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 .
This action is in response to the amendment filed on 01/09/2026. Claims 1-12 and 14-20 are pending in the case. All claims are examined and rejected accordingly.
Applicant Response
3. In Applicant’s response dated 01/09/2026, Applicant amended Claims 1-3, 7-8, 10, 12, 14-15, 17, and 19-20, cancelled claim 13 and argued against all objections and rejections previously set forth in the Office Action dated 10/10/2025.
Claim Rejections - 35 USC § 101
4. 35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
5. Claims 1-12 and 14-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed towards an abstract idea, without significantly more.
Step 1
According to the first part of the analysis, in the instant case, claim is directed to a computer implemented method, which is a process and falls within one of the four statutory categories (i.e., process, machine, manufacture, or composition of matter).
Regarding Claim 1, 19 and 20,
At step 2A, prong 1, Does the claim recite a judicial exception?
Claim 1 further recites the steps of :
generating basic training data corresponding to a combination of device data and simulation result data using a compact model, the compact model configured to generate the simulation result data and model uncertainty value by performing a simulation based on the device data, the simulation result data indicating characteristics of a semiconductor device corresponding to the device data (This step relies on mathematical modeling and mathematical simulation to produce output which falls into the “Mathematical Concepts” grouping of abstract ideas.)
training the deep learning model based on the basic training data such that the deep learning model is configured to output prediction data and uncertainty data, the prediction data indicating the characteristics of the semiconductor device the uncertainty data indicating uncertainty of the prediction data; (This step relies on training a deep learning model includes mathematical operation which falls into the “Mathematical Concepts” grouping of abstract ideas.), and
retraining the deep learning model based on the uncertainty data (This step relies on retraining a deep learning model using uncertainty data/metrics includes mathematical/ statistical optimization, which falls into the “Mathematical Concepts” grouping of abstract ideas.),
performing a first retraining based on the model uncertainty when the model uncertainty data is larger than a model reference value, wherein the model reference value is based on a target performance of the trained deep learning model (This step relies on training a deep learning model includes mathematical operation which falls into the “Mathematical Concepts” grouping of abstract ideas.);
performing a second retraining based on the data uncertainty value when the data uncertainty value is larger than a data reference value, wherein the data reference value is based on the target performance of the trained deep learning model (This step relies on training a deep learning model includes mathematical operation which falls into the “Mathematical Concepts” grouping of abstract ideas.),
The claim recites a process of performing simulations, creating /combining datasets to train deep learning model, computing predictions and uncertainties to retain the ML model, i.e., “Mathematical Concept” groupings of abstract ideas. Accordingly, the claims recite an abstract idea.
Step 2A prong 2: Does the claim recite additional elements? Do those additional elements, individually and in combination, integrate the judicial exception into a practical application?
Further, the claim does not recite any additional element which could integrate this abstract idea into a practical application, because the additional elements recited of consist of:
“… the method being performed by executing program codes by at least one processor, the program codes being stored in computer readable media …” (claim1), “the method being performed by executing program codes by at least one processor, the program codes being stored in computer readable media …”, (claim 19) and “A computing device comprising: at least one processor; and a computer readable medium storing program codes and a compact model, the program codes being executed by the at least one processor … ”( claim 20) ( Generic computer components on which to implement the math abstract idea (see MPEP 2106.05(f));
“… device data and simulation result data …” ( These are insignificant extra-solution activity of data gathering (see MPEP 2106.05(g))
data indicating characteristics of a semiconductor device corresponding to the device data ( These are references to field of use of semiconductor device and their characteristics and abstract calculation to field of use does not integrate the abstract idea to practical application (see MPEP 2106.05(h))
The additional elements are recited at a high level of generality and do not amount to significantly more than the abstract idea (MPEP 2106.05(f)). The claim use a computer to perform a math and does not improve the function of the computer or other technology. Accordingly, the claim does not integrate the abstract idea into practical application.
Thus, the claim is directed towards the abstract idea.
Step 2B: Do the additional elements, considered individually and in combination, amount to significantly more than the judicial exception?
No, As shown above with respect to integration of the abstract idea into a practical application, the additional element of “… the method being performed by executing program codes by at least one processor, the program codes being stored in computer readable media …” (claim1), “the method being performed by executing program codes by at least one processor, the program codes being stored in computer readable media …”, (claim 19) and “A computing device comprising: at least one processor; and a computer readable medium storing program codes and a compact model, the program codes being executed by the at least one processor … ”( claim 20) ( Generic computer components on which to implement the math abstract idea (see MPEP 2106.05(f));
“… device data and simulation result data …” ( These are insignificant extra-solution activity of data gathering (see MPEP 2106.05(g))
data indicating characteristics of a semiconductor device corresponding to the device data ( These are references to field of use of semiconductor device and their characteristics and abstract calculation to field of use does not integrate the abstract idea to practical application (see MPEP 2106.05(h)).
The additional elements, alone and in combination, fail to integrate the abstract idea into a practical application or add “significantly more.” Thus, the claims are not patent eligible. Mere instructions to apply an exception using generic computer components cannot provide an inventive concept. Neither can insignificant extra-solution activity. All of these additional elements as generically claimed are thus considered well-understood, routine, and conventional. Therefore, these limitations, taken alone or in combination, do not integrate the abstract idea into a practical application or recite significantly more that the abstract idea.
Thus, these independent claims are not patent eligible.
The dependent claims respectively recite a judicial exception in limitations of: “wherein model uncertainty value indicates the uncertainty of the prediction data caused by insufficiency of the basic training data.” (claims 2), “wherein the retraining the deep learning model includes: comparing the model uncertainty value with the model reference value; generating addition training data using the compact model when the model uncertainty value is larger than the model reference value; and retraining the deep learning model based on the addition training data, wherein the addition training data is different from the basic training data..” (claims 3, 14), “wherein retraining the deep learning model further includes: determining an addition data range corresponding to a range of data such that the model uncertainty value is larger than the model reference value.” (claims 4), “wherein the addition training data correspond to a combination of the device data included in the addition data range and the simulation result data.”, (claim 5, 18), “wherein the operations further comprise: comparing the first output against the second output, wherein the performing the action is further based on the comparing.” (claims 6, 19), “wherein the uncertainty data includes a data uncertainty value, the data uncertainty value indicating the uncertainty of the prediction data caused by noises in the basic training data.” (claims 7, 20), “wherein retraining the deep learning model includes: comparing the data uncertainty value with the data reference value; providing measurement data by measuring the characteristics of the semiconductor device, when the data uncertainty value is larger than the data reference value; correcting the compact model based on the measurement data; generating updated training data using the corrected compact model; and retraining the deep learning model based on the updated training data.”(Claim 8), “wherein retraining the deep learning model further includes: determining a measurement data range corresponding to a range of data such that the data uncertainty value is larger than the data reference value.”(claim 9), “wherein the characteristics of the semiconductor device corresponds to the device data included in the measurement data range.”(Claim 10), “wherein the deep learning model that has been trained based on the basic training data is initialized, and the initialized deep learning model is trained based on the measurement data.”,(claim 11), “wherein the model uncertainty value indicates the uncertainty of the prediction data caused by insufficiency of the basic training data, and the data uncertainty value indicating the uncertainty of the prediction data caused by noises of the basic training data.”(claim 12), “wherein whether to perform the first retraining is first determined based on the model uncertainty value, and when it is determined that the first retraining is not performed, subsequently determining whether to perform the second retraining is determined based on the data uncertainty value.”(claim 14), “wherein whether to perform the first retraining and whether to perform the second retraining are determined independently of each other.”(claim 15), “wherein the deep learning model includes a Bayesian Neural Network (BNN).”(claim 16), “wherein the device data indicats structure and operation condition of the semiconductor device, the simulation result data and the prediction data indicate electrical characteristics of the semiconductor device, and the device data is included in input data of the deep learning model.”(claim 17), “wherein the input data of the deep learning model further includes process data indicating a condition of manufacturing process of the semiconductor device.”(claim 18).
These additional limitations (in claims 2-18 ) also constitute concepts performed in the human mind which fall within the “Mental Processes” groupings of abstract ideas.
This judicial exception is not integrated into a practical application. Additional elements “computer readable medium comprising: computer program code (in claims 2-18), all amount to no more than adding insignificant extra-solution activity/specifications related to data gathering, data input, or data transmittal. These additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The dependent claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of non-transitory computer readable medium comprising: computer program code are again insignificant extra-solution activity steps that cannot provide an inventive concept. All of these additional elements as generically claimed are considered well-understood, routine, and conventional.
Therefore, these limitations, taken alone or in combination, do not integrate the abstract idea into a practical application or recite significantly more that the abstract idea. Thus, all of the dependent claims are also not patent eligible.
Claim 20 is rejected under 35 U.S.C. 101 because claims 20 is directed to a “computer readable storage medium” that could be non-transitory medium or transitory medium since the specification is silent with respect to which the “computer readable storage medium” includes or excludes. The specification in paragraph [0050] states ambiguous definition of the storage medium as “processor or values obtained from arithmetic processing performed by the processor may be stored in a transitory and/or non-transitory computer-readable medium.” As such, in a broadest reasonable interpretation, the claimed medium can include signal per se which is non-statutory. Examiner recommends that the claims be amended to “non-transitory computer readable storage medium” in order to overcome these 101 rejections.
Appropriate correction is required.
Examiner Comments
7. 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 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.
Claim Rejections - 35 USC § 103
8. 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.
9. Claims 1-12 and 14-20 are rejected under 35 U.S.C. 103 as being unpatentable over Dandy (Pat. No.: US 20210081592 B1, Pub. Date: 2021-03-18) in view of ASENOV (Pub. No US 20170103153 A1, Pub. Date 2017-04-13.) in further view of MIDDLEBROOKS (US 20210286270 A1 2021-09-16)
Dandy teaches a method of generating a deep learning model (see Dandy: Fig.3, [0032], system 300 also includes a machine 310 that hosts a machine learning facility 312 configured to use supervised or unsupervised machine learning techniques.”), the method being performed by executing program code by at least one processor, the program codes being stored in computer readable media the method comprising:
generating basic training data corresponding to a combination of device data and simulation result data (see Dandy: Fig.1, [0023], At 106, a training dataset that includes the simulation model component parameters ( i.e. device data) from operation 102 and the corresponding simulated values produced in operation 104 ( i.e. simulation result data) are provided as training data input to a machine learning facility.”), using a [circuit simulation model ] (see Dandy: Fig.1, [0022], “The simulator accepts the input values and, at operation 104, uses a circuit simulator (i.e. compact model) to generate simulated output values that are stored in the particular instance that was used to generate the output values”), the [circuit simulation model ] configured to generate the simulation result data and a model uncertainty value by performing a simulation based on the device data (see Dandy: Fig.1, [0014], “a dataset of generated values are combined into a simulation result value for each of the simulation model component parameters. The inputs to the simulation may include values for the parameters of the individual components in the circuit identified in operation 102, and the output of the simulation may include a very large dataset of individual instances of the inputs to the simulation along with simulated circuit outputs.”… [0028], “he measured data may be manipulated to account for measurement uncertainty and reflect a set of possible measurements. For example, if the DC gain of a certain measurement had a ±2% accuracy, then signals that were up to 2% higher and up to 2% lower could be stored as an additional training set or as additional instances of the original training set.”), the simulation result data indicating characteristics of a semiconductor device corresponding to the device data (see Dandy: Fig.1, [0022], The output values simulated by the simulator and stored in the dataset in operation 104 may include current, voltage, etc., as shown in Table 1 (i.e. characteristics of a semiconductor device). In some embodiments the simulated output values may also include waveforms of particular testing nodes of the network at particular time.”)
training the deep learning model based on the basic training data such that the deep learning model is configured to output prediction data (see Dandy: Fig.1, [0024], “After the machine learning facility receives its training data in operation 106, the trained network is created in operation 108. In operation 108, for example, a neural network may read the inputs from the first instance and generate a predicted outcome. Then the neural network compares its generated predicted outcome to the data used to create the simulated results, also included in the instance, and uses back propagation to modify weights and biases within the neural network so that its next prediction will be closer to the original data value.”) and uncertainty data (see Dandy: Fig.1, [0025], “The machine learning facility may apply a particular technique, such as a Bayesian approach, Random Forest, regression models, or classification models.”, i.e. Bayesian approach is a principled method for predicting uncertainty by providing a full probability distribution), […]
Dandy does not teach the system wherein:
the compact model configured to generate the simulation result data and a model uncertainty value indicating the characteristics of semiconductor device;
the uncertainty data indicating uncertainty of the prediction data; and
performing a first retraining based on the model uncertainty when the model uncertainty data is larger than a model reference value, wherein the model reference value is based on a target performance of the trained deep learning model: and
performing a second retraining based on the data uncertainty value when the data uncertainty value is larger than a data reference value, wherein the data reference value is based on the target performance of the trained deep learning model.
However, ASENOV teaches a system wherein:
a compact model is configured to generate simulation result data indicating the characteristics of semiconductor device (see ASENOV: Fig.1, [0053], “Compact transistor models such as BSIM4 (Berkeley Short-channel IGFET Model 4) and BSIM-CMG (Berkeley Short-channel IGFET Model—Common Multi-Gate) are simplified physical models typically employed in circuit simulators, for example SPICE (Simulation Program with Integrated Circuit Emphasis), to model the behavior of semiconductor devices such as CMOS field effect transistors in integrated circuits. The set of compact model parameters that specify the behavior of a particular semiconductor device are stored in a data structure called a model card, which is used as an input to a SPICE simulation process.”)
Because both Dandy and ASENOV are in the same/similar field of semiconductor device characteristics simulation, accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention, to modify the teaching of medium of Dandy to include the system, method and medium of wherein a compact model is configured to generate simulation result data indicating the characteristics of semiconductor device as taught by ASENOV. After modification of Dandy, the circuit simulation model simulate device data to generate a training data for the deep learning model can also incorporate the use of compact model to generate simulation result data as taught by as taught by ASENOV. One would have been motivated to make such a combination in order to provide effective and efficient semiconductor device production by increasing production speed and quality.
Dandy and ASENOV does not explicitly teach the system wherein:
the uncertainty data indicating uncertainty of the prediction data;
performing a first retraining based on the model uncertainty when the model uncertainty data is larger than a model reference value, wherein the model reference value is based on a target performance of the trained deep learning model: and
performing a second retraining based on the data uncertainty value when the data uncertainty value is larger than a data reference value, wherein the data reference value is based on the target performance of the trained deep learning model.
However, MIDDLEBROOKS teaches the system wherein :
the uncertainty data indicating uncertainty of the prediction data (see MIDDLEBROOKS: Fig.1, [0020], “using the determined variability in the predicted multiple posterior distributions and/or the quantified uncertainty to adjust the machine learning model to decrease the uncertainty of the machine learning model by making the machine learning model more descriptive or including more diverse training data.”)
performing a first retraining based on the model uncertainty when the model uncertainty data is larger than a model reference value (see MIDDLEBROOKS: Fig.1, [0027], “using the determined variability in the predicted multiple posterior distributions to adjust the one or more parameters of the machine learning model to decrease the uncertainty of the machine learning model comprises training the machine learning model with additional and more diverse training samples”, retraining), wherein the model reference value is based on a target performance of the trained deep learning model (see MIDDLEBROOKS: Fig.1, [0024], “adjusting the machine learning model to decrease the uncertainty of the machine learning model comprises increasing a training set size and/or adding to a dimensionality of a latent space associated with the machine learning model.”), and
performing a second retraining based on the data uncertainty value when the data uncertainty value is larger than a data reference value (see MIDDLEBROOKS: Fig.1, [0043], “the uncertainty of the parameterized model is related to the uncertainty of weights of parameters of the parameterized model, and the size and descriptiveness of the latent space, such that uncertainty in the weights manifests in uncertainty in the output, causing increased output variance.”), wherein the data reference value is based on the target performance of the trained deep learning model (see MIDDLEBROOKS: Abstract, “adjusting the machine learning model to decrease the uncertainty of the machine learning model comprises increasing a training set size and/or adding to a dimensionality of a latent space associated with the machine learning model.”)
Because both Dandy, ASENOV and MIDDLEBROOKS are in the same/similar field of semiconductor device characteristics simulation, accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention, to modify the teaching of medium of Dandy to include the wherein the uncertainty data indicating uncertainty of the prediction data and retraining the deep learning model based on the uncertainty data as taught by MIDDLEBROOKS. After modification of Dandy, the circuit simulation model simulate device data to generate a training data for the deep learning model can also incorporate the use of compact model to generate simulation result data as taught by as taught by MIDDLEBROOKS. One would have been motivated to make such a combination in order to increase optimization of learning models by generating and analyzing extensive prediction data for diagnosis and prognosis for the device.
Regarding Claim 2,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 1. Dandy further teaches the method wherein:
the uncertainty data include a model uncertainty value indicating the uncertainty of the prediction data caused by insufficiency of the basic training data ( see MIDDLEBROOKS: Fig.1, [0043], “he uncertainty of the parameterized model is related to the uncertainty of weights of parameters of the parameterized model, and the size and descriptiveness of the latent space, such that uncertainty in the weights manifests in uncertainty in the output, causing increased output variance.”)
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention, to modify the teaching of medium of Dandy to include the wherein the uncertainty data include a model uncertainty value indicating the uncertainty of the prediction data caused by insufficiency of the basic training data as taught by MIDDLEBROOKS. One would have been motivated to make such a combination in order to increase optimization of learning models by generating and analyzing extensive prediction data for diagnosis and prognosis for the device.
Regarding Claim 3,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 2. Dandy further teaches the method wherein:
the retraining the deep learning model (see Dandy: Fig.1,0024, “The training continues progressing through all of the instances until the network is fully trained. In some embodiments the training makes multiple passes through the training database, as indicated through the optional loopback 109. In some embodiments the neural network may change the order of the instances as they appear in the training dataset to avoid training biases.”), includes:
comparing the model uncertainty value with a model reference value (see Dandy: Fig1, [0024], “Then the neural network compares its generated predicted outcome to the data used to create the simulated results, also included in the instance, and uses back propagation to modify weights and biases within the neural network so that its next prediction will be closer to the original data value.”)
generating addition training data using the compact model when the model uncertainty value is larger than the model reference value (see Dandy: Fig1, [0028], “Once these measurements are obtained, the measurements acquired in operation 202 may be applied to the trained learning network, in an operation 204, to predict or infer a set of revised simulation model parameters that better match the measured result than the original predicted model. In certain implementations, the measured data may be manipulated to account for measurement uncertainty and reflect a set of possible measurements”; and
retraining the deep learning model based on the addition training data, wherein the addition training data is different from the basic training data (see Dandy: Fig1, [0028], “The machine learning network could then be improved (i.e., updated or retrained) in an operation 206 with the additional instances or new training dataset as described above. The result after re-training the machine learning network in operation 206 would be a set of nominal model parameters, each of which would have a certain range of uncertainty due to measurement errors.”)
Regarding Claim 4,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 3. Dandy further teaches the method wherein:
retraining the deep learning model further (see Dandy: Fig.1, [0024], “The training continues progressing through all of the instances until the network is fully trained. In some embodiments the training makes multiple passes through the training database, as indicated through the optional loopback 109. In some embodiments the neural network may change the order of the instances as they appear in the training dataset to avoid training biases.”), includes:
determining an addition data range corresponding to a range of data such that the model uncertainty value is larger than the model reference value (see Dandy: Fig.1,[0028], “In certain implementations, the measured data may be manipulated to account for measurement uncertainty and reflect a set of possible measurements. For example, if the DC gain of a certain measurement had a ±2% accuracy, then signals that were up to 2% higher and up to 2% lower could be stored as an additional training set or as additional instances of the original training set.”)
Regarding Claim 5,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 1. Dandy further teaches the method wherein:
the addition training data correspond to a combination of the device data included in the addition data range (see Dandy: Fig.1, [0019], “the input parameters for the dataset may be bound within a range of values. For example, the values of the resistance parameter for each of the three resistors in the above example may be bound between 100 Ohms and 1000 Ohms when generating the training dataset.”), and the simulation result data (see Dandy: Fig.1, [0020], “the generation of the dataset produced in the operation 104, the simulation may create instances by varying the resistance values for each of the three resistors independently, as well as simulate and store output values for the various testing nodes within (or at the endpoints of) the resistor network.”)
Regarding Claim 6,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 3. Dandy further teaches the method wherein:
the deep learning model that has been trained based on the basic training data is further trained based on the addition training data (see Dandy: Fig.2, [0028], “for example, if the DC gain of a certain measurement had a ±2% accuracy, then signals that were up to 2% higher and up to 2% lower could be stored as an additional training set or as additional instances of the original training set. The machine learning network could then be improved (i.e., updated or retrained) in an operation 206 with the additional instances or new training dataset as described above.”)
Regarding Claim 7,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 1. Dandy further teaches the method wherein:
the uncertainty data includes a data uncertainty value, the data uncertainty value indicating the uncertainty of the prediction data caused by noises in the basic training data (see Dandy: Fig.2, [0028], “he result after re-training the machine learning network in operation 206 would be a set of nominal model parameters, each of which would have a certain range of uncertainty due to measurement errors.”)
Regarding Claim 8,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 7. Dandy further teaches the method wherein:
retraining the deep learning model (see Dandy: Fig.1, [0024], “The training continues progressing through all of the instances until the network is fully trained. In some embodiments the training makes multiple passes through the training database, as indicated through the optional loopback 109. In some embodiments the neural network may change the order of the instances as they appear in the training dataset to avoid training biases.”), includes:
comparing the data uncertainty value with the data reference value (see Dandy: Fig.3, [0035], “he machine learning facility 312 may apply machine learning to the simulated values to generate a predicted value for each inputted parameter and compare the predicted value to its corresponding simulated value. The machine learning facility 312 may adjust the simulation model based on the comparison of the predicted to the simulated values.”);
providing measurement data by measuring the characteristics of the semiconductor device, when the data uncertainty value is larger than the data reference value (see Dandy: Fig.3, [0033], “The input 324 may be structured to receive measurement data either directly from a measurement or data extracted from another measurement device. The input 324 may also be structured to receive data that has been previously stored or data from an information cloud.”);
correcting the compact model based on the measurement data (see Dandy: Fig.2, [0031], “If the physical embodiment, such as a circuit, is not working as expected, the updated simulation model may generate possible insight into what is different from the intended design, as well as which parameters may be contributing to these differences. The adjusted simulation model may also provide a validated starting point for design revisions.”);
generating updated training data using the corrected compact model (see Dandy: Fig.2, [0028], “e machine learning network could then be improved (i.e., updated or retrained) in an operation 206 with the additional instances or new training dataset as described above. The result after re-training the machine learning network in operation 206 would be a set of nominal model parameters, each of which would have a certain range of uncertainty due to measurement errors.”); and
retraining the deep learning model based on the updated training (see Dandy: Fig.2, [0028], “machine learning network could then be improved (i.e., updated or retrained) in an operation 206 with the additional instances or new training dataset as described above. The result after re-training the machine learning network in operation 206 would be a set of nominal model parameters, each of which would have a certain range of uncertainty due to measurement errors.”)
Regarding Claim 9,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 8. Dandy further teaches the method wherein:
retraining the deep learning model (see Dandy: Fig.1, [0024], “The training continues progressing through all of the instances until the network is fully trained. In some embodiments the training makes multiple passes through the training database, as indicated through the optional loopback 109. In some embodiments the neural network may change the order of the instances as they appear in the training dataset to avoid training biases.”), further includes:
determining a measurement data range corresponding to a range of data such that the data uncertainty value is larger than the data reference value ( see Dandy: Fig.1, [0039], “The machine learning network may be trained with data that replaces the nominal circuit with varying possible manufacturing defects (e.g., missing components, shorts across nets, incorrect parts, etc.). In such implementations, one can vary parameters across possible fault values in addition to normal tolerance ranges. The measured data can be fed into a trained neural network to infer failures on a production line based on what defect is likely to be present, for example. Such implementations facilitate automated or semi-automated troubleshooting and corrective action efforts.”
Regarding Claim 10,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 9. ASENOV further teaches the method wherein:
the characteristics of the semiconductor device corresponds to the device data included in the measurement data range ( see ASENOV: Fig.1, [0050], “a drift-diffusion model is calibrated using the results of the ensemble Monte Carlo simulation. This is done by using the EMC generated current-voltage characteristics for a particular transistor to determine values of the parameters of a mobility model that is incorporated in the drift-diffusion model of this particular transistor.”)
See the motivation to combine Dandy, and ASENOV in claim 1.
Regarding Claim 11,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 8. Dandy further teaches the method wherein:
the deep learning model that has been trained based on the basic training data is initialized (see Dandy: Fig.1, [0023], “At 106, a training dataset that includes the simulation model component parameters from operation 102 and the corresponding simulated values produced in operation 104 are provided as training data input to a machine learning facility. Table 1 may be considered an overly simplified training dataset for explanation purposes.”), and the initialized deep learning model is trained based on the measurement data (see Dandy: Fig.1, [0024], “after the machine learning facility receives its training data in operation 106, the trained network is created in operation 108. In operation 108, for example, a neural network may read the inputs from the first instance and generate a predicted outcome. Then the neural network compares its generated predicted outcome to the data used to create the simulated results, also included in the instance, and uses back propagation to modify weights and biases within the neural network so that its next prediction will be closer to the original data value.”)
Regarding Claim 12,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 1. Dandy further teaches the method wherein:
the model uncertainty value indicates the uncertainty of the prediction data caused by insufficiency of the basic training data (see Dandy: Fig.1, [0028] “Once these measurements are obtained, the measurements acquired in operation 202 may be applied to the trained learning network, in an operation 204, to predict or infer a set of revised simulation model parameters that better match the measured result than the original predicted model. In certain implementations, the measured data may be manipulated to account for measurement uncertainty and reflect a set of possible measurements.”), and the data uncertainty value indicates the uncertainty of the prediction data caused by noises of the basic training data (see Dandy: Fig.1, [0028], “The result after re-training the machine learning network in operation 206 would be a set of nominal model parameters, each of which would have a certain range of uncertainty due to measurement errors.”)
Regarding Claim 14,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 1. Dandy further teaches the method wherein:
whether to perform the first retraining is first determined based on the model uncertainty value, and when it is determined that the first retraining is not performed, subsequently determining whether to perform the second retraining is determined based on the data uncertainty value ( see Dandy: Fig.1, [0024], “back propagation to modify weights and biases within the neural network so that its next prediction will be closer to the original data value. In the example of Table 1, the inputs to the network are the simulated results (Node 1 Predicted voltage, Node 2 Predicted Voltage, etc.), and the predicted output is a predicted value for R1, R2, and R3, which is compared, during training, to the original values of R1, R2, and R3 that were used to create the simulated results. The training continues progressing through all of the instances until the network is fully trained. In some embodiments the training makes multiple passes through the training database, as indicated through the optional loopback 109. In some embodiments the neural network may change the order of the instances as they appear in the training dataset to avoid training biases.”)
Regarding Claim 15,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 1. Dandy further teaches the method wherein:
whether to perform the first retraining and whether to perform the second retraining are determined independently of each other (see Dandy: Fig.1, [0026], “The machine learning network could then be improved (i.e., updated or retrained) in an operation 206 with the additional instances or new training dataset as described above. The result after re-training the machine learning network in operation 206 would be a set of nominal model parameters, each of which would have a certain range of uncertainty due to measurement errors.”)
Regarding Claim 16,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 1. Dandy further teaches the method wherein:
the deep learning model includes a Bayesian Neural Network (BNN) ( see Dandy: Fig.1, [0025], “ Operation 108 may use supervised machine learning or unsupervised machine learning. Supervised machine learning as used herein generally refers to machine learning that is based upon training sets that contain labeled data. Unsupervised machine learning generally refers to ‘learning’ on training sets that contain mostly unlabeled data to train the neural network. The machine learning facility may apply a particular technique, such as a Bayesian approach, Random Forest, regression models, or classification models.”)
Regarding Claim 17,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 1. Dandy further teaches the method wherein:
wherein the device data indicates structure and an operation condition of the semiconductor device (see Dandy: Fig.1, [0053], “Example 7 is the method of any one of examples 1-6, in which the simulation model is a circuit simulation model, and wherein the circuit simulation model includes at least one component selected from the group consisting of: a resistor, a transistor, a capacitor, an inductor, a diode, an operational amplifier, a voltage source, a current source, and a transmission line”
the simulation result data and the prediction data indicate electrical characteristics of the semiconductor device (see Dandy: Fig.1, [0054], “the simulation model is a circuit simulation model, and wherein the at least one parameter is selected from the group consisting of: resistance, impedance, temperature coefficient, parasitic capacitance, transmission line length, transmission line width, material dielectric constant, and geometry.”), and
the device data is included in input data of the deep learning model (see Dandy: Fig.1, [0023], At 106, a training dataset that includes the simulation model component parameters ( i.e. device data) from operation 102 and the corresponding simulated values produced in operation 104 ( i.e. simulation result data) are provided as training data input to a machine learning facility.”),
Regarding Claim 18,
As shown above, Dandy, ASENOV and MIDDLEBROOKS teaches all the limitations of claim 17. Dandy further teaches the method wherein:
the input data of the deep learning model further includes process data indicating a condition of manufacturing process of the semiconductor device (see Dandy: Fig.1, [0013], “the machine learning facility receives its training data in operation 106, the trained network is created in operation 108. In operation 108, for example, a neural network may read the inputs from the first instance and generate a predicted outcome. Then the neural network compares its generated predicted outcome to the data used to create the simulated results, also included in the instance, and uses back propagation to modify weights and biases within the neural network so that its next prediction will be closer to the original data value.”)
Regarding Claim independent 19,
Claim 19 is directed to a method claim and has similar/same claim limitation as claim 1 and is rejected under the same rationale.
Regarding Claim independent 20,
Claim 20 is directed to a computing device claim and has the same/similar claim limitations claim 1 and is rejected under the same rationale.
Response to Arguments
Claim Rejections - 35 U.S.C. § 103,
Applicant’s arguments with respect to claim amendments have been considered but are moot considering the new combination of references being used in the current rejection. The new combination of references was necessitated by Applicant’s claim amendments. Therefore, the claims are rejected under the new combination of references as indicated above.
Claim Rejections - 35 U.S.C. § 101,
Regarding the 35 U.S.C. 101 rejection for being directed non-statutory subject matter has been sustained based on applicant amendments and. Therefore, the 35 U.S.C. 101 rejection has been updated and sustained.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
PGPUB
NUMBER:
INVENTOR-INFORMATION:
TITLE / DESCRIPTION
US 20210390232 A1
KHANDELWAL; Ashish
Title: AUTOMATED ANALOG AND MIXED-SIGNAL CIRCUIT DESIGN AND VALIDATION
Description: Technique for designing circuits including receiving a data object representing a circuit for a first process technology, the circuit including a first sub-circuit, the first sub-circuit including a first electrical component and a second electrical component arranged in a first topology.
US 20220343171 A1
Cihangir; Neslihan Kos
Title: METHODS AND APPARATUS TO CALIBRATE ERROR ALIGNED UNCERTAINTY FOR REGRESSION AND CONTINUOUS STRUCTURED PREDICTION TASKS
Description: his disclosure relates generally to deep learning and, more particularly, to calibrating uncertainty for regression and continuous structured model prediction tasks.
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 ZELALEM W SHALU whose telephone number is (571)272-3003. The examiner can normally be reached M- F 0800am- 0500pm.
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, Cesar Paula can be reached at (571) 272-4128. 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.
/Zelalem Shalu/Examiner, Art Unit 2145
/CESAR B PAULA/Supervisory Patent Examiner, Art Unit 2145
Prosecution Insights