.e. an observation or evaluation that can practically be performed in the human mind. That the prediction model is “edge-based” represents no more than mere instructions to apply the judicial exception on a computer, and therefore does not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f). Accordingly, claim 6 fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 6 is also patent ineligible under 35 U.S.C. § 101.
Claim 7
Claim 7 recites “determining one or more anomalies from current physical and environmental conditions based on the model representation of the physical and environmental conditions.” However, when given its broadest, reasonable interpretation, “determining one or more anomalies from current physical and environmental conditions based on the model representation of physical and environmental conditions” is considered a mental process, i.e. an observation, evaluation or judgment that can practically be performed in the human mind. Accordingly, claim 7 fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 7 is also patent ineligible under 35 U.S.C. § 101.
Claim 8
Regarding “Step 1,” independent claim 8 is to a statutory category as claim 8 is directed to a system, which can be considered a machine or manufacture.
Accordingly, the analysis proceeds to “Step 2A, Prong One” to determine if the claim recites a judicial exception. In this case, claim 8 recites a mathematical concept and mental processes and thus recites a judicial exception. In particular, similar to claim 1 described above, the recitation in claim 8 of “determine residuals…via subtracting a time averaged mean from a measurement…” is indicative of a mathematical concept. Also similar to claim 1, the following limitations in claim 8 are considered recitations of mental processes: “wherein in response to one of more of the residuals exceeding a zero value and exceeding a threshold value an action is performed;” “generate a forecast of one or more future conditions for the physical environment…using one or more forecasting…models, the one or more forecasting…models incorporating depth gradients of the multiple sensors and the residuals to generate the forecast;” and “…classifying the forecast as an anomalous forecast….”
Because claim 8 recites a judicial exception (i.e. a mathematical concept, mental processes), the analysis proceeds to “Step2A, Prong Two.” But here the claim does not recite additional elements that integrate the judicial exception into a practical application. In particular, in addition to the above-noted mental processes and mathematical concept, claim 8 recites that the forecast of one or more future conditions is generated “based on the graph neural network using one or more forecasting machine learning models.” Claim 8 also recites, “incorporate, into a graph neural network, data received from a group of multiple sensors distributed in a physical environment, the data relating to physical and environmental conditions of the physical environment” and that the residuals are determined “via the multiple sensors.” In addition, claims 8 recites “present an alert indicating the anomalous forecast” in response to classifying the forecast as an anomalous forecast. However, “incorporate[ing]….data received from a group of multiple sensors distributed in a physical environment, the data relating to physical and environmental conditions of the physical environment,” wherein the residuals are determined “via the multiple sensors,” and “present[ing] an alert indicating the anomalous forecast” are indicative of insignificant extra-solution activity, i.e. mere data gathering, and is insufficient to integrate the abstract idea into a practical application. See MPEP § 2106.5(g). That this data is incorporated into a graph neural network, and that the forecast of one or more future conditions is generated “based on the graph neural network using one or more forecasting machine learning models” represent no more than mere instructions to apply the judicial exception on a computer, and thus also do not integrate the judicial exception into a practical application. See MPEP § 2106.05(f). Particularly, the claim omits any detail as to how the graph neural network is involved in generating the forecast of one or more future conditions; the graph neural network appears to be invoked merely as a tool for performing the judicial exception. As such, the graph neural network is tantamount to a generic computer.
Accordingly, as claim 8 does not recite additional elements that integrate the judicial exception into a practical application, the analysis proceeds to “Step 2B” to determine whether the claims recite additional elements that amount to significantly more than the judicial exception. However, in this case, the claim does not. As noted above, in addition to the above-noted mental processes and mathematical concept, claim 8 recites “incorporate….data received from a group of multiple sensors distributed in a physical environment, the data relating to physical and environmental conditions of the physical environment,” wherein the residuals are determined “via the multiple sensors,” and “present an alert indicating the anomalous forecast.” As further noted above, these limitations are indicative of insignificant extra-solution activity, i.e. mere data gathering. See MPEP § 2106.5(g). Such data gathering is also well-understood, routine and conventional. See, e.g., Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014). That the forecast of one or more future conditions is generated “based on the graph neural network using one or more forecasting machine learning models” represents mere instructions to apply the abstract idea on a generic computer, as is described above, and thus does not amount to significantly more than the judicial exception.
Consequently, claim 8 recites an abstract idea but does not include additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea. As a result, and for the reasons described above, claim 8 is rejected as being patent ineligible under 35 U.S.C. § 101.
Claim 13
Claim 13 recites “wherein the one or more forecasting machine learning models comprise an edge-based prediction model that generates a model representation of the physical and environmental conditions.” However, when given its broadest reasonable interpretation, “wherein the one or more forecasting…models comprise [a]…prediction model that generates a model representation of the physical and environmental conditions” is considered a characteristic of a mental process, i.e. an observation or evaluation that can practically be performed in the human mind. That the prediction model is “edge-based” represents no more than mere instructions to apply the judicial exception on a computer, and therefore does not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f). Accordingly, claim 13 fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 13 is also patent ineligible under 35 U.S.C. § 101.
Claim 14
Claim 14 recites “determine one or more anomalies from current physical and environmental conditions based on the model representation of the physical and environmental conditions.” However, when given its broadest, reasonable interpretation, “determine one or more anomalies from current physical and environmental conditions based on the model representation of physical and environmental conditions” is considered a mental process, i.e. an observation, evaluation or judgment that can practically be performed in the human mind. Accordingly, claim 14 fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 14 is also patent ineligible under 35 U.S.C. § 101.
Claim 15
Regarding “Step 1,” independent claim 15 is to a statutory category as claim 15 is directed to a computer program product comprising one or more computer readable storage media, which can be considered a manufacture.
Accordingly, the analysis proceeds to “Step 2A, Prong One” to determine if the claim recites a judicial exception. In this case, claim 15 recites a mathematical concept and mental processes and thus recites a judicial exception. In particular, similar to claim 1 described above, the recitation in claim 15 of “determine residuals…via subtracting a time averaged mean from a measurement…” is indicative of a mathematical concept. Also similar to claim 1, the following limitations in claim 15 are considered recitations of mental processes: “wherein in response to one of more of the residuals exceeding a zero value and exceeding a threshold value an action is performed;” “generate a forecast of one or more future conditions for the physical environment…using one or more forecasting…models, the one or more forecasting…models incorporating depth gradients of the multiple sensors and the residuals to generate the forecast;” and “…classifying the forecast as an anomalous forecast….”
Because claim 15 recites a judicial exception (i.e. a mathematical concept, mental processes), the analysis proceeds to “Step2A, Prong Two.” But here the claim does not recite additional elements that integrate the judicial exception into a practical application. In particular, in addition to the above-noted mental processes and mathematical concept, claim 15 recites that the forecast of one or more future conditions is generated “based on the graph neural network using one or more forecasting machine learning models.” Claim 15 also recites, “incorporate, into a graph neural network, data received from a group of multiple sensors distributed in a physical environment, the data relating to physical and environmental conditions of the physical environment” and that the residuals are determined “via the multiple sensors.” In addition, claims 15 recites “present an alert indicating the anomalous forecast” in response to classifying the forecast as an anomalous forecast. However, “incorporate[ing]….data received from a group of multiple sensors distributed in a physical environment, the data relating to physical and environmental conditions of the physical environment,” wherein the residuals are determined “via the multiple sensors,” and “present[ing] an alert indicating the anomalous forecast” are indicative of insignificant extra-solution activity, i.e. mere data gathering, and is insufficient to integrate the abstract idea into a practical application. See MPEP § 2106.5(g). That this data is incorporated into a graph neural network, and that the forecast of one or more future conditions is generated “based on the graph neural network using one or more forecasting machine learning models” represent no more than mere instructions to apply the judicial exception on a computer, and thus also do not integrate the judicial exception into a practical application. See MPEP § 2106.05(f). Particularly, the claim omits any detail as to how the graph neural network is involved in generating the forecast of one or more future conditions; the graph neural network appears to be invoked merely as a tool for performing the judicial exception. As such, the graph neural network is tantamount to a generic computer. Claim 15 also recites that the computer program product comprises one or more computer readable storage media with program instructions for performing the tasks recited in claim 15. Again, however, the one or more computer readable storage media represent mere instructions to apply the judicial exception on a computer, and thus also do not integrate the judicial exception into a practical application.
Accordingly, as claim 15 does not recite additional elements that integrate the judicial exception into a practical application, the analysis proceeds to “Step 2B” to determine whether the claims recite additional elements that amount to significantly more than the judicial exception. However, in this case, the claim does not. As noted above, in addition to the above-noted mental processes and mathematical concept, claim 15 recites “incorporate….data received from a group of multiple sensors distributed in a physical environment, the data relating to physical and environmental conditions of the physical environment,” wherein the residuals are determined “via the multiple sensors,” and “present an alert indicating the anomalous forecast.” As further noted above, these limitations are indicative of insignificant extra-solution activity, i.e. mere data gathering. See MPEP § 2106.5(g). Such data gathering is also well-understood, routine and conventional. See, e.g., Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014). That the forecast of one or more future conditions is generated “based on the graph neural network using one or more forecasting machine learning models” represents mere instructions to apply the abstract idea on a generic computer, as is described above, and thus does not amount to significantly more than the judicial exception. Similarly, the recitation in claim 15 of “one or more computer readable storage media” comprising program instructions for performing the noted tasks in claim 15 also represents mere instructions to apply the abstract idea on a generic computer, and thus does not amount to significantly more than the judicial exception.
Consequently, claim 15 recites an abstract idea but does not include additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea. As a result, and for the reasons described above, claim 15 is rejected as being patent ineligible under 35 U.S.C. § 101.
Claim 20
Claim 20 recites “wherein the one or more forecasting machine learning models comprise an edge-based prediction model that generates a model representation of the physical and environmental conditions.” However, when given its broadest reasonable interpretation, “wherein the one or more forecasting…models comprise [a]…prediction model that generates a model representation of the physical and environmental conditions” is considered a characteristic of a mental process, i.e. an observation or evaluation that can practically be performed in the human mind. That the prediction model is “edge-based” represents no more than mere instructions to apply the judicial exception on a computer, and therefore does not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f). Accordingly, claim 20 fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 20 is also patent ineligible under 35 U.S.C. § 101.
Claim 21
Claim 21 recites “determine one or more anomalies from current physical and environmental conditions based on the model representation of the physical and environmental conditions.” However, when given its broadest, reasonable interpretation, “determine one or more anomalies from current physical and environmental conditions based on the model representation of physical and environmental conditions” is considered a mental process, i.e. an observation, evaluation or judgment that can practically be performed in the human mind. Accordingly, claim 21 fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 21 is also patent ineligible under 35 U.S.C. § 101.
Claim 22
Claim 22 recites that “the physical environment comprises an aquaculture farm to be affected via the anomalous forecast.” However, this limitation does no more than generally link the judicial exception (i.e. mental processes, mathematical concept) noted above in independent claim 1 to a field of use or technological environment, and thus fails to integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(h).
Claim 23
Claim 23 recites that “the aquaculture farm comprises one or more cages to be affected via the anomalous forecast.” However, this limitation does no more than generally link the judicial exception (i.e. mental processes, mathematical concept) noted above in independent claim 1 to a field of use or technological environment, and thus fails to integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(h).
Claim 24
Claim 24 recites that “the one or more future conditions comprise toxic algae concentration for a body of water.” However, this is considered a characteristic of the mental process recited in independent claim 1 (i.e. “generating a forecast of one or more conditions…”). As such, claim 24 fails to recite additional elements (i.e. elements other than the mental processes, mathematical concepts) that would integrate the judicial exception into a practical application or amount to significantly more than the judicial exception.
Claim 25
Claim 25 recites that “the physical environment comprises an aquaculture farm to be affected via the anomalous forecast.” However, this limitation does no more than generally link the judicial exception (i.e. mental processes, mathematical concept) noted above in independent claim 8 to a field of use or technological environment, and thus fails to integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(h).
Claim 26
Claim 26 recites that “the aquaculture farm comprises one or more cages to be affected via the anomalous forecast.” However, this limitation does no more than generally link the judicial exception (i.e. mental processes, mathematical concept) noted above in independent claim 8 to a field of use or technological environment, and thus fails to integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(h).
Claim 27
Claim 27 recites that “the one or more future conditions comprise toxic algae concentration for a body of water.” However, this is considered a characteristic of the mental process recited in independent claim 8 (i.e. “generate a forecast of one or more conditions…”). As such, claim 27 fails to recite additional elements (i.e. elements other than the mental processes, mathematical concepts) that would integrate the judicial exception into a practical application or amount to significantly more than the judicial exception.
Claim 28
Claim 28 recites that “the physical environment comprises an aquaculture farm to be affected via the anomalous forecast.” However, this limitation does no more than generally link the judicial exception (i.e. mental processes, mathematical concept) noted above in independent claim 15 to a field of use or technological environment, and thus fails to integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(h).
Claim 29
Claim 29 recites that “the aquaculture farm comprises one or more cages to be affected via the anomalous forecast.” However, this limitation does no more than generally link the judicial exception (i.e. mental processes, mathematical concept) noted above in independent claim 15 to a field of use or technological environment, and thus fails to integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(h).
Claim 30
Claim 30 recites that “the one or more future conditions comprise toxic algae concentration for a body of water.” However, this is considered a characteristic of the mental process recited in independent claim 15 (i.e. “generating a forecast of one or more conditions…”). As such, claim 30 fails to recite additional elements (i.e. elements other than the mental processes, mathematical concepts) that would integrate the judicial exception into a practical application or amount to significantly more than the judicial exception.
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1, 8 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over the article entitled, “Time-Series Graph Network for Sea Surface Temperature Prediction” by Sun et al. (“Sun”), over WIPO Publication No. WO 2021/044192 A1 to Soualhia et al. (“Soualhia”), over the article entitled, “Prediction of 3-D Ocean Temperature by Multilayer Convolutional LSTM” by Zhang et al. (“Zhang”), and also over the article entitled, “Sea Surface Temperature and High Water Temperature Occurrence Prediction Using a Long Short-Term Memory Model” by Kim et al. (“Kim”).
Regarding claims 1, 8 and 15, Sun generally presents “a time-series graph network (TSGN) that can jointly capture graph-based spatial correlation and temporal dynamics.” (Abstract). Like claimed, Sun particularly teaches:
incorporating, into a graph neural network, data received from a group of multiple sensors distributed in a physical environment, the data relating to physical and environmental conditions of the physical environment (Sun generally teaches applying the TSGN to the task of sea surface temperature prediction:
The sea surface temperature (SST) [1] is a physical quantity reflecting the thermal conditions of seawater and a key parameter in the field of marine research. The various effects of the sea-water temperature field largely affect the distribution of fishery resources, the direction of marine pollution, the development of oil and gas resources, and military activities. In addition, global climate anomalies are inseparably related to changes in the sea-water surface temperature. Therefore, it is important to make accurate and efficient predictions of seawater surface temperatures.
(Section 1 “Introduction”; emphasis added).
For the complex dependencies of sea surface temperature data in the time and space dimensions, we propose a graph neural network called a time-series graph network (TSGN) by combining the advantages of a long short-term memory (LSTM) network in processing temporal information. The model is based on the graph structure of the sensor network and uses the topology between sensors to model the geometric structure of the sensors. LSTM aggregation is adopted for SST data containing temporal information. This model combines graph convolution and gated time convolution to extract the most useful spatial features and coherently capture the most basic temporal features.
The main contributions of this paper are as follows:
(1) A graph learning network, i.e., a TSGN, is designed, which has strong learning ability for data prediction tasks containing time series information.
(2) A TSGN employs an LSTM to aggregate the node features to better capture the time correlation and improve the prediction accuracy.
(3) A TSGN is applied to an SST prediction task, which is a regression fitting, and the experimental results show good prediction results.
(Section 1 “Introduction”; emphasis added).
In particular, to predict sea surface temperatures, Sun teaches inputting historical sea surface temperature data into the TSGN, i.e. a graph neural network, whereby the TSGN outputs the predicted sea surface temperature for a future period:
As shown in Fig.1, the task of graph-based SST prediction can be defined as follows: build a graph neural network, input historical temperature data
X
∈
R
N
×
T
i
n
×
D
(where N is the number of sensors,
T
i
n
is the length of the time window, and D is the input feature dimension) and a graph G reflecting the spatial connectivity between temperature sensors, and output the temperature for a future period
Y
∈
R
N
×
T
o
u
t
×
D
(where
T
o
u
t
is the prediction step and d is the number of temperatures).
(Section 2.1 “Problem Definition”; emphasis added).
Sun further discloses that the historical temperature data is received from a sensor network comprising a plurality of sensors distributed in a physical environment:
To implement the sea surface temperature prediction function based on graph learning, we propose a prediction model framework, as shown in Fig. 2.
The key implementations in the framework are described as follows: constructing a graph based on the distribution of sensors on the sea surface and the covariance matrix calculated from the observed data, modeling the complex spatial dependencies of the sensors, and preprocessing the historically monitored data. The temperature data corresponding to each moment are used as the node feature input, and the creation of edges is based on the Pearson correlation coefficient between each sensor. Then, a TSGN, which is a graph neural network, was used to implement the information interaction between nodes, and all nodes were put in during sampling, feature aggregation was implemented using LSTM networks, and the final output prediction results were obtained. In the following section, we explain in detail the contents of the main modules in the framework.
(Section 3 “Framework overview”; emphasis added).
The sensor network was first modeled as an undirected graph. The distribution of nodes on the graph corresponds to the distribution of sensors within the sea area, and the edges indicate the relationship between the two sensors. Fig. 3 shows the method for constructing the undirected graph G. Here,
X
a
denotes sensor a;
e
a
,
b
denotes whether there is an edge between sensor and sensor b, that is,
e
a
,
b
=
1
indicates that a and b are connected, and there is an edge;
e
a
,
b
=
0
shows that a and b are not connected and there is no edge;
X
a
consists of the temperature data at the corresponding moment; and
e
a
,
b
is related to the Pearson correlation coefficient between two points.
(Section 3.1 “Undirected graph construction for sensor networks”; emphasis added).
The experimental data are the sea surface temperature data of the Northwest Pacific Ocean (range : 0 –
60
o
N
,
100
o
E
-
180
o
E
) from 2001 to 2005. There are four variables, and the variable names and data formats are as follows:
(1) Time: This indicates January 1, 2001, 0:00 to December 31, 2005, 18:00, where the time step is 6 h, with a total of 7304 moments, and text-type data, such as “20011010100” and “20011010106”;
(2) Latitude: This indicates the distribution of latitude 0
60
o
N
, with a step size of
0.5
o
, and is an array of 121 ×1;
(3) Longitude: The distribution of longitude
100
o
E
180
o
E
, with a step size of
0.5
o
, is an array of 161 ×1;
(4) sst01_05: This indicates the sea surface temperature, which is an array of 121×161×7304, corresponding to the latitude, longitude, and time. The sea surface temperature of the north-west Pacific Ocean at 0:00 on January 1, 2005, was selected as a sample map, as shown in Fig.7.
The actual data collection and storage process will encounter numerous problems, which may lead to missing and abnormal final data; thus, the obtained data should be pre-processed by the abnormal and missing values. In the above dataset, not all data belong to the observed area. We filtered out 13447 observation points belonging to the Northwest Pacific Ocean. Each observation point contained a sensor. The sensors and graph nodes have a one-to-one relation and therefore, there are 13447 nodes in the undirected graph of the sensor network.
(Section 4.1 “Dataset and data pre-processing”; emphasis added).
Sun also teaches that the TSGN is trained using such historical sea surface temperature data to predict future temperatures:
The prediction task in our experiments was defined as using 28 (
T
i
n
=
24
, sampling period of 6 h) historical temperature data of all nodes to predict the next 4 (
T
o
u
t
=
4
, sampling period of 6 h) temperature values, that is, the temperature values of the first 7 consecutive days are input to predict the temperature values of the next day. A sliding window with a window size of 32 was used to generate 7304 data samples with a step size of 1. The data are divided into a training set, an evaluation set, and a test set at a ratio of 7:1:2 in chronological order.
We used samples from the training set to optimize the model parameters. The performance of the model was evaluated by optimizing the hyperparameters of the execution results with the evaluation set and validating them with the test set.
(Section 4.2 “Experimental setup”; emphasis added).
Accordingly, Sun teaches incorporating, into a graph neural network, i.e. a TSGN, sea surface temperature data received from a group of multiple sensors distributed in a physical environment, the sea surface temperature data relating to physical and environmental conditions of the physical environment.); and
generating a forecast of one or more future conditions for the physical environment based on the graph neural network using one or more forecasting machine learning models (like noted above, Sun teaches inputting historical sea surface temperature data into the TSGN, i.e. a graph neural network, whereby the TSGN outputs a predicted sea surface temperature for a future period:
For the complex dependencies of sea surface temperature data in the time and space dimensions, we propose a graph neural network called a time-series graph network (TSGN) by combining the advantages of a long short-term memory (LSTM) network in processing temporal information. The model is based on the graph structure of the sensor network and uses the topology between sensors to model the geometric structure of the sensors. LSTM aggregation is adopted for SST data containing temporal information. This model combines graph convolution and gated time convolution to extract the most useful spatial features and coherently capture the most basic temporal features.
The main contributions of this paper are as follows:
(1) A graph learning network, i.e., a TSGN, is designed, which has strong learning ability for data prediction tasks containing time series information.
(2) A TSGN employs an LSTM to aggregate the node features to better capture the time correlation and improve the prediction accuracy.
(3) A TSGN is applied to an SST prediction task, which is a regression fitting, and the experimental results show good prediction results.
(Section 1 “Introduction”; emphasis added).
As shown in Fig.1, the task of graph-based SST prediction can be defined as follows: build a graph neural network, input historical temperature data
X
∈
R
N
×
T
i
n
×
D
(where N is the number of sensors,
T
i
n
is the length of the time window, and D is the input feature dimension) and a graph G reflecting the spatial connectivity between temperature sensors, and output the temperature for a future period
Y
∈
R
N
×
T
o
u
t
×
D
(where
T
o
u
t
is the prediction step and d is the number of temperatures).
(Section 2.1 “Problem Definition”; emphasis added).
The prediction task in our experiments was defined as using 28 (
T
i
n
=
24
, sampling period of 6 h) historical temperature data of all nodes to predict the next 4 (
T
o
u
t
=
4
, sampling period of 6 h) temperature values, that is, the temperature values of the first 7 consecutive days are input to predict the temperature values of the next day. A sliding window with a window size of 32 was used to generate 7304 data samples with a step size of 1. The data are divided into a training set, an evaluation set, and a test set at a ratio of 7:1:2 in chronological order.
(Section 4.2 “Experimental setup”; emphasis added).
The TSGN as trained to predict a sea surface temperature for a future period is considered a “forecasting model” like claimed. Accordingly, Sun teaches generating a forecast of one or more future conditions, i.e. sea surface temperatures, for the physical environment based on the graph neural network using one or more forecasting machine learning models, i.e. a trained TSGN.).
Sun thus teaches a method similar to that of claim 1. Sun discloses that such teachings can be implemented via computer program instructions that are executable by one or more processors of a computer system (see section 4.2 “Experimental setup”). Such a computer system comprising executable instructions to carry out the above-described teachings of Sun is considered a system similar to that of claim 8. The memory necessary in such a computer system to store the executable instructions is considered a computer program product similar to that of claim 15. Sun, however, does not teach determining residuals via the multiple sensors and via subtracting a time averaged mean from a measurement, wherein in response to one or more of the residuals exceeding a zero value and exceeding a threshold value an action is performed, as is required by claims 1, 8 and 15. Moreover, Sun does not explicitly disclose that the one or more forecasting machine learning models incorporate depth gradients of the multiple sensors and the residuals to generate the forecast, and that an alert indicating an anomalous forecast is presented in response to the one or more forecasting machine learning models classifying the forecast as an anomalous forecast, as is further required by claims 1, 8 and 15.
Soualhia teaches that statistical properties of features used to train machine learning models can change over time, and thereby result in degraded performance of the models (see e.g. page 1, lines 6-22). To remedy this, Soualhia teaches detecting and/or predicting data drift in an input data stream, determining a compensation function, and applying the compensation function to the input data stream to offset at least part of the data drift (see e.g. page 1, line 25 – column 2, line 8). Soualhia particularly discloses that data drift can be detected in an input data stream by subtracting a time averaged mean (i.e. a feature’s mean value calculated over a training window) from an online/streamed feature value (i.e. determining the difference between the streamed current value and the mean value of the feature) (see e.g. page 9, line 30 – page 10, line 3; page 10, lines 17-25; page 12, lines 17-21; page 12, line 30 – page 13, line 9; page 13, lines 21-31; and page 18, line 26 – page 19, line 13). In response to the drift exceeding a zero value and a threshold value (i.e. exceeding the feature’s mean deviation or a second threshold), an action is performed (i.e. a compensation function is applied to the data to offset the drift, or the machine learning model is retrained) (see e.g. page 3, lines 7-19; page 13, lines 2-6; page 14, lines 1-10; page 18, lines 3-19; and page 19, lines 1-31).
It would have been obvious to one of ordinary skill in the art, having the teachings of Sun and Soualhia before the effective filing date of the claimed invention, to modify the method, system and computer program product taught by Sun so as to determine any drift of the input data (i.e. the data obtained via the multiple sensors) by subtracting a time averaged mean from a current value (i.e. from a sensor measurement), and wherein in response to the drift exceeding a zero value and exceeding a threshold value, an action is performed like taught by Soualhia. The drift can be considered a “residual” like claimed. It would have been advantageous to one of ordinary skill to utilize such a combination because it can provide for more accurate predictions from the machine learning model, as is taught by Soualhia (see e.g. page 1, lines 6-22; and page 8, lines 16-26).
Zhang generally teaches generating a forecast of one or more future conditions (i.e. future ocean temperatures) for a physical environment using one or more forecasting machine learning models, wherein the one or more forecasting machine learning models incorporate depth gradients (i.e. depth levels), understandably of multiple sensors, to generate the forecast:
Sea surface temperature (SST) prediction has raised considerable attention in various ocean-related fields. However, these methods were only limited to the time-sequence prediction of some isolated points, and their spatial linkage was not considered. Furthermore, these studies only predict the temperature of sea surface, but the subsurface temperature in the inner ocean is much more important. In this letter, we propose a model of multilayer convolutional long- and short-term memory (M-convLSTM) to predict 3-D ocean temperature, comprising convolutional neural networks (CNNs), long- and short-term memory (LSTM), and multiple layer stacking to consider the horizontal and vertical temperature variations from sea surface to subsurface to be about 2000 m below. Global marine environment observation data (ARGO) are used to conduct the prediction of 3-D ocean temperature in this letter, and the results demonstrate the overall good accuracy of forecast and ARGO data.
(Abstract; emphasis added).
The ARGO standard global 1
°
gridded monthly average data set ranging from 2005 to 2017 (156 months) is collected from http://www.ARGO.ucsd.edu. The data size of one month is 180 × 360 × 27, where 27 represents the number of layers in different depths. To speed up the training of the M-convLSTM model, a typical rectangular area (60
°
× 40
°
) of (E130
°
–190
°
, N10
°
–50
°
) in the Pacific Ocean is selected here. Fig. 1 shows the ARGO temperature of sea surface (SST, in the depth of 0 m) in global and selected areas in January 2005.
We construct input data in the following format: [128, 28, 27, 60, 40], where 128 represents the data capacity, and the M-convLSTM predicts one step every 28 months. As shown in Fig. 2, x is the batch of training the temperature data set and y is the temperature true value of the next month of x. Through a continuous backpropagation of error process, the entire network finally “learns” the temperature change of different depth layers in the selected region, thereby predicting the data of the next time.
(Section II.A “DATA”; emphasis added).
The key to the CNN concept is to extract the local image features using the convolution kernel. In the convolution pane, the spatial position of each pixel is extracted, revealing that the CNN is suitable for fields such as image processing and feature recognition. However, LSTM is suitable for learning information on time series data. The global gridded ARGO monthly average data set can be processed as a time-serialized image data, and the prediction or backtracking tasks based on the data set are similar to the prediction of the image-generation algorithm.
Fig. 3 shows the structure of M-convLSTM. A total of 27 layers represent the input data set of 27 depth layers. For each layer, the input information
x
t
enters the same time cell from the previous time cell (A). It activates the function “tanh” and then passes it to the next time cell. The information received by each cell of the LSTM includes the hidden layer state
h
t
-
1
at the last moment. Both the input
x
t
and output of the previous time pass through four different activation functions (corresponding to different “gates”).
(Section II.B. “M-ConvLSTM Model”; emphasis added).
The experimental result dimension is (64, 40, 27, 8), where 27 represents the number of depth layers and 8 represents the time step. For evaluation, we reserved the true values of temperature values for several moments to evaluate the prediction accuracy.
(Section III. “RESULTS”; emphasis added).
In this letter, the M-convLSTM model that combines CNN and LSTM with layer stacking is proposed to predict the 3-D ocean temperature. It can learn the horizontal variation and vertical variation of temperature across different depths in the inner ocean. Experimental results show that the M-convLSTM offers good prediction accuracy. For several moments of prediction, the upper layer offers good accuracy as the time step increases. The accuracy of the upper layers is better than that of the deeper layers, possibly because the variation of the deeper layers is smaller than that of the upper layers. For future work, the M-convLSTM should be able to improve accuracy in the middle and deeper layers.
(Section IV. “CONCLUSION”; emphasis added).
It would have been obvious to one of ordinary skill in the art, having the teachings of Sun, Soualhia and Zhang before the effective filing date of the claimed invention, to modify the method, system and computer program product taught by Sun and Soualhia such that, in addition to the data drifts (i.e. residuals), the one or more forecasting machine learning models incorporate depth gradients of the multiple sensors to generate the forecast, as is taught by Zhang. It would have been advantageous to one of ordinary skill to utilize such a combination because it would enable subsurface temperatures to be predicted, which is important, as is suggested by Zhang (see e.g. the Abstract).
Similar to Sun, Soualhia and Zhang, Kim teaches generating a forecast of one or more future conditions (i.e. sea surface temperatures) for a physical environment using one or more forecasting machine learning models (see e.g. Section 1 “Introduction,” which recites “[t]herefore, in this study, we present a recurrent neural network (RNN)-based long short-term memory (LSTM) model based on deep-learning technology [1,2], to predict sea surface temperatures (SSTs).”). Regarding the claimed invention, Kim further suggests presenting an alert indicating an anomalous forecast (i.e. high water temperatures) in response to the one or more forecasting machine learning models classifying the forecast as an anomalous forecast:
Due to global warming, high water temperatures (HWTs) are frequently observed along the coast of the Korean Peninsula. This phenomenon has led to mass mortality of farmed fish, resulting in massive economic losses to fishermen. The HWT warning period lasted for a total of 32 days in 2017, but persisted for 43 days in 2018. If this trend continues, the damage resulting from HWTs will likely be further exacerbated. To prevent and mitigate exposure risk, it is necessary to predict HWT occurrence accurately in advance. Therefore, in this study, we present a recurrent neural network (RNN)-based long short-term memory (LSTM) model based on deep-learning technology [1,2], to predict sea surface temperatures (SSTs).
Generally, extreme MHW is defined as the top 10% of all SST values observed over the past 30 years for a particular body of water [3,4]. However, in this study, we define HWTs in the context of typical water temperatures in Korea. We follow the criteria defined by the Korean Ministry of Maritime Affairs and Fisheries, which operates a HWT alert system to prevent damage to the aquaculture sector and respond as needed to these events. The HWT alert system consists of the following three stages: level of interest, which occurs 7 days prior to onset of the temperature increase; level of watch, when the water temperature reaches 28
°
C
; and level of warning, when the water temperature exceeds 28
°
C
and lasts for 3 or more days. Since the Korea Ministry of Maritime Affairs evaluates HWT based on a threshold of 28
°
C
, we defined HWT as > 28
°
C
. In this study, the area selected for HWT prediction was the coastal region extending from Goheung to Yeosu, Jeollanam-do, Korea. This region has a high concentration of fish farms. Thus, persistent HWTs cause serious damage to the local fishing industry, as observed in the 2018 disaster, which involved the loss of 54.1 million fish/shellfish in Jeollanam-do, with a total property damage of USD 38 million [5].
(Section 1 “Introduction”; emphasis added).
The HWT area and frequency are increasing in coastal areas off the Korean Peninsula due to global warming [32,33]. Figure 2 shows average SSTs for the target area. The maximum SST in the 7-year period between 2009 and 2015 was about 27 ◦C, and that from 2016 to 2018 (3 years) was about 28.5 ◦C. Thus, the target area, which contains numerous fish farms, experienced concentrated HWTs during the past 3 years. If HWT events in this area can be forecast in advance, the fishing industry may be able to respond more quickly to moderate the economic damage.
(Section 2.1 “Study Area”).
It would have been obvious to one of ordinary skill in the art, having the teachings of Sun, Soualhia, Zhang and Kim before the effective filing date of the claimed invention, to modify the method, system and computer program product taught by Sun, Soualhia and Zhang so as to present an alert indicating an anomalous forecast in response to the one or more forecasting machine learning models classifying the forecast as an anomalous forecast, as is taught by Kim. It would have been advantageous to one of ordinary skill to utilize such a combination because it can “moderate economic damage,” as is suggested by Kim (see e.g. the portion of Section 2.1 “Study Area” cited above). Accordingly, Sun, Soualhia, Zhang and Kim are considered to teach, to one of ordinary skill in the art, a method like that of claim 1, a system like that of claim 8, and a computer program product like that of claim 15.
Claims 6, 7, 13, 14, 20 and 21 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Sun, Soualhia, Zhang and Kim, which is described above, and also over the article entitled, “Multivariate Seawater Quality Prediction Based on PCA-RVM Supported by Edge Computing Towards Smart Ocean” by Sun et al. (“Sun II”).
Regarding claims 6, 13 and 20, Sun, Soualhia, Zang and Kim teach a method like that of claim 1, a system like that of claim 8, and a computer program product like that of claim 15, as is described above, and which entail generating a forecast of one or more future conditions (i.e. future sea surface temperatures) using one or more forecasting machine learning models. Sun further teaches that the one or more forecasting machine learning models generate a model representation of the physical and environmental conditions (Like noted above, Sun discloses that historical sea surface temperature data is used to train a graph neural network, i.e. a TSGN, to predict future sea surface temperatures based on a sequence of historical sea surface temperature data:
For the complex dependencies of sea surface temperature data in the time and space dimensions, we propose a graph neural network called a time-series graph network (TSGN) by combining the advantages of a long short-term memory (LSTM) network in processing temporal information. The model is based on the graph structure of the sensor network and uses the topology between sensors to model the geometric structure of the sensors. LSTM aggregation is adopted for SST data containing temporal information. This model combines graph convolution and gated time convolution to extract the most useful spatial features and coherently capture the most basic temporal features.
The main contributions of this paper are as follows:
(1) A graph learning network, i.e., a TSGN, is designed, which has strong learning ability for data prediction tasks containing time series information.
(2) A TSGN employs an LSTM to aggregate the node features to better capture the time correlation and improve the prediction accuracy.
(3) A TSGN is applied to an SST prediction task, which is a regression fitting, and the experimental results show good prediction results.
(Section 1 “Introduction”; emphasis added).
As shown in Fig.1, the task of graph-based SST prediction can be defined as follows: build a graph neural network, input historical temperature data
X
∈
R
N
×
T
i
n
×
D
(where N is the number of sensors,
T
i
n
is the length of the time window, and D is the input feature dimension) and a graph G reflecting the spatial connectivity between temperature sensors, and output the temperature for a future period
Y
∈
R
N
×
T
o
u
t
×
D
(where
T
o
u
t
is the prediction step and d is the number of temperatures).
(Section 2.1 “Problem Definition”; emphasis added).
The prediction task in our experiments was defined as using 28 (
T
i
n
=
24
, sampling period of 6 h) historical temperature data of all nodes to predict the next 4 (
T
o
u
t
=
4
, sampling period of 6 h) temperature values, that is, the temperature values of the first 7 consecutive days are input to predict the temperature values of the next day. A sliding window with a window size of 32 was used to generate 7304 data samples with a step size of 1. The data are divided into a training set, an evaluation set, and a test set at a ratio of 7:1:2 in chronological order.
We used samples from the training set to optimize the model parameters. The performance of the model was evaluated by optimizing the hyperparameters of the execution results with the evaluation set and validating them with the test set.
(Section 4.2 “Experimental setup”; emphasis added).
Such a TSGN trained to predict future sea surface temperatures based on a sequence of historical sea surface temperature data is considered a model representation of physical and environmental conditions, i.e. of sea surface temperatures, and is generated using a prediction model, e.g. a TSGN.). Accordingly, Sun, Soualhia, Zhang and Kim teach a method similar to that of claim 6, a system similar to that of claim 13, and a computer program product similar to that of claim 20. Sun, Soualhia, Zhang and Kim, however, do not explicitly disclose that the one or more forecasting machine-learning models comprise an “edge-based” prediction model that generates the model representation, as is required by claims 6, 13 and 20.
Sun II generally describes “a hybrid multivariate prediction model for seawater quality assessment in an edge computing environment, considering the combination of principal component analysis (PCA) and relevance vector machine (RVM).” (Abstract). Regarding the claimed invention, Sun II particularly teaches an edge-based prediction model that generates a model representation of physical and environmental conditions (Sun II generally teaches that edge computing can provide a solution to handling massive data generated by oceanic sensor devices:
Massive data, generated from oceanic sensor devices, inevitably requires high-bandwidth to guarantee efficient transmission. Along with an explosive increase of such data, the communication network may be finally overwhelmed by the overload transmission. It throws a severe challenge to online data processing. In this situation, mobile edge computing (MEC) is viewed as a workable solution. Through deployment of cloud-like infrastructure near data sources, data can be preferentially handled at the edge, largely relaxing the whole network [12]. Given this, this paper aims at investigating the MEC-based seawater quality prediction, in order to make the quick response for potential ocean emergencies.
(Page 54506).
Sun II further teaches generating a model representation of physical and environmental conditions, e.g. of PH or DO, for deployment on an edge computing node:
The overall prediction framework in edge computing environment is shown in the Fig 1. The edge computing node can gather data from oceanic collectors, such as buoy, surveying vessel, underwater vehicle and aircraft. Our seawater quality prediction can be achieved in this edge. Functionally, this framework consists of two key functional components, namely the PCA analyzer and the RVM predictor. The former lies in reasonably reducing the dimension of multivariate ocean data, to obtain good clustering quality, while the latter realize a good nonlinear approximation on this basis.
Firstly, various types of seawater quality data are collected through sensor devices deployed in ocean. Generally speaking, multivariate time series often contain more dynamic information of the seawater quality than the univariate time series. Then this multivariate series is injected into the PCA analyzer for data extraction, where redundant information hidden in multiple variables can be eliminated by linear transformation. Actually, this data preprocess can provide more excellent initialization input for predictor. Finally, we consider the multi-input-single-output RVM predictor for seawater quality modeling, where the input signal is a combination of all seawater quality factors, and the model readout is either pH or DO. In order to better understand our proposal, we describe these characteristics of both PCA and RVM in detail.
(Pages 54507-54508).
The model as deployed on the edge computing device is considered an edge-based prediction model.).
It would have been obvious to one of ordinary skill in the art, having the teachings of Sun, Soualhia, Zhang, Kim and Sun II before the effective filing date of the claimed invention, to modify the method, system and computer program product taught by Sun, Soualhia, Zhang and Kim such that the one or more machine learning models comprise a prediction model deployed on an edge computing node (i.e. is edge-based) to generate the model representation of the physical and environmental conditions, like taught by Sun II. It would have been advantageous to one of ordinary skill to utilize such edge computing because it can reduce the load on the communication network, as is suggested by Sun II (see e.g. page 54506). Accordingly, Sun, Soualhia, Zhang, Kim and Sun II are considered to teach, to one of ordinary skill in the art, a method like that of claim 6, a system like that of claim 13, and a computer program product like that of claim 20.
As per claims 7, 14 and 21, it would have been obvious, as is described above (see the rejection for claims 1, 8 and 15), to modify the method, system and computer program product taught by Sun, Soualhia and Zhang so as to present an alert indicating an anomalous forecast in response to the one or more forecasting machine learning models classifying the forecast as an anomalous forecast, as is taught by Kim. Kim particularly suggests determining one or more anomalies (i.e. high water temperatures, HWTs) from current physical and environmental conditions based on the model representation of physical and environmental conditions (see e.g. section 6 “Conclusions”). Accordingly, the above-described combination of Sun, Soualhia, Zhang, Kim and Sun II is further considered to teach a method like that of claim 7, a system like that of claim 14, and a computer program product like that of claim 21.
Claims 22, 23, 25, 26, 28 and 29 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Sun, Soualhia, Zhang and Kim, which is described above, and also over the article entitled, “3D Ensemble Simulation of Seawater Temperature – An Application for Aquaculture Operations” by Shettigar et al. (“Shettigar”).
Regarding claims 22, 25 and 28, Sun, Soualhia, Zang and Kim teach a method like that of claim 1, a system like that of claim 8, and a computer program product like that of claim 15, as is described above, and which entail incorporating into a graph neural network data (e.g. sea surface temperature data) received from a group of multiple sensors distributed in a physical environment, and generating a forecast of one or more future conditions (i.e. seawater temperature) for the physical environment based on the graph neural network. Sun particularly teaches that the physical environment can comprise an area of an ocean, e.g. the Northwest Pacific Ocean (see section 4.1 “Dataset and data pre-processing”). Sun, Soualhia, Zang and Kim however do not explicitly disclose that the physical environment comprises an aquaculture farm to be affected via the anomalous forecast, as is required by claims 22, 25 and 28.
Shettigar nevertheless suggests predicting sea water temperatures for a physical environment using one or more models, wherein the physical environment comprises an aquaculture farm that would be affected via an anomalous prediction (see e.g. “INTRODUCTION” on pages 1-2 and “Aquaculture in Greece” on page 3). Shettigar teaches that forecasts of sea water temperatures, including anomalous forecasts (e.g. extreme seawater temperature), can benefit aquaculture farms to plan short and long-term operations in advance (see e.g. “INTRODUCTION”).
It would have been obvious to one of ordinary skill in the art, having the teachings of Sun, Soualhia, Zhang, Kim and Shettigar before the effective filing date of the claimed invention, to modify the method, system and computer program product taught by Sun, Soualhia, Zhang and Kim such that the physical environment for which the forecast is generated comprises an aquaculture farm like taught by Shettigar, which would be affected by the anomalous forecast. It would have been advantageous to one of ordinary skill to utilize such a combination, because the aquaculture farm would then be able to beneficially apply the model forecast to plan operations, as is taught by Shettigar (see e.g. “INTRODUCTION”). Accordingly, Sun, Soualhia, Zhang, Kim and Shettigar are considered to teach, to one of ordinary skill in the art, a method like that of claim 22, a system like that of claim 25, and a computer program product like that of claim 28.
As per claims 23, 26 and 29, it would have been obvious, as is described above, to modify the method, system and computer program product taught by Sun, Soualhia, Zhang and Kim such that the physical environment for which the forecast is generated comprises an aquaculture farm like taught by Shettigar, which would be affected by the anomalous forecast. Kim particularly teaches that the aquaculture farm can comprise one or more cages that would be affected via the anomalous forecast (see e.g. “INTRODUCTION” on pages 1-2 and “Aquaculture in Greece” on page 3). Accordingly, Sun, Soualhia, Zhang, Kim and Shettigar are further considered to teach a method like that of claim 23, a system like that of claim 26, and a computer program product like that of claim 29.
Claims 24, 27 and 30 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Sun, Soualhia, Zhang and Kim, which is described above, and also over U.S. Patent Application Publication No. 2022/0198303 to Ito et al. (“Ito”).
Regarding claims 24, 27 and 30, Sun, Soualhia, Zang and Kim teach a method like that of claim 1, a system like that of claim 8, and a computer program product like that of claim 15, as is described above, and which entail incorporating into a graph neural network data received from a group of multiple sensors distributed in a physical environment, and generating a forecast of one or more future conditions for the physical environment based on the graph neural network. Sun, Soualhia, Zang and Kim, however, do not explicitly disclose that the one or more future conditions comprise toxic algae concentration for a body of water, as is required by claims 24, 27 and 30.
Ito nevertheless teaches incorporating, into a machine learning model, data received from e.g. sensors in a physical environment, and generating a forecast of one or more future conditions for the physical environment, wherein the one or more future conditions comprise toxic algae concentration (e.g. red tide) for a body or water (see e.g. paragraphs 0002-0006, 0011-0012, 0018, 0022, 0040, 0050-0051 and 0055).
It would have been obvious to one of ordinary skill in the art, having the teachings of Sun, Soualhia, Zhang, Kim and Ito before the effective filing date of the claimed invention, to modify the method, system and computer program product taught by Sun, Soualhia, Zhang and Kim such that the one or more future conditions additionally or alternatively include toxic algae concentration for a body of water, as is taught by Ito. It would have been advantageous to one of ordinary skill to utilize such a combination, because by predicting toxic algae concentrations, damage to marine industry may be prevented or mitigated, as is suggested by Ito (see e.g. paragraph 0002). Accordingly, Sun, Soualhia, Zhang, Kim and Ito are considered to teach, to one of ordinary skill in the art, a method like that of claim 24, a system like that of claim 27, and a computer program product like that of claim 30.
Response to Arguments
The Examiner acknowledges the Applicant’s amendments to claims 1, 6, 8 and 15, and addition of new claims 22-30. In response to these amendments, the objections presented in the previous Office Action to claims 6 and 7 are respectfully withdrawn.
Regarding the 35 U.S.C. § 101 rejections, the Applicant argues that the claims, as amended, are not directed to an abstract idea.
The Examiner however respectfully disagrees. Independent claim 1 recites “generating a forecast of one or more future conditions for the physical environment…using one or more forecasting…models, the one or more forecasting…models incorporating depth gradients of the multiple sensors and the residuals to generate the forecast.” Independent claims 8 and 15 recite similar limitations. For the reasons described in the 35 U.S.C. § 101 rejections presented above, the generation of a forecast of one or more future conditions using one or more forecasting models, when given its broadest reasonable interpretation, can be considered an evaluation or judgement that can practically be performed in the human mind. Moreover, the newly-added limitation of “determining residuals…via subtracting a time averaged mean from a measurement” can be considered a recitation of a mathematical concept.
Further regarding the 35 U.S.C. § 101 rejections, the Applicant argues that the claims, even if directed to an abstract idea, are integrated into a practical application. In particular, the Applicant submits that the claims provide additional elements that reflect an improvement in anomaly or risk detection that is performed in a collaborative forecasting operation.
The Examiner, however, respectfully disagrees. The claims do not reflect an improvement in computer capabilities per se, as the claims employ generic computing machines and rely on the use of generic machine learning technology. Other than a general disclosure of receiving data from a group of multiple sensors, the claims do not reflect an improvement in anomaly or risk detection. It is not clear from the disclosure as to how the incorporation of residuals would improve computer capabilities or anomaly or risk detection. The Examiner also respectfully notes that an improvement in the abstract idea itself is not an improvement in technology. See, e.g., Trading Technologies Int’l v. IBG, 921 F.3d 1084, 1093-94, 2019 USPQ2d 138290 (Fed. Cir. 2019). Accordingly, the Examiner respectfully maintains that the claims do not recite additional elements that integrate the judicial exception into a practical application. The 35 U.S.C. § 101 rejections are thus respectfully maintained.
The Applicant’s arguments concerning the 35 U.S.C. § 103 rejections presented in the previous Office Action have been considered, but are moot in view of the new grounds of rejection presented above.
Conclusion
The prior art made of record on form PTO-892 and not relied upon is considered pertinent to applicant’s disclosure. The applicant is required under 37 C.F.R. §1.111(C) to consider these references fully when responding to this action. In particular, the article by Hobday et al. cited therein (“Seasonal forecasting for decision support in marine fisheries and aquaculture”) teaches generating a forecast of one or more future conditions for a physical environment that comprises an aquaculture farm having one or more cages.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BLAINE T BASOM whose telephone number is (571)272-4044. The examiner can normally be reached Monday-Friday, 9:00 am - 5:30 pm, 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, Matt Ell can be reached at (571)270-3264. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/BTB/
1/24/2026
/MATTHEW ELL/Supervisory Patent Examiner, Art Unit 2141