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 .
Response to Amendment
Applicant’s submission filed on 07/29/2025 has been entered. The status of claims is as follows:
Claims 1-2, 4-21 remain pending in the application.
Claim 3 is cancelled.
Claims 1, 9 and 16 are amended.
Response to Arguments
In reference to the Claim Rejections under 35 U.S.C 101:
Applicant asserts on Remarks pg. 1-3 that the amended are patent-eligible under 101 because, although they recite a judicial exception, the claims integrate that exception into a practical application that improves the functioning and reliability of a computer system – specifically, an artificial neural network (ANN) retrained in situ within the same memory after deployment. The specification explains that remotely deployed sensors face reliability, error-propagation, and maintenance challenges, and the invention addresses these issues by enabling in-memory retraining when performance falls below a threshold. This disclosed improvement enhances mean-time-to-failure and overall system reliability, demonstrating a technological advancement as required by MPEP 2106.04(d). Therefore, the Applicant contends that the Examiner’s characterization of the retraining as insignificant extra-solution activity is incorrect because the in-situ post-deployment retraining forms the core technological solution to the identified problems.
Applicant’s arguments have been fully considered but are not persuasive. Although the amended claims now recite that the remote sensor system may be deployed in locations such as 5G network, smart camera, radar, or space/ subsea equipment, and further recite in-situ retraining of the artificial neural network (ANN) after evaluating its performance, the claims still recite an abstract idea and do not integrate that abstract idea into a practical application as required by Step 2A Prong Two of the eligibility analysis. The claim continues to be directed to the abstract idea of mathematical analysis and mathematical manipulation of data, including classification, identification, analysis, comparison of ANN outputs to known outputs, determining a difference, and updating weights/ biases – all of which fall within the judicial exception of mathematical concepts. The recited “retraining circuitry”, “memory” and “controller” are described generically and perform only their well-understood functions of storing data and executing mathematical operations. The specification likewise describes the retraining as conventional ANN training activity.
The additional elements – deploying the sensor system in various environments, performing ANN evaluation post-deployment, and retraining “in-situ” – do not impose any meaningful limits that would amount to a practical application. The claim does not recite any specific technological improvement to the memory, controller, communication hardware, or ANN architecture. Instead, the claim merely applies the abstract mathematical operations in the context of a generic remote sensor system, which is insufficient under MPEP 2106.04(d). Improving “mean time to failure” of the sensor system is stated only as a result of performing the abstract mathematical processes, not as a technological improvement rooted in the claim’s structure. As such, the claims do not effect an improvement to the functioning of a computer or other technology, but instead use generic computer components as tools to implement the abstract idea. Accordingly, the claims fail to integrate the judicial exception into a practical application and do not amount to significantly more under Step 2B. The 101 rejection is therefore maintained.
Applicant’s arguments filed 07/29/2025 have been fully considered but they are not persuasive.
In reference to the Claim Rejections under 35 U.S.C 103:
Applicant argues on Remarks pg. 4 that the cited prior arts do not teach, suggest or render obvious every limitation of independent claims 1, 9 and 16, wherein each of the independent claims recites “wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure”.
Examiner respectfully disagrees and notes that although the Examiner indicated during the interview that the proposed amendments appeared to overcome the 103 rejections, upon further consideration of the full claim amendments as submitted in the response, the Examiner finds that the previously cited references still teach or render obvious the newly added limitations “wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure”. Specifically, Turcot discloses in ¶[0078]: “The sensor data can be obtained from the vehicle occupant along with the video data or the audio data, instead of the video data or the audio data, etc. In embodiments, the sensor data can include one or more of vehicle temperature, outside temperature, time of day, level of daylight, weather conditions, headlight activation, windshield wiper activation, entertainment center selection, or entertainment center volume.”, ¶[0079]: “Images, which can include facial or torso data, human perception state data, audio data, and physiological data, can be collected using multiple mobile devices. The image data can be applied to neural network training, where the neural network training can enable deep learning. The deep learning can include in situ retraining.”, and ¶[0081]: “A mobile device can include a front-side camera and/or a back-side camera that can be used to collect expression data. A mobile device can include a microphone, audio transducer, or other audio capture apparatus that can be used to capture the speech and nonspeech vocalizations. Sources of expression data can include a webcam 1222, a phone camera 1242, a tablet camera 1252, a wearable camera 1262, and a mobile camera 1230. A wearable camera can comprise various camera devices, such as a watch camera 1272. Sources of audio data 1282 can include a microphone 1280.”. Accordingly, the 103 rejections are maintained.
Applicant’s arguments filed 07/29/2025 have been fully considered but they are not persuasive.
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 1-2, 4-21 are rejected under U.S.C 101 for containing an abstract idea without significantly more.
Regarding claim 1:
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is a process.
Step 2A – Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, the claim recites an abstract idea.
evaluating performance of the artificial neural network by the controller post-deployment of the remote sensor system; - This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
comparing an output of the artificial neural network to a known output for the representative dataset; - This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
determining a difference between the output and the known output; - This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
Step 2A – Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
operating an artificial neural network allocated to memory post-deployment of a remote system including the memory and a controller coupled to the memory, wherein the controller includes retraining circuitry; – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
wherein operating the artificial neural network includes performing operations with respect to data from the sensor, the operations comprising one or more of classification, identification, and analysis of the data from the sensor. – This limitation is directed to field of use (see MPEP 2106.05(h))
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure; – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
wherein [evaluating comprises] inputting a representative dataset to the artificial neural network This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)).
retraining, in-situ by the retraining circuitry post-deployment of the remote system, the artificial neural network at least partially in response to the evaluation yielding a sub-threshold result. This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)).
wherein the method is performed by the controller; – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
operating the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
No, there are no additional elements that amount to significantly more than the judicial exception. The additional elements are:
operating an artificial neural network allocated to memory post-deployment of a remote system including the memory and a controller coupled to the memory, wherein the controller includes retraining circuitry; – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
wherein operating the artificial neural network includes performing operations with respect to data from the sensor, the operations comprising one or more of classification, identification, and analysis of the data from the sensor. – This limitation is directed to field of use (see MPEP 2106.05(h))
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure; – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
wherein [evaluating comprises] inputting a representative dataset to the artificial neural network – This limitation is directed to receiving or transmitting data over a network. The courts have recognized receiving or transmitting data over a network as well understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity (see MPEP 2106.05(d) II.).
retraining, in-situ by the retraining circuitry post-deployment of the remote system, the artificial neural network at least partially in response to the evaluation yielding a sub-threshold result. This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)).
wherein the method is performed by the controller; – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
operating the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
Regarding claim 2,
Claim 2 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 1 which includes an abstract idea (see rejection for claim 1). The additional limitations:
evaluating performance of the artificial neural network after retraining; and - This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
determining whether to retrain the artificial neural network again based on results of the evaluation. This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
Regarding claim 4,
Claim 4 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 1 which includes an abstract idea (see rejection for claim 1). The additional limitations:
wherein evaluating comprises periodically evaluating performance of the artificial neural network. This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
Regarding claim 5,
Claim 5 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 4 which includes an abstract idea (see rejection for claim 4). The additional limitations:
receiving an indication of a desired frequency of the periodic evaluation from a host of the memory; and – This limitation is directed to receiving or transmitting data over a network. The courts have recognized receiving or transmitting data over a network as well understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity (see MPEP 2106.05(d) II.).
wherein periodically evaluating comprises periodically evaluating at the desired frequency. This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
Regarding claim 6,
Claim 6 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 4 which includes an abstract idea (see rejection for claim 4). The additional limitations:
detecting an error in the memory; and This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
wherein retraining further comprises retraining at least partially in response to the error. This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)), and also directed to well-understood, routine and conventional as evidenced by Huang et al. (US 20210271809 A1) ([0091]: “For example, in the classification task, features of objects to be labelled passing the verification can be used as features of the training samples and the manual labelling results as labels of the training samples to generate the training samples to retrain or incrementally train and the machine learning model. The retraining or incremental training process of the model is well known in the art, which is not repeated here.”)
Regarding claim 7,
Claim 7 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 6 which includes an abstract idea (see rejection for claim 6). The additional limitations:
wherein retraining accounts for the error without remapping storage of weights in the memory. This claim merely recites a further limitation on the retraining at least partially in response to the error from Claim 6 which was directed to insignificant extra-solution activity (see MPEP 2106.05(g)), and also directed to well-understood, routine and conventional as evidenced by Huang et al. (US 20210271809 A1) ([0091]: “For example, in the classification task, features of objects to be labelled passing the verification can be used as features of the training samples and the manual labelling results as labels of the training samples to generate the training samples to retrain or incrementally train and the machine learning model. The retraining or incremental training process of the model is well known in the art, which is not repeated here.”)
Regarding claim 8,
Claim 8 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 6 which includes an abstract idea (see rejection for claim 6). The additional limitations:
wherein detecting the error comprises detecting a bit error or bit freeze in the memory. This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
Regarding claim 9,
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is a process.
Step 2A – Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, the claim recites an abstract idea.
compare an output of the artificial neural network to a known output for the representative dataset to periodically evaluate performance of an artificial neural network post-deployment of the remote sensor system at a predefined frequency, - This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
determining a difference between the output and the known output; - This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
Step 2A – Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
a sensor This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
a memory device; and This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
a controller coupled to the memory device and configured to; – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
operate an artificial neural network allocated to the memory device post-deployment of the remote sensor system, Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
wherein operating the artificial neural network includes performing operations with respect to data from the sensor, the operations comprising one or more of classification, identification, and analysis of the data from the sensor – This limitation is directed to field of use (see MPEP 2106.05(h))
input a representative dataset to the artificial neural network This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)).
wherein the artificial neural network is allocated to the memory device; and – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
retrain the artificial neural network in-situ post-deployment of the remote sensor system at least partially in response to a sub-threshold result of the periodic performance evaluation. This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)).
operate the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system; Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure. – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
No, there are no additional elements that amount to significantly more than the judicial exception. The additional elements are:
a sensor This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
a memory device; and This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
a controller coupled to the memory device and configured to; – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
operate an artificial neural network allocated to the memory device post-deployment of the remote sensor system, Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
wherein operating the artificial neural network includes performing operations with respect to data from the sensor, the operations comprising one or more of classification, identification, and analysis of the data from the sensor – This limitation is directed to field of use (see MPEP 2106.05(h))
input a representative dataset to the artificial neural network – This limitation is directed to receiving or transmitting data over a network. The courts have recognized receiving or transmitting data over a network as well understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity (see MPEP 2106.05(d) II.).
wherein the artificial neural network is allocated to the memory device; and – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
retrain the artificial neural network in-situ post-deployment of the remote sensor system at least partially in response to a sub-threshold result of the periodic performance evaluation. This limitation is also directed to well-understood, routine and conventional as evidenced by Huang et al. (US 20210271809 A1) ([0091]: “For example, in the classification task, features of objects to be labelled passing the verification can be used as features of the training samples and the manual labelling results as labels of the training samples to generate the training samples to retrain or incrementally train and the machine learning model. The retraining or incremental training process of the model is well known in the art, which is not repeated here.”)
operate the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system; Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure. – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
Regarding claim 10,
Claim 10 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 9 which includes an abstract idea (see rejection for claim 9). The additional limitations:
wherein the predefined frequency is a user-defined parameter based at least in part on reliability or lifetime expectations of the apparatus. This claim merely recites a further limitation on the periodically evaluate performance of an artificial neural network at a predefined frequency from Claim 9 which was directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
Regarding claim 11,
Claim 11 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 9 which includes an abstract idea (see rejection for claim 9). The additional limitations:
wherein the apparatus stores [[a]] the representative dataset and [[a]] the known output of the artificial neural network for the representative dataset; and – This limitation is directed to storing and retrieving information in memory, which the court have recognized as well-understood, routine, and conventional activity (see MPEP 2106.05 (d) II. iv)
Regarding claim 12,
Claim 12 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 11 which includes an abstract idea (see rejection for claim 11). The additional limitations:
wherein the sub-threshold result comprises a greater difference between the output of the artificial neural network for the representative dataset and the known output than a predefined threshold. This claim merely recites a further limitation on the retrain the artificial neural network at least partially in response to a sub-threshold result of the periodic performance evaluation from Claim 9 which was directed to insignificant extra-solution activity (see MPEP 2106.05(g)), and also directed to well-understood, routine and conventional as evidenced by Huang et al. (US 20210271809 A1) ([0091]: “For example, in the classification task, features of objects to be labelled passing the verification can be used as features of the training samples and the manual labelling results as labels of the training samples to generate the training samples to retrain or incrementally train and the machine learning model. The retraining or incremental training process of the model is well known in the art, which is not repeated here.”)
Regarding claim 13,
Claim 13 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 12 which includes an abstract idea (see rejection for claim 12). The additional limitations:
wherein the controller is configured to – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
apply the predefined threshold irrespective of a quantity of bit errors present in the memory device. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
Regarding claim 14,
Claim 14 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 9 which includes an abstract idea (see rejection for claim 9). The additional limitations:
wherein the controller is configured to – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
retrain the artificial neural network irrespective of a quantity of bit errors present in the memory device. This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)), and also directed to well-understood, routine and conventional as evidenced by Huang et al. (US 20210271809 A1) ([0091]: “For example, in the classification task, features of objects to be labelled passing the verification can be used as features of the training samples and the manual labelling results as labels of the training samples to generate the training samples to retrain or incrementally train and the machine learning model. The retraining or incremental training process of the model is well known in the art, which is not repeated here.”)
Regarding claim 15,
Claim 15 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 9 which includes an abstract idea (see rejection for claim 9). The additional limitations:
wherein the controller is configured to – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
detect a plurality of bit errors in the memory device; and This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
wherein the controller being configured to – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
retrain the artificial neural network comprises This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)), and also directed to well-understood, routine and conventional as evidenced by Huang et al. (US 20210271809 A1) ([0091]: “For example, in the classification task, features of objects to be labelled passing the verification can be used as features of the training samples and the manual labelling results as labels of the training samples to generate the training samples to retrain or incrementally train and the machine learning model. The retraining or incremental training process of the model is well known in the art, which is not repeated here.”)
the controller being configured to – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
map weights to memory cells of the memory device without adjusting weights mapped to particular memory cells corresponding to the bit errors. This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
Regarding Claim 16,
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is a process.
Step 2A – Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, the claim recites an abstract idea.
comparing an output of the artificial neural network to a known output for the representative dataset to perform the periodic performance evaluation post-deployment of the remote sensor system at a frequency based on the second definition; and - This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
determine a difference between the output and the known output; - This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
Step 2A – Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
A non-transitory machine-readable medium storing machine-readable instructions, which when executed by a machine, cause the machine to This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
receive a first definition of a threshold for results of a periodic performance evaluation of an artificial neural network; This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)).
receive a second definition of at least one of a reliability expectation and a lifetime expectation of a memory device; This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)).
operate an artificial neural network allocated to the memory device post-deployment of a remote sensor system including the memory device, a sensor, and a controller coupled to the memory device and the sensor – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
wherein the instructions to operate the artificial neural network include instructions to perform operations with respect to data from the sensor, – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
the operations comprising one or more classification, identification, and analysis of the data from the sensor; – This limitation is directed to field of use (see MPEP 2106.05(h))
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure– This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
input a representative dataset to the artificial neural network This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)).
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
retrain the artificial neural network in-situ post-deployment of the remote system at least partially in response to a particular periodic performance evaluation not meeting the first definition. This limitation is directed to insignificant extra-solution activity (see MPEP 2106.05(g)).
operating the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
No, there are no additional elements that amount to significantly more than the judicial exception. The additional elements are:
A non-transitory machine-readable medium storing machine-readable instructions, which when executed by a machine, cause the machine to This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
receive a first definition of a threshold for results of a periodic performance evaluation of an artificial neural network; – This limitation is directed to receiving or transmitting data over a network. The courts have recognized receiving or transmitting data over a network as well understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity (see MPEP 2106.05(d) II.).
receive a second definition of at least one of a reliability expectation and a lifetime expectation of a memory device; – This limitation is directed to receiving or transmitting data over a network. The courts have recognized receiving or transmitting data over a network as well understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity (see MPEP 2106.05(d) II.).
operate an artificial neural network allocated to the memory device post-deployment of a remote sensor system including the memory device, a sensor, and a controller coupled to the memory device and the sensor – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
wherein the instructions to operate the artificial neural network include instructions to perform operations with respect to data from the sensor, – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
the operations comprising one or more classification, identification, and analysis of the data from the sensor; – This limitation is directed to field of use (see MPEP 2106.05(h))
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure– This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
input a representative dataset to the artificial neural network – This limitation is directed to receiving or transmitting data over a network. The courts have recognized receiving or transmitting data over a network as well understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity (see MPEP 2106.05(d) II.).
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
retrain the artificial neural network in-situ post-deployment of the remote system at least partially in response to a particular periodic performance evaluation not meeting the first definition. This limitation is also directed to well-understood, routine and conventional as evidenced by Huang et al. (US 20210271809 A1) ([0091]: “For example, in the classification task, features of objects to be labelled passing the verification can be used as features of the training samples and the manual labelling results as labels of the training samples to generate the training samples to retrain or incrementally train and the machine learning model. The retraining or incremental training process of the model is well known in the art, which is not repeated here.”)
operating the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the exception into a practical application.
Regarding claim 17,
Claim 17 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 16 which includes an abstract idea (see rejection for claim 16). The additional limitations:
further storing a table comprising correspondences between respective frequencies of the periodic performance evaluation and respective reliability expectations and respective lifetime expectations; and – This limitation is directed to storing and retrieving information in memory, which the court have recognized as well-understood, routine, and conventional activity (see MPEP 2106.05 (d) II. iv)
further storing instructions to select the frequency from the table based on the second definition. – This limitation is directed to storing and retrieving information in memory, which the court have recognized as well-understood, routine, and conventional activity (see MPEP 2106.05 (d) II. iv)
Regarding claim 18,
Claim 18 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 16 which includes an abstract idea (see rejection for claim 16). The additional limitations:
further comprising instructions to – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
detect a plurality of bit errors in the memory device; and - This limitation is directed to the abstract idea of a mental process (including an observation, evaluation, judgement, opinion) which can be performed in the human mind, or by a human using pen and paper (see MPEP 2106.04(a)(2) Ill. C.)
wherein the instructions to – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
retrain the artificial neural network comprise This limitation is also directed to well-understood, routine and conventional as evidenced by Huang et al. (US 20210271809 A1) ([0091]: “For example, in the classification task, features of objects to be labelled passing the verification can be used as features of the training samples and the manual labelling results as labels of the training samples to generate the training samples to retrain or incrementally train and the machine learning model. The retraining or incremental training process of the model is well known in the art, which is not repeated here.”)
instructions to – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
retrain the artificial neural network irrespective of the plurality of errors. This limitation is also directed to well-understood, routine and conventional as evidenced by Huang et al. (US 20210271809 A1) ([0091]: “For example, in the classification task, features of objects to be labelled passing the verification can be used as features of the training samples and the manual labelling results as labels of the training samples to generate the training samples to retrain or incrementally train and the machine learning model. The retraining or incremental training process of the model is well known in the art, which is not repeated here.”)
Regarding claim 19,
Claim 19 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 16 which includes an abstract idea (see rejection for claim 16). The additional limitations:
further storing [[a]] the representative dataset and [[a]] the known output of the artificial neural network for the representative dataset; and This limitation is directed to storing and retrieving information in memory, which the court have recognized as well-understood, routine, and conventional activity (see MPEP 2106.05 (d) II. iv)
Regarding claim 20,
Claim 20 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 16 which includes an abstract idea (see rejection for claim 16). The additional limitations:
cause weights of the artificial neural network to be stored in a plurality of memory cells of the memory device prior to operation of the artificial neural network; and This limitation is directed to storing and retrieving information in memory, which the court have recognized as well-understood, routine, and conventional activity (see MPEP 2106.05 (d) II. iv)
cause weights of the retrained artificial neural network to be stored in the plurality of memory cells. This limitation is directed to storing and retrieving information in memory, which the court have recognized as well-understood, routine, and conventional activity (see MPEP 2106.05 (d) II. iv)
Regarding claim 21,
Claim 21 is rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 16 which includes an abstract idea (see rejection for claim 16). The additional limitations:
instructions to operate the retrained artificial neural network. – This limitation is directed to a computer merely used as a tool to perform an existing process (see MPEP 2106.05(f) (2)).
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.
Claim(s) 1, 2, 4, 6-9, 11-13, 15-16, 19-21 are rejected under 35 U.S.C. 103 as being unpatentable over Qin et al. (US 2019/0073259 A1) (hereafter referred to as “Qin”) in view of Abeysooriya et al. (US9336483) (hereafter referred to as “Abeysooriya”), Mizuno et al. (5,577,166) (hereafter referred to as “Mizuno”) and further in view of Turcot et al (US 2021/0125065 A1) (hereafter referred to as “Turcot”) and Miserendino et al (US 10,121,108 B2) (hereafter referred to as “Miserendino”)
Regarding Claim 1:
Qin explicitly discloses:
operating an artificial neural network allocated to memory post-deployment of a remote system including the memory and a controller coupled to the memory, wherein the controller includes retraining circuitry; (Qin, Abstract: “The processor is configured to construct the neural network based on a structure of the neural network and a subset of the weights stored by the plurality of memory cells.”, ¶[0024]: “The method further includes for the weight data related to the neural network, bypassing error correction encoding and generating Cyclic Redundancy Check (CRC) data. The method further includes storing the weight data and the CRC data into the nonvolatile memory.”, ¶[0025]: “The weight value may be represented as a binary array, and estimating whether the weight value includes an error may include detecting an odd number of bit flips in the weight value. The weight values may be represented as binary arrays, and the CRC data may include a respective check bit appended to each weight value”) [Examiner’s note: The fact that the weight values of the neural network being stored in the memory array is being interpreted as the ANN being allocated to the memory]
retraining, in-situ by the retraining circuitry post-deployment of the remote system, the artificial neural network at least partially in response to the evaluation yielding a sub-threshold result. (Qin, [0073]: “If the neural network constructor 220 detects faulty cells based on the indicators, the neural network constructor 220 determines a number of faulty cells in operation 540… If the number of faulty cells is larger than the predetermined threshold, the neural network constructor 220 causes the neural network trainer 210 to retrain the neural network 135 (or a portion of the neural network 135) in operation 545, and the neural network constructor 220 constructs the neural network 135 according to the retrained weight values in operation 560.”) [Examiner’s note: The fact that a portion of the neural network is retrained in operation is being interpreted as the partially retraining of the neural network in-situ post-deployment]
wherein the method is performed by the controller (Qin, ¶[0021]: “Various embodiments disclosed herein are related to a method of storing and constructing a neural network.”, ¶[0059]: “The neural network constructor 220 is a component that constructs the neural network 135. The neural network constructor 220 may retrieve structure data from the structure storage 140 and weight data from the weight storage 150, and construct the neural network 135 based on the structure data and the weight data retrieved.”) [Examiner’s note: “a controller” is being interpreted as the “neural network constructor” in this context because the constructor performs the evaluation function which aligns with the definition of a controller in ¶0022] of the Instant Specification.
Qin fails to disclose:
a remote sensor system including a sensor, the memory, and a controller coupled to the memory, wherein the controller includes retraining circuitry;
evaluating performance of the artificial neural network post-deployment of the remote system;
wherein evaluating comprises inputting a representative dataset to the artificial neural network and comparing an output of the artificial neural network to a known output for the representative dataset; and
wherein operating the artificial neural network includes performing operations with respect to data from the sensor, the operations comprising one or more of classification, identification, and analysis of the data from the sensor
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure.
wherein retraining comprises: determining a difference between the output and the known output;
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network
operating the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system
However, Abeysooriya explicitly discloses:
evaluating performance of the artificial neural network post-deployment of the remote system; (Abeysooriya, Col. 1, Lines 52-57: “After the neural network is deployed and used by a content distribution network, additional input and output data associated with the content distribution network may be received and used to continuously or periodically evaluate the performance level of the neural network”, Col. 23, Lines 44-51: “At periodic intervals, the content management server 102 may transmit the most recent batch of predictive analysis input data and observed results to the neural network management server 610. In other embodiments, the predictive analysis data received in step 901 may be individual data records, each including a set of inputs and one or more outputs, that may be collected and used to evaluate the neural network 612 in real-time or near real-time.”)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Abeysooriya. Qin teaches training a neural network within the memory storage of a device. Abeysooriya teaches continuously or periodically evaluating the performance of an artificial neural network after being deployed by using additional input and output data associated with the content distribution network. One of ordinary skill would have motivation to combine Qin and Abeysooriya to enable timely corrective actions such as retraining, fine-tuning or model replacement. This proactive approach ensures that the deployed ANN continues to meet accuracy, efficiency, and robustness requirements while minimizing the risk of costly errors or downtime (Abeysooriya, Col. 22, Lines 35-41: “Additionally, step 804 may optional in some embodiments, but when performed, may potentially improve the speed and efficiency of the system when subsequent processes are used to evaluate the primary neural network data structure 612 and/or to generate and train potential replacement candidate neural networks, as described below in FIG. 9”)
However, Mizuno explicitly discloses:
wherein evaluating comprises inputting a representative dataset to the artificial neural network and comparing an output of the artificial neural network to a known output for the representative dataset; and (Mizuno, Col. 1, Lines 51-62: “In FIG. 1, when the input pattern 18 includes m items or elements Xi (i=1, 2, ... , m), the input pattern 18 can be represented as a vector X… Assume the input pattern 18 to be classified into any one of n categories. The output pattern 19, as the result of classification, is then expressed as a vector Y including n elements Yi (i=1, 2, ... , n).”, and Col. 2, Lines 1-11: “In the training of the neural network 11, the input pattern, namely, the vector X is supplied to the neural network 11 so as to cause the neural network 11 to compute the output pattern, namely, the vector Y. In this situation, a teacher pattern (correct output pattern) T is instructed as a desirable output pattern for the input pattern. Like the vector Y, the pattern T also includes n elements T1 to Tn. In response to the instructed pattern T, the neural network 11 corrects weights of the respective arcs to minimize the difference between the current output pattern Y and the desired output pattern T.”) [The “representative dataset” of the input is “vector X”. The highlighted indicates the process of adjusting weights based on the comparison of the neural network’s output and the known correct output (i.e., the teacher pattern output T)]
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Mizuno. Qin teaches training a neural network within the memory storage of a device. Mizuno teaches evaluating neural network’s performance and retraining the network. One of ordinary skill would have motivation to combine Qin and Mizuno to improve the re-training process of the neural network by automatically accomplish the re-training process to the user whenever the average error exceeds the threshold value (Mizuno, Col. 9, Lines 14-29)
However, Miserendino explicitly discloses:
operating the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system (Miserendino, Col. 8, Lines 51-58: “In-situ training allows each instance of a sensor to tailor itself on data not available to anyone but the local user; this method effectively guarantees all in-situ trained classifier models are unique. In other words, the set of all malware detection models is heterogeneous instead of homogeneous. Bad actors will no longer be able to rely on pre-testing their malware and incur greater risk of discovery across the community of users.”, Col. 9, Lines 3-11: “The implementation of in-situ learning described here is based on a combination of third-party and in-situ datasets allowing the user the benefits of tailoring without requiring the user to release data to the third-party. Due to the blending of datasets and a tightly controlled and automated machine learning process, the user is less prone to unintentional errors introduced by poor machine learning methods that could result in poor performance.”)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Miserendino. . Qin teaches training a neural network within the memory storage of a device. Miserendino teaches system and method for in-situ classifier retraining for malware identification and model heterogeneity. One of ordinary skill would have motivation to combine Qin and Miserendino because MPEP 2143 sets forth the Supreme Court rationales for obviousness including: (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results; (E): “Obvious to try” choosing from a finite number of identified, predictable solutions, with a reasonable expectation of success; (F) Known work in one field of endeavor may prompt variations of it for use in either the same field or a different one based on design incentives or other market forces if the variations are predictable to one of the ordinary skill in the art.
However, Turcot explicitly discloses:
wherein operating the artificial neural network includes performing operations with respect to data from the sensor, the operations comprising one or more of classification, identification, and analysis of the data from the sensor (Turcot, ¶[0005]: “The deep learning neural network can be trained to analyze the obtained data and to identify human perception states. The deep learning neural network can be adapted or "retrained", as more data is analyzed by the network, to speed operation, to improve convergence, and so on. The deep learning network can be retrained using in situ retraining.”, ¶[0033]: “The human perception state analysis is based on obtaining images that include facial data from the individual. The images can include video images, still images, intermittently obtained images, and so on. The images can include visible light images, near-infrared light images, etc”, ¶[0082]: “In other embodiments, the wearable device is another device, such as an earpiece with a camera, a helmet or hat with a camera, a clip-on camera attached to clothing, or any other type of wearable device with a camera or other sensor for collecting expression data.”)
a remote sensor system including a sensor, the memory, and a controller coupled to the memory and the retraining circuitry; (Turcot, ¶[0078]: “programmable apparatus which executes any of the above-mentioned computer program products or computer implemented methods may include one or more microprocessors, microcontrollers, embedded microcontrollers, programmable digital signal processors, programmable devices, programmable gate arrays, programmable array logic, memory devices, application specific integrated circuits, or the like. Each may be suitably employed or configured to process computer program instructions, execute computer logic, store computer data, and so on.”)
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure. (Turcot, ¶[0078]: “The sensor data can be obtained from the vehicle occupant along with the video data or the audio data, instead of the video data or the audio data, etc. In embodiments, the sensor data can include one or more of vehicle temperature, outside temperature, time of day, level of daylight, weather conditions, headlight activation, windshield wiper activation, entertainment center selection, or entertainment center volume.”, ¶[0079]: “Images, which can include facial or torso data, human perception state data, audio data, and physiological data, can be collected using multiple mobile devices. The image data can be applied to neural network training, where the neural network training can enable deep learning. The deep learning can include in situ retraining.”, and ¶[0081]: “A mobile device can include a front-side camera and/or a back-side camera that can be used to collect expression data. A mobile device can include a microphone, audio transducer, or other audio capture apparatus that can be used to capture the speech and nonspeech vocalizations. Sources of expression data can include a webcam 1222, a phone camera 1242, a tablet camera 1252, a wearable camera 1262, and a mobile camera 1230. A wearable camera can comprise various camera devices, such as a watch camera 1272. Sources of audio data 1282 can include a microphone 1280.”. Accordingly, the 103 rejections are maintained.)
wherein retraining comprises: determining a difference between the output and the known output; (Turcot, ¶[0045]: “The training includes analyzing the training data and comparing the analysis results to the known good or expected results. When the analysis results differ from the known good results, weights can be readjusted or retrained until the analysis of the second set of training data yields the expected results.”)
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network. (Turcot, ¶[0047]: “The flow 200 further includes backward propagation from the additional nodes 230. Backward propagation or "backpropagation" can be used to update weights, biases, or other values associated with nodes within layers of a neural network such as a deep learning neural network. In backpropagation, weight values can be iteratively and recursively updated based on a function, an algorithm, a heuristic, and so on. In embodiments, the backpropagation can be based on an algorithm such as a gradient-based algorithm for optimization.”, ¶[0048]: “Discussed throughout, the training, retraining, or pruning of weights, etc., can take place on a server device. A server device can include a local server, a remote server, a cloud server, a mesh server, and so on… The flow 200 further includes modifying the set of weights 252 that were trained based on the additional set of weights and the topology. The modifying the weights can include updating the weight values, retraining the weight values, and so on.”)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Turcot. Qin teaches training a neural network within the memory storage of a device. Turcot teaches a method for deep learning in-situ retraining. One of ordinary skill would have motivation to combine Qin and Turcot to maintain model accuracy and adaptability in dynamic or remote environments where external access is limited or unavailable. This self-correcting mechanism ensures that the neural network remains robust and reliable over time, without the need for manual intervention or re-deployment.
Regarding Claim 2, the combination of Qin, Abeysooriya, Miserendino, Turcot and Mizuno discloses all the limitations of Claim 1 (as shown in the rejection above).
Mizuno further discloses:
evaluating performance of the artificial neural network after retraining; and (Mizuno, Figure 4:
PNG
media_image1.png
741
566
media_image1.png
Greyscale
) [Blocks 38, 39 and 310 in Figure 4 assess whether the average error has increased and decide whether to continue re-training or not. If the average error has not increased, it applies satisfactory performance, hence no explicit performance evaluation step after retraining is shown in the figure]
determining whether to retrain the artificial neural network again based on results of the evaluation. (Mizuno, Col. 6, Lines 47-53: “Furthermore, the average error, which has been kept retained from the classification start point up to the current point, is updated (step 37). When the average error exceeds a preset constant (step 38), the monitored result 112 of execution history is presented on the screen of the display 24 to notify the increase in the average error to the user. The control processing module 15 requests the user to input an instruction so as to determine whether or not a re-training of the neural network 11 is to be executed.”)
Regarding Claim 4, the combination of Qin, Miserendino, Turcot, Abeysooriya and Mizuno discloses all the limitations of Claim 1 (as shown in the rejection above).
Qin in view of Abeysooriya, Miserendino, Turcot and Mizuno further discloses:
wherein evaluating comprises periodically evaluating performance of the artificial neural network. (Abeysooriya, Col. 1, Lines 52-57: “After the neural network is deployed and used by a content distribution network, additional input and output data associated with the content distribution network may be received and used to continuously or periodically evaluate the performance level of the neural network”)
Regarding Claim 6, the combination of Qin, Miserendino, Turcot, Abeysooriya and Mizuno discloses all the limitations of Claim 1 (as shown in the rejection above).
Qin in view of Mizuno, Miserendino, Turcot and Abeysooriya further discloses:
detecting an error in the memory; and (Qin, [0041]: “Disclosed is a device for storing and constructing neural network by detecting weight values (also referred to as "weights" herein) of the neural network associated with faulty cells.”) [The examiner interprets that the “faulty cells” here is the “error in the memory”]
wherein retraining further comprises retraining at least partially in response to the error. (Qin, [0061]: “Hence, if the number of faulty cells is more the predetermined threshold, the neural network constructor 220 may cause the neural network trainer 210 to retrain the neural network 135 or a portion of the neural network 135, and construct the neural network 135 after retraining.”) [The examiner interprets that the “faulty cells” here is the “error in the memory”]
Regarding Claim 7, the combination of Qin, Miserendino, Turcot, Abeysooriya and Mizuno discloses all the limitations of Claim 1 (as shown in the rejection above).
Qin in view of Mizuno, Miserendino, Turcot and Abeysooriya further discloses:
wherein retraining accounts for the error without remapping storage of weights in the memory. (Qin, [0073]: “If the number of faulty cells is less than a predetermined threshold, the neural network constructor 220 performs weight substitution by applying a default value ( e.g., zero) for weight values stored by the faulty cells in operation 550.”) [The examiner interprets the “weight substitution” here as adjusting weight values without changing their location in the memory (i.e., without re-mapping the storage of weights in the memory.)]
Regarding Claim 8, the combination of Qin, Miserendino, Turcot, Abeysooriya and Mizuno discloses all the limitations of Claim 6 (as shown in the rejection above).
Qin in view of Mizuno, Miserendino, Turcot and Abeysooriya further discloses:
wherein detecting the error comprises detecting a bit error or bit freeze in the memory. (Qin, [0057]: “If any odd number of bits is flipped from their target values, the CRC detector 230 determines that the vector of memory cells stores an incorrect weight value. Because each bit error may occur with a small probability, if any error exists in the vector of memory cells, the vector of memory cells will likely include a single faulty cell.”)
Regarding Claim 9, Qin explicitly discloses:
An apparatus, comprising: a memory device; and (Qin, [0014]: “In one or more embodiments, the structure of the neural network is stored by a first memory device, and the weights are stored by the plurality of memory cells of a second memory device different from the first memory device.”)
a controller coupled to the memory device and configured to: (Qin, Abstract: “The device further includes a processor coupled to the plurality of memory cells.”, ¶[0055]: “In one embodiment, the processor 130 forms a neural network trainer 210, a neural network constructor 220, a CRC detector 230, a stuck-at-fault cell detector 240, and a feature detector 250.”, ¶[0059]: “In one approach, the neural network constructor 220 may determine faulty cells storing incorrect weight values, and perform weight substitution or weight elimination based on the determined faulty cells before constructing the neural network 135.) [Examiner’s note: “a controller” is being interpreted as the “neural network constructor” in this context because the constructor performs the evaluation function which aligns with the definition of a controller in ¶0022] of the Instant Specification. The processor here is coupled to memory cells and forms a neural network constructor, that means the constructor is also coupled to the memory]
wherein the artificial neural network is allocated to the memory device; and (Qin, Abstract: “The processor is configured to construct the neural network based on a structure of the neural network and a subset of the weights stored by the plurality of memory cells.”)
retrain the artificial neural network in-situ post-deployment of the remote system at least partially in response to a sub-threshold result of the periodic performance evaluation. (Qin, [0073]: “If the neural network constructor 220 detects faulty cells based on the indicators, the neural network constructor 220 determines a number of faulty cells in operation 540… If the number of faulty cells is larger than the predetermined threshold, the neural network constructor 220 causes the neural network trainer 210 to retrain the neural network 135 (or a portion of the neural network 135) in operation 545, and the neural network constructor 220 constructs the neural network 135 according to the retrained weight values in operation 560.”) [Examiner’s note: The fact that a portion of the neural network is retrained in operation is being interpreted as the partially retraining of the neural network in-situ post-deployment]
Qin fails to disclose:
A remote sensor system not accessible by technicians or other computing infrastructure, comprising: a sensor
operate an artificial neural network allocated to the memory device post-deployment of the remote sensor system, wherein operating the artificial neural network includes performing operations with respect to data from the sensor, the operations comprising one or more of classification, identification, and analysis of the data from the sensor;
input a representative dataset to the artificial neural network and comparing an output of the artificial neural network to a known output for the representative dataset
periodically evaluate performance of an artificial neural network post-deployment of the remote system at a predefined frequency
determining a difference between the output and the known output;
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network
operate the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system;
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure
However, Mizuno explicitly discloses:
wherein evaluating comprises inputting a representative dataset to the artificial neural network and comparing an output of the artificial neural network to a known output for the representative dataset; and (Mizuno, Col. 1, Lines 51-62: “In FIG. 1, when the input pattern 18 includes m items or elements Xi (i=1, 2, ... , m), the input pattern 18 can be represented as a vector X… Assume the input pattern 18 to be classified into any one of n categories. The output pattern 19, as the result of classification, is then expressed as a vector Y including n elements Yi (i=1, 2, ... , n).”, and Col. 2, Lines 1-11: “In the training of the neural network 11, the input pattern, namely, the vector X is supplied to the neural network 11 so as to cause the neural network 11 to compute the output pattern, namely, the vector Y. In this situation, a teacher pattern (correct output pattern) T is instructed as a desirable output pattern for the input pattern. Like the vector Y, the pattern T also includes n elements T1 to Tn. In response to the instructed pattern T, the neural network 11 corrects weights of the respective arcs to minimize the difference between the current output pattern Y and the desired output pattern T.”) [The “representative dataset” of the input is “vector X”. The highlighted indicates the process of adjusting weights based on the comparison of the neural network’s output and the known correct output (i.e., the teacher pattern output T)]
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Mizuno. Qin teaches training a neural network within the memory storage of a device. Mizuno teaches evaluating neural network’s performance and retraining the network. One of ordinary skill would have motivation to combine Qin and Mizuno to improve the re-training process of the neural network by automatically accomplish the re-training process to the user whenever the average error exceeds the threshold value (Mizuno, Col. 9, Lines 14-29)
However, Abeysooriya explicitly discloses:
periodically evaluate performance of an artificial neural network post-deployment of the remote sensor system at a predefined frequency (Abeysooriya, Col. 1, Lines 52-57: “After the neural network is deployed and used by a content distribution network, additional input and output data associated with the content distribution network may be received and used to continuously or periodically evaluate the performance level of the neural network”, Col. 23, Lines 44-51: “At periodic intervals, the content management server 102 may transmit the most recent batch of predictive analysis input data and observed results to the neural network management server 610. In other embodiments, the predictive analysis data received in step 901 may be individual data records, each including a set of inputs and one or more outputs, that may be collected and used to evaluate the neural network 612 in real-time or near real-time.”)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Abeysooriya. Qin teaches training a neural network within the memory storage of a device. Abeysooriya teaches continuously or periodically evaluating the performance of an artificial neural network after being deployed by using additional input and output data associated with the content distribution network. One of ordinary skill would have motivation to combine Qin and Abeysooriya to enable timely corrective actions such as retraining, fine-tuning or model replacement. This proactive approach ensures that the deployed ANN continues to meet accuracy, efficiency, and robustness requirements while minimizing the risk of costly errors or downtime (Abeysooriya, Col. 22, Lines 35-41: “Additionally, step 804 may optional in some embodiments, but when performed, may potentially improve the speed and efficiency of the system when subsequent processes are used to evaluate the primary neural network data structure 612 and/or to generate and train potential replacement candidate neural networks, as described below in FIG. 9”)
However, Miserendino explicitly discloses:
operate the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system (Miserendino, Col. 8, Lines 51-58: “In-situ training allows each instance of a sensor to tailor itself on data not available to anyone but the local user; this method effectively guarantees all in-situ trained classifier models are unique. In other words, the set of all malware detection models is heterogeneous instead of homogeneous. Bad actors will no longer be able to rely on pre-testing their malware and incur greater risk of discovery across the community of users.”, Col. 9, Lines 3-11: “The implementation of in-situ learning described here is based on a combination of third-party and in-situ datasets allowing the user the benefits of tailoring without requiring the user to release data to the third-party. Due to the blending of datasets and a tightly controlled and automated machine learning process, the user is less prone to unintentional errors introduced by poor machine learning methods that could result in poor performance.”)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Miserendino. . Qin teaches training a neural network within the memory storage of a device. Miserendino teaches system and method for in-situ classifier retraining for malware identification and model heterogeneity. One of ordinary skill would have motivation to combine Qin and Miserendino because MPEP 2143 sets forth the Supreme Court rationales for obviousness including: (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results; (E): “Obvious to try” choosing from a finite number of identified, predictable solutions, with a reasonable expectation of success; (F) Known work in one field of endeavor may prompt variations of it for use in either the same field or a different one based on design incentives or other market forces if the variations are predictable to one of the ordinary skill in the art.
However, Turcot explicitly discloses:
a remote sensor system including a sensor, the memory, and a controller coupled to the memory and the retraining circuitry; (Turcot, ¶[0078]: “programmable apparatus which executes any of the above-mentioned computer program products or computer implemented methods may include one or more microprocessors, microcontrollers, embedded microcontrollers, programmable digital signal processors, programmable devices, programmable gate arrays, programmable array logic, memory devices, application specific integrated circuits, or the like. Each may be suitably employed or configured to process computer program instructions, execute computer logic, store computer data, and so on.”)
operate an artificial neural network allocated to the memory device post-deployment of the remote sensor system, wherein operating the artificial neural network includes performing operations with respect to data from the sensor, the operations comprising one or more of classification, identification, and analysis of the data from the sensor; (Turcot, ¶[0005]: “The deep learning neural network can be trained to analyze the obtained data and to identify human perception states. The deep learning neural network can be adapted or "retrained", as more data is analyzed by the network, to speed operation, to improve convergence, and so on. The deep learning network can be retrained using in situ retraining.”, ¶[0033]: “The human perception state analysis is based on obtaining images that include facial data from the individual. The images can include video images, still images, intermittently obtained images, and so on. The images can include visible light images, near-infrared light images, etc”, ¶[0082]: “In other embodiments, the wearable device is another device, such as an earpiece with a camera, a helmet or hat with a camera, a clip-on camera attached to clothing, or any other type of wearable device with a camera or other sensor for collecting expression data.”)
wherein retraining comprises: determining a difference between the output and the known output; (Turcot, ¶[0045]: “The training includes analyzing the training data and comparing the analysis results to the known good or expected results. When the analysis results differ from the known good results, weights can be readjusted or retrained until the analysis of the second set of training data yields the expected results.”)
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network. (Turcot, ¶[0047]: “The flow 200 further includes backward propagation from the additional nodes 230. Backward propagation or "backpropagation" can be used to update weights, biases, or other values associated with nodes within layers of a neural network such as a deep learning neural network. In backpropagation, weight values can be iteratively and recursively updated based on a function, an algorithm, a heuristic, and so on. In embodiments, the backpropagation can be based on an algorithm such as a gradient-based algorithm for optimization.”, ¶[0048]: “Discussed throughout, the training, retraining, or pruning of weights, etc., can take place on a server device. A server device can include a local server, a remote server, a cloud server, a mesh server, and so on… The flow 200 further includes modifying the set of weights 252 that were trained based on the additional set of weights and the topology. The modifying the weights can include updating the weight values, retraining the weight values, and so on.”)
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure. (Turcot, ¶[0078]: “The sensor data can be obtained from the vehicle occupant along with the video data or the audio data, instead of the video data or the audio data, etc. In embodiments, the sensor data can include one or more of vehicle temperature, outside temperature, time of day, level of daylight, weather conditions, headlight activation, windshield wiper activation, entertainment center selection, or entertainment center volume.”, ¶[0079]: “Images, which can include facial or torso data, human perception state data, audio data, and physiological data, can be collected using multiple mobile devices. The image data can be applied to neural network training, where the neural network training can enable deep learning. The deep learning can include in situ retraining.”, and ¶[0081]: “A mobile device can include a front-side camera and/or a back-side camera that can be used to collect expression data. A mobile device can include a microphone, audio transducer, or other audio capture apparatus that can be used to capture the speech and nonspeech vocalizations. Sources of expression data can include a webcam 1222, a phone camera 1242, a tablet camera 1252, a wearable camera 1262, and a mobile camera 1230. A wearable camera can comprise various camera devices, such as a watch camera 1272. Sources of audio data 1282 can include a microphone 1280.”. Accordingly, the 103 rejections are maintained.)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Turcot. Qin teaches training a neural network within the memory storage of a device. Turcot teaches a method for deep learning in-situ retraining. One of ordinary skill would have motivation to combine Qin and Turcot to maintain model accuracy and adaptability in dynamic or remote environments where external access is limited or unavailable. This self-correcting mechanism ensures that the neural network remains robust and reliable over time, without the need for manual intervention or re-deployment.
Regarding Claim 11, the combination of Qin, Miserendino, Turcot and Abeysooriya discloses all the limitations of Claim 9 (as shown in the rejection above).
Qin in view of Abeysooriya, Mizuno, Miserendino and Turcot further discloses:
wherein the apparatus stores [[a]] the representative dataset and [[a]] the known output of the artificial neural network for the representative dataset (Mizuno, Col 6, Lines 43-47: “When the correct output pattern 111 is supplied via the pattern input device 23 (step 33), the execution history monitoring module 13 stores, in the execution history file 17, the input and output patterns beforehand kept therein and the received correct output pattern.”, Col. 1, Lines 51-62: “In FIG. 1, when the input pattern 18 includes m items or elements Xi (i=1, 2, ... , m), the input pattern 18 can be represented as a vector X… Assume the input pattern 18 to be classified into any one of n categories. The output pattern 19, as the result of classification, is then expressed as a vector Y including n elements Yi (i=1, 2, ... , n).”, and Col. 2, Lines 1-11: “In the training of the neural network 11, the input pattern, namely, the vector X is supplied to the neural network 11 so as to cause the neural network 11 to compute the output pattern, namely, the vector Y. In this situation, a teacher pattern (correct output pattern) T is instructed as a desirable output pattern for the input pattern.)
Regarding Claim 12, the combination of Qin, Miserendino, Turcot and Abeysooriya discloses all the limitations of Claim 11 (as shown in the rejection above).
Qin in view of Abeysooriya, Miserendino, Turcot further discloses:
wherein the sub-threshold result comprises a greater difference between the output of the artificial neural network for the representative dataset and the known output than a predefined threshold. (Mizuno, Col. 1, Lines 51-62: “In FIG. 1, when the input pattern 18 includes m items or elements Xi (i=1, 2, ... , m), the input pattern 18 can be represented as a vector X… Assume the input pattern 18 to be classified into any one of n categories. The output pattern 19, as the result of classification, is then expressed as a vector Y including n elements Yi (i=1, 2, ... , n).”, and Col. 2, Lines 1-11: “In the training of the neural network 11, the input pattern, namely, the vector X is supplied to the neural network 11 so as to cause the neural network 11 to compute the output pattern, namely, the vector Y. In this situation, a teacher pattern (correct output pattern) T is instructed as a desirable output pattern for the input pattern. Like the vector Y, the pattern T also includes n elements T1 to Tn. In response to the instructed pattern T, the neural network 11 corrects weights of the respective arcs to minimize the difference between the current output pattern Y and the desired output pattern T.”) [The “representative dataset” of the input is “vector X”. The highlighted indicates the process of adjusting weights based on the comparison of the neural network’s output and the known correct output (i.e., the teacher pattern output T)]
Regarding Claim 13, the combination of Qin, Miserendino, Turcot, Abeysooriya and Mizuno discloses all the limitations of Claim 12 (as shown in the rejection above)
Qin in view of Abeysooriya, Miserendino, Turcot and Mizuno further discloses:
wherein the controller is configured to apply the predefined threshold irrespective of a quantity of bit errors present in the memory device. (Qin, [0052]: “As temperature increases, the mobility of the ions also increases causing the programming threshold for the conductive bridge memory cell to decrease. Thus, the conductive bridge memory element may have a wide range of programming thresholds over temperature.”) [The highlighted indicates that the memory has a wide range of programming thresholds over temperature, which is irrespective of a quantity of bit errors present in the memory ]
Regarding Claim 15, the combination of Qin, Abeysooriya, Turcot, Miserendino and Mizuno discloses all the limitations of Claim 9 (as shown in the rejection above).
Qin in view of Miserendino, Mizuno, Turcot and Abeysooriya further discloses:
wherein the controller is configured to detect a plurality of bit errors in the memory device; and (Qin, [0041]: “Disclosed is a device for storing and constructing neural network by detecting weight values (also referred to as "weights" herein) of the neural network associated with faulty cells.”) [The examiner interprets that the “faulty cells” here is the “error in the memory”]
wherein the controller being configured to retrain the artificial neural network comprises the controller being configured to map weights to memory cells of the memory device without adjusting weights mapped to particular memory cells corresponding to the bit errors. (Qin, [0073]: “If the number of faulty cells is less than a predetermined threshold, the neural network constructor 220 performs weight substitution by applying a default value ( e.g., zero) for weight values stored by the faulty cells in operation 550.”) [The examiner interprets the “weight substitution” (i.e., map weights to memory cells) here as adjusting weight values without changing their location in the memory (i.e., without re-mapping the storage of weights in the memory.)]
Regarding Claim 16, Qin explicitly discloses:
A non-transitory machine-readable medium storing machine-readable instructions, which when executed by a machine, cause the machine to: (Qin, [0055]: “The processor 130 may execute instructions stored by a non-transitory computer readable medium to form these components as software modules.”)
receive a first definition of a threshold for results of a periodic performance evaluation of an artificial neural network; (Qin, [0008]: “In one or more embodiments, the processor is configured to perform a cyclic redundancy check on the plurality of memory cells, and detect the one or more faulty cells from the plurality of memory cells based on the cyclic redundancy check.”, [0061]: “In this example, if the neural network constructor 220 determines that less than 2.5% of memory cells are faulty cells, the neural network constructor 220 performs weight substitution or weight elimination on the faulty cells, and constructs the neural network 135 based on the weight substitution. On the other hand, if the neural network constructor 220 determines that more than 2.5% of memory cells are faulty cells, the neural network constructor 220 retrains the portion of the neural network 135 including the faulty cells or the entire neural network 135, and constructs the neural network 135 based on the retraining.) [The examiner interprets the “cyclic redundancy check” performed on the memory cells as the “periodic performance evaluation” of a neural network, and the “first definition of a threshold” here is the ”2.5%” figure as it is a corresponding threshold that triggers retraining process]
receive a second definition of at least one of a reliability expectation and a lifetime expectation of a memory device; (Qin, [0052]: “a conductive bridge memory element may include two solid metal electrodes, one relatively inert (e.g., tungsten) and the other electrochemically active (e.g., silver or copper), with a thin film of the solid electrolyte between the two electrodes. As temperature increases, the mobility of the ions also increases causing the programming threshold for the conductive bridge memory cell to decrease. Thus, the conductive bridge memory element may have a wide range of programming thresholds over temperature.”) [The ability of a memory cell to maintain data integrity and functionality over varying temperatures is a reliability feature, as it indicates resilience to one of the most common stressors for electronic components. Reliable operation in diverse temperature conditions also correlates with a longer functional lifespan of the memory.]
operate an artificial neural network allocated to the memory device post-deployment of a remote sensor system including the memory device and a controller coupled to the memory device; (Qin, Abstract: “The processor is configured to construct the neural network based on a structure of the neural network and a subset of the weights stored by the plurality of memory cells.”, ¶[0024]: “The method further includes for the weight data related to the neural network, bypassing error correction encoding and generating Cyclic Redundancy Check (CRC) data. The method further includes storing the weight data and the CRC data into the nonvolatile memory.”, ¶[0025]: “The weight value may be represented as a binary array, and estimating whether the weight value includes an error may include detecting an odd number of bit flips in the weight value. The weight values may be represented as binary arrays, and the CRC data may include a respective check bit appended to each weight value”) [Examiner’s note: The fact that the weight values of the neural network being stored in the memory array is being interpreted as the ANN being allocated to the memory]
retrain the artificial neural network in-situ post-deployment of the remote sensor system at least partially in response to a particular periodic performance evaluation not meeting the first definition. (Qin, [0073]: “If the neural network constructor 220 detects faulty cells based on the indicators, the neural network constructor 220 determines a number of faulty cells in operation 540… If the number of faulty cells is larger than the predetermined threshold, the neural network constructor 220 causes the neural network trainer 210 to retrain the neural network 135 (or a portion of the neural network 135) in operation 545, and the neural network constructor 220 constructs the neural network 135 according to the retrained weight values in operation 560.”) [Examiner’s note: The fact that a portion of the neural network is retrained in operation is being interpreted as the partially retraining of the neural network in-situ post-deployment]
Qin fails to disclose:
Wherein the instructions to operate the artificial neural network includes performing operations with respect to data from the sensor, the operations comprising one or more of classification, identification, and analysis of the data from the sensor
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure.
input a representative dataset to the artificial neural network and comparing an output of the artificial neural network to a known output for the representative dataset
perform the periodic performance evaluation post-deployment of the remote system at a frequency based on the second definition; and
determining a difference between the output and the known output;
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network
operating the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system
However, Abeysooriya explicitly discloses:
perform the periodic performance evaluation post-deployment of the remote system at a frequency based on the second definition; and (Abeysooriya, Col. 1, Lines 52-57: “After the neural network is deployed and used by a content distribution network, additional input and output data associated with the content distribution network may be received and used to continuously or periodically evaluate the performance level of the neural network”, Col. 23, Lines 44-51: “At periodic intervals, the content management server 102 may transmit the most recent batch of predictive analysis input data and observed results to the neural network management server 610. In other embodiments, the predictive analysis data received in step 901 may be individual data records, each including a set of inputs and one or more outputs, that may be collected and used to evaluate the neural network 612 in real-time or near real-time.”)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Abeysooriya. Qin teaches training a neural network within the memory storage of a device. Abeysooriya teaches continuously or periodically evaluating the performance of an artificial neural network after being deployed by using additional input and output data associated with the content distribution network. One of ordinary skill would have motivation to combine Qin and Abeysooriya to enable timely corrective actions such as retraining, fine-tuning or model replacement. This proactive approach ensures that the deployed ANN continues to meet accuracy, efficiency, and robustness requirements while minimizing the risk of costly errors or downtime (Abeysooriya, Col. 22, Lines 35-41: “Additionally, step 804 may optional in some embodiments, but when performed, may potentially improve the speed and efficiency of the system when subsequent processes are used to evaluate the primary neural network data structure 612 and/or to generate and train potential replacement candidate neural networks, as described below in FIG. 9”)
However, Mizuno explicitly discloses:
input a representative dataset to the artificial neural network and comparing an output of the artificial neural network to a known output for the representative dataset (Mizuno, Col. 1, Lines 51-62: “In FIG. 1, when the input pattern 18 includes m items or elements Xi (i=1, 2, ... , m), the input pattern 18 can be represented as a vector X… Assume the input pattern 18 to be classified into any one of n categories. The output pattern 19, as the result of classification, is then expressed as a vector Y including n elements Yi (i=1, 2, ... , n).”, and Col. 2, Lines 1-11: “In the training of the neural network 11, the input pattern, namely, the vector X is supplied to the neural network 11 so as to cause the neural network 11 to compute the output pattern, namely, the vector Y. In this situation, a teacher pattern (correct output pattern) T is instructed as a desirable output pattern for the input pattern. Like the vector Y, the pattern T also includes n elements T1 to Tn. In response to the instructed pattern T, the neural network 11 corrects weights of the respective arcs to minimize the difference between the current output pattern Y and the desired output pattern T.”) [The “representative dataset” of the input is “vector X”. The highlighted indicates the process of adjusting weights based on the comparison of the neural network’s output and the known correct output (i.e., the teacher pattern output T)]
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Mizuno. Qin teaches training a neural network within the memory storage of a device. Mizuno teaches evaluating neural network’s performance and retraining the network. One of ordinary skill would have motivation to combine Qin and Mizuno to improve the re-training process of the neural network by automatically accomplish the re-training process to the user whenever the average error exceeds the threshold value (Mizuno, Col. 9, Lines 14-29)
However, Miserendino explicitly discloses:
operate the retrained artificial neural network in the remote sensor system thereby improving a mean time to failure of the remote sensor system (Miserendino, Col. 8, Lines 51-58: “In-situ training allows each instance of a sensor to tailor itself on data not available to anyone but the local user; this method effectively guarantees all in-situ trained classifier models are unique. In other words, the set of all malware detection models is heterogeneous instead of homogeneous. Bad actors will no longer be able to rely on pre-testing their malware and incur greater risk of discovery across the community of users.”, Col. 9, Lines 3-11: “The implementation of in-situ learning described here is based on a combination of third-party and in-situ datasets allowing the user the benefits of tailoring without requiring the user to release data to the third-party. Due to the blending of datasets and a tightly controlled and automated machine learning process, the user is less prone to unintentional errors introduced by poor machine learning methods that could result in poor performance.”)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Miserendino. . Qin teaches training a neural network within the memory storage of a device. Miserendino teaches system and method for in-situ classifier retraining for malware identification and model heterogeneity. One of ordinary skill would have motivation to combine Qin and Miserendino because MPEP 2143 sets forth the Supreme Court rationales for obviousness including: (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results; (E): “Obvious to try” choosing from a finite number of identified, predictable solutions, with a reasonable expectation of success; (F) Known work in one field of endeavor may prompt variations of it for use in either the same field or a different one based on design incentives or other market forces if the variations are predictable to one of the ordinary skill in the art.
However, Turcot explicitly discloses:
a remote sensor system including a sensor, the memory, and a controller coupled to the memory and the retraining circuitry; (Turcot, ¶[0078]: “programmable apparatus which executes any of the above-mentioned computer program products or computer implemented methods may include one or more microprocessors, microcontrollers, embedded microcontrollers, programmable digital signal processors, programmable devices, programmable gate arrays, programmable array logic, memory devices, application specific integrated circuits, or the like. Each may be suitably employed or configured to process computer program instructions, execute computer logic, store computer data, and so on.”)
wherein the instructions to operate the artificial neural network includes performing operations with respect to data from the sensor, the operations comprising one or more of classification, identification, and analysis of the data from the sensor (Turcot, ¶[0005]: “The deep learning neural network can be trained to analyze the obtained data and to identify human perception states. The deep learning neural network can be adapted or "retrained", as more data is analyzed by the network, to speed operation, to improve convergence, and so on. The deep learning network can be retrained using in situ retraining.”, ¶[0033]: “The human perception state analysis is based on obtaining images that include facial data from the individual. The images can include video images, still images, intermittently obtained images, and so on. The images can include visible light images, near-infrared light images, etc”, ¶[0082]: “In other embodiments, the wearable device is another device, such as an earpiece with a camera, a helmet or hat with a camera, a clip-on camera attached to clothing, or any other type of wearable device with a camera or other sensor for collecting expression data.”)
wherein the remote sensor system is deployed in one of a 5G network, smart camera, radar, space exploration equipment, and subsea exploration equipment and therefore is not accessible by technicians or other computing infrastructure. (Turcot, ¶[0078]: “The sensor data can be obtained from the vehicle occupant along with the video data or the audio data, instead of the video data or the audio data, etc. In embodiments, the sensor data can include one or more of vehicle temperature, outside temperature, time of day, level of daylight, weather conditions, headlight activation, windshield wiper activation, entertainment center selection, or entertainment center volume.”, ¶[0079]: “Images, which can include facial or torso data, human perception state data, audio data, and physiological data, can be collected using multiple mobile devices. The image data can be applied to neural network training, where the neural network training can enable deep learning. The deep learning can include in situ retraining.”, and ¶[0081]: “A mobile device can include a front-side camera and/or a back-side camera that can be used to collect expression data. A mobile device can include a microphone, audio transducer, or other audio capture apparatus that can be used to capture the speech and nonspeech vocalizations. Sources of expression data can include a webcam 1222, a phone camera 1242, a tablet camera 1252, a wearable camera 1262, and a mobile camera 1230. A wearable camera can comprise various camera devices, such as a watch camera 1272. Sources of audio data 1282 can include a microphone 1280.”. Accordingly, the 103 rejections are maintained.)
wherein retraining comprises: determining a difference between the output and the known output; (Turcot, ¶[0045]: “The training includes analyzing the training data and comparing the analysis results to the known good or expected results. When the analysis results differ from the known good results, weights can be readjusted or retrained until the analysis of the second set of training data yields the expected results.”)
applying corrections to reduce the difference, including updating weights and/or biases of the artificial neural network. (Turcot, ¶[0047]: “The flow 200 further includes backward propagation from the additional nodes 230. Backward propagation or "backpropagation" can be used to update weights, biases, or other values associated with nodes within layers of a neural network such as a deep learning neural network. In backpropagation, weight values can be iteratively and recursively updated based on a function, an algorithm, a heuristic, and so on. In embodiments, the backpropagation can be based on an algorithm such as a gradient-based algorithm for optimization.”, ¶[0048]: “Discussed throughout, the training, retraining, or pruning of weights, etc., can take place on a server device. A server device can include a local server, a remote server, a cloud server, a mesh server, and so on… The flow 200 further includes modifying the set of weights 252 that were trained based on the additional set of weights and the topology. The modifying the weights can include updating the weight values, retraining the weight values, and so on.”)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin and Turcot. Qin teaches training a neural network within the memory storage of a device. Turcot teaches a method for deep learning in-situ retraining. One of ordinary skill would have motivation to combine Qin and Turcot to maintain model accuracy and adaptability in dynamic or remote environments where external access is limited or unavailable. This self-correcting mechanism ensures that the neural network remains robust and reliable over time, without the need for manual intervention or re-deployment.
Regarding Claim 19, the combination of Qin, Miserendino, Turcot and Abeysooriya explicitly discloses all the limitations of Claim 16 (as shown in the rejection above).
Qin in view of Abeysooriya, Miserendino, Turcot further discloses:
further storing [[a]] the representative dataset and [[a]] the known output of the artificial neural network for the representative dataset (Mizuno, Col 6, Lines 43-47: “When the correct output pattern 111 is supplied via the pattern input device 23 (step 33), the execution history monitoring module 13 stores, in the execution history file 17, the input and output patterns beforehand kept therein and the received correct output pattern.”, Col. 1, Lines 51-62: “In FIG. 1, when the input pattern 18 includes m items or elements Xi (i=1, 2, ... , m), the input pattern 18 can be represented as a vector X… Assume the input pattern 18 to be classified into any one of n categories. The output pattern 19, as the result of classification, is then expressed as a vector Y including n elements Yi (i=1, 2, ... , n).”, and Col. 2, Lines 1-11: “In the training of the neural network 11, the input pattern, namely, the vector X is supplied to the neural network 11 so as to cause the neural network 11 to compute the output pattern, namely, the vector Y. In this situation, a teacher pattern (correct output pattern) T is instructed as a desirable output pattern for the input pattern.)
Regarding Claim 20, the combination of Qin, Turcot, Miserendino, Abeysooriya and Mizuno discloses all the limitations of Claim 16 (as shown in the rejection above).
Qin in view of Turcot, Miserendino, Mizuno and Abeysooriya further discloses:
cause weights of the artificial neural network to be stored in a plurality of memory cells of the memory device prior to operation of the artificial neural network; and (Qin, [0005]: “The processor is configured to construct the neural network based on a structure of the neural network and a subset of the weights stored by the plurality of memory cells.”) [The highlighted indicates the process of constructing a neural network based on its structure and set of weights, which means this process must happen prior to the operation]
cause weights of the retrained artificial neural network to be stored in the plurality of memory cells. (Qin, [0011]: “In one or more embodiments, the processor is configured to retrain the another subset of the weights stored by the one or more memory cells and bypass retraining the subset of the weights stored by the subset of the plurality of memory cells, in response to detecting the one or more faulty cells.”)
Regarding Claim 21, the combination of Qin, Turcot, Miserendino, Abeysooriya and Mizuno discloses all the limitations of Claim 16 (as shown in the rejection above).
Qin in view of Turcot, Miserendino, Mizuno and Abeysooriya further discloses:
instructions to operate the retrained artificial neural network. (Qin, [0075]: “The CRC detector 230 reads the stored indicator and the vector of memory cells in operation 615, for example, in response to an instruction to construct the neural network 135 from the processor 130.”. [0073]: “… the neural network constructor 220 constructs the neural network 135 according to the retrained weight values in operation 560.”)
Claim(s) 5, 14 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Qin et al. (US 2019/0073259 A1) in view of Abeysooriya et al (US 9,336,483 B1) (hereafter referred to as “Abeysooriya”), Mizuno et al. (5,577,166) (hereafter referred to as “Mizuno”), Turcot et al (US 2021/0125065 A1) (hereafter referred to as “Turcot”) and Miserendino et al (US 10,121,108 B2) (hereafter referred to as “Miserendino”) and in further view of Sun et al. (US 11,080,152 B2)
Regarding Claim 5, the combination of Qin, Turcot, Miserendino, Abeysooriya and Mizuno discloses all the limitations of Claim 4 (as shown in the rejection above).
Qin in view of Abeysooriya, Turcot, Miserendino and Mizuno further discloses:
wherein periodically evaluating comprises periodically evaluating at the desired frequency.(Abeysooriya, Col. 23, Lines 44-51: “At periodic intervals, the content management server 102 may transmit the most recent batch of predictive analysis input data and observed results to the neural network management server 610. In other embodiments, the predictive analysis data received in step 901 may be individual data records, each including a set of inputs and one or more outputs, that may be collected and used to evaluate the neural network 612 in real-time or near real-time.”)
Qin in view of Abeysooriya, Turcot, Miserendino and Mizuno fails to disclose:
further comprising receiving an indication of a desired frequency of the periodic evaluation from a host of the memory; and
However, Sun explicitly discloses:
receiving an indication of a desired frequency of the periodic evaluation from a host of the memory; and (Sun, Col. 15, Lines 6-16: “For example, weights that may be modified more frequently may be stored in a more reliable type of memory with a weaker ECC to allow for faster access to those weights. In another example, file data (which may be less important than the weights of a neural network) and which may be modified less frequently may be stored in a less reliable type of memory with a stronger ECC. This may allow the controller 330 and/or data storage or memory device 320 to tune and/or optimize how different types of data are protected and/or how quickly different types of data may be accessed.”) [Examiner’s note: According to ¶[0010] of the Instant Specification, “an indication of a desired frequency of the periodic evaluation” is being interpreted as the reliability expectation of the memory implementing the neural network. The highlight indicates performing the process of evaluation at a desired frequency because a typical optimization process will require evaluation process.]
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin, Mizuno, Turcot, Miserendino, Abeysooriya and Sun. Qin teaches training a neural network within the memory storage of a device. Abeysooriya teaches continuously or periodically evaluating the performance of an artificial neural network after being deployed by using additional input and output data associated with the content distribution network. Turcot teaches a method for deep learning in-situ retraining. Miserendino teaches system and method for in-situ classifier retraining for malware identification and model heterogeneity. Mizuno teaches evaluating neural network’s performance and retraining the network. Sun teaches optimizing the neural network data organization. One of ordinary skill would have motivation to combine Qin, Abeysooriya, Turcot, Miserendino, Mizuno and Sun to allow the controller to access the metadata of a neural network more quickly and more efficiently (Sun, Col. 13, Lines 1-4)
Regarding Claim 14, the combination of Qin, Turcot, Miserendino, Mizuno and Abeysooriya discloses all the limitations of Claim 9 (as shown in the rejection above).
Qin in view of Abeysooriya, Turcot, Miserendino and Mizuno fails to disclose:
wherein the controller is configured to retrain the artificial neural network irrespective of a quantity of bit errors present in the memory device.
However, Sun explicitly discloses:
wherein the controller is configured to retrain the artificial neural network irrespective of a quantity of bit errors present in the memory device. (Sun, Col. 16, Lines 48-54: “In one embodiment, the bits in colunms associated with lower bit significances ( or importance) may be updated more frequently than bits associated in colunms associated with higher bit significance. This may be at least partly due to the nature weights in neural networks and how weights may be updated.”) [The updating or retraining process happens more frequently for less significant bits, which indicates that the retraining does not rely on the bit errors]
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin, Mizuno, Turcot, Miserendino, Abeysooriya and Sun. Qin teaches training a neural network within the memory storage of a device. Abeysooriya teaches continuously or periodically evaluating the performance of an artificial neural network after being deployed by using additional input and output data associated with the content distribution network. Turcot teaches a method for deep learning in-situ retraining. Miserendino teaches system and method for in-situ classifier retraining for malware identification and model heterogeneity. Mizuno teaches evaluating neural network’s performance and retraining the network. Sun teaches optimizing the neural network data organization. One of ordinary skill would have motivation to combine Qin, Abeysooriya, Turcot, Miserendino, Mizuno and Sun to allow the controller to access the metadata of a neural network more quickly and more efficiently (Sun, Col. 13, Lines 1-4)
Regarding Claim 18, the combination of Qin, Turcot, Miserendino, Mizuno and Abeysooriya discloses all the limitations of Claim 16 (as shown in the rejection above).
Qin in view of Turcot, Miserendino, Abeysooriya and Mizuno further discloses:
further comprising instructions to detect a plurality of bit errors in the memory device; and (Qin, [0041]: “Disclosed is a device for storing and constructing neural network by detecting weight values (also referred to as "weights" herein) of the neural network associated with faulty cells.”) [The examiner interprets that the “faulty cells” here is the “error in the memory”]
Qin in view of Abeysooriya, Turcot, Miserendino and Mizuno fails to disclose:
wherein the instructions to retrain the artificial neural network comprise instructions to retrain the artificial neural network irrespective of the plurality of errors.
However, Sun explicitly discloses:
wherein the instructions to retrain the artificial neural network comprise instructions to retrain the artificial neural network irrespective of the plurality of errors. (Sun, Col. 16, Lines 48-54: “In one embodiment, the bits in colunms associated with lower bit significances ( or importance) may be updated more frequently than bits associated in colunms associated with higher bit significance. This may be at least partly due to the nature weights in neural networks and how weights may be updated.”) [The updating or retraining process happens more frequently for less significant bits, which indicates that the retraining does not rely on the bit errors]
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin, Mizuno, Turcot, Miserendino, Abeysooriya and Sun. Qin teaches training a neural network within the memory storage of a device. Abeysooriya teaches continuously or periodically evaluating the performance of an artificial neural network after being deployed by using additional input and output data associated with the content distribution network. Turcot teaches a method for deep learning in-situ retraining. Miserendino teaches system and method for in-situ classifier retraining for malware identification and model heterogeneity. Mizuno teaches evaluating neural network’s performance and retraining the network. Sun teaches optimizing the neural network data organization. One of ordinary skill would have motivation to combine Qin, Abeysooriya, Turcot, Miserendino, Mizuno and Sun to allow the controller to access the metadata of a neural network more quickly and more efficiently (Sun, Col. 13, Lines 1-4)
Claim(s) 10, 17 are rejected under 35 U.S.C. 103 as being unpatentable over Qin et al. (US 2019/0073259 A1) in view of Abeysooriya et al (US 9,336,483 B1) (hereafter referred to as “Abeysooriya”), Mizuno et al. (5,577,166) (hereafter referred to as “Mizuno”), Turcot et al (US 2021/0125065 A1) (hereafter referred to as “Turcot”) and Miserendino et al (US 10,121,108 B2) (hereafter referred to as “Miserendino”) and further in view of Song et al. (US 2020/0036193 A1) (hereafter referred to as “Song”)
Regarding Claim 10, the combination of Qin, Mizuno, Turcot, Miserendino and Abeysooriya discloses all the limitations of Claim 9 (as shown in the rejection above).
Qin in view of Abeysooriya, Mizuno, Turcot, Miserendino fails to disclose:
wherein the predefined frequency is a user-defined parameter based at least in part on reliability or lifetime expectations of the apparatus.
However, Song explicitly discloses:
wherein the predefined frequency is a user-defined parameter based at least in part on reliability or lifetime expectations of the apparatus. (Song, [0055]: “To collect data for determining the first coefficient and the second coefficient, charge cycle tests are repeatedly performed under a predefined C-rate condition applied for each example. Additionally, success/failure is determined by evaluating the life performance for each example. The success standard of life performance is the capacity retention rate of 80% on the basis of 300 cycles.”) [The examiner interprets that the “predefined C-rate condition” here is the “predefined frequency” as setting a predefined C rate condition would be part of defining the parameters of the cycle test or evaluation to assess the battery life expectation]
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin, Abeysooriya, Mizuno, Turcot, Miserendino and Song. Qin teaches training a neural network within the memory storage of a device. Turcot teaches a method for deep learning in-situ retraining. Miserendino teaches system and method for in-situ classifier retraining for malware identification and model heterogeneity. Mizuno teaches evaluating neural network’s performance and retraining the network. Sun teaches optimizing the neural network data organization. Abeysooriya teaches continuously or periodically evaluating the performance of an artificial neural network after being deployed by using additional input and output data associated with the content distribution network. Song teaches evaluating the charging control apparatus. One of ordinary skill would have motivation to combine Qin, Abeysooriya, Turcot, Miserendino, Mizuno and Song to improve the charging performance of the battery by checking the device performance in cycles and tuning the negative electrode materials (Song, [0007])
Regarding Claim 17, the combination of Qin, Turcot, Mizuno, Miserendino and Abeysooriya discloses all the limitations of Claim 16 (as shown in the rejection above).
Qin in view of Abeysooriya, Mizuno, Turcot and Miserendino fails to disclose:
further storing a table comprising correspondences between respective frequencies of the periodic performance evaluation and respective reliability expectations and respective lifetime expectations; and
further storing instructions to select the frequency from the table based on the second definition.
However, Song explicitly discloses:
further storing a table comprising correspondences between respective frequencies of the periodic performance evaluation and respective reliability expectations and respective lifetime expectations; and (Song, [0056]: “FIG. 1 is a graph showing the life performance evaluation results obtained from the charge cycle tests for each example in the above Table 1. In the graph, the ■ mark indicates an example evaluated as having failed the life performance, and the * mark indicates an example evaluated as having succeeded the life performance.”) [The “charge cycle tests” here is the “respective frequencies of the periodic performance evaluation” and the “life performance evaluation results” is the “respective reliability expectations”]
further storing instructions to select the frequency from the table based on the second definition. (Song, [0055]: “Additionally, success/failure is determined by evaluating the life performance for each example. The success standard of life performance is the capacity retention rate of 80% on the basis of 300 cycles. Accordingly, the capacity retention rate of 80% or more is a success, and the capacity retention rate of less than 80% is a failure.”) [The examiner interprets that the performance evaluation frequency of 300 cycles is selected as it indicates a success with more than 80% of capacity retention rate ]
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Qin, Abeysooriya, Turcot, Mizuno, Miserendino and Song. Qin teaches training a neural network within the memory storage of a device. Turcot teaches a method for deep learning in-situ retraining. Miserendino teaches system and method for in-situ classifier retraining for malware identification and model heterogeneity. Mizuno teaches evaluating neural network’s performance and retraining the network. Sun teaches optimizing the neural network data organization. Abeysooriya teaches continuously or periodically evaluating the performance of an artificial neural network after being deployed by using additional input and output data associated with the content distribution network. Song teaches evaluating the charging control apparatus. One of ordinary skill would have motivation to combine Qin, Abeysooriya, Turcot, Mizuno, Miserendino and Song to improve the charging performance of the battery by checking the device performance in cycles and tuning the negative electrode materials (Song, [0007])
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to AMY TRAN whose telephone number is (571)270-0693. The examiner can normally be reached Monday - Friday 7:30 am - 5:00 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, David Yi can be reached at (571) 270-7519. 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.
/AMY TRAN/Examiner, Art Unit 2126
/DAVID YI/Supervisory Patent Examiner, Art Unit 2126
Prosecution Insights