Circuit optimization methods, apparatus and computer equipment
By optimizing analog circuits using Gaussian process models and multi-point prediction techniques, the problem of low circuit optimization efficiency was solved, enabling rapid circuit optimization and shortening electromagnetic simulation time, thereby improving circuit design efficiency.
Patent Information
- Application Number
- CN202311315782.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-10-11
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-10-11
AI Technical Summary
In analog circuit design, the increased complexity of transistor models leads to low circuit optimization efficiency, excessively long electromagnetic simulation time, and an increased number of iterative optimizations, which affects the efficiency of circuit performance evaluation.
A Gaussian process model is used to construct the acquisition function, expectation boosting function, and probability boosting function. Circuit optimization is performed through multi-point prediction processing. The model is reconstructed using sample data and prediction performance data, and iterative optimization conditions are set to obtain the final optimization parameters.
It improves circuit optimization efficiency, reduces the number of electromagnetic simulation iterations, shortens the circuit optimization cycle, and enables rapid search for the globally optimal variable design.
Smart Images

Figure CN117391018B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of integrated circuit design technology, and in particular to a circuit optimization method, apparatus, computer equipment, storage medium, and computer program product. Background Technology
[0002] With the adoption of new processes, transistor models in analog circuit design are becoming increasingly complex, and higher-order effects and process deviations are having a more and more serious impact on circuit performance.
[0003] As analog circuits become increasingly large and device models become more complex, the time required for electromagnetic simulation of analog circuits and for evaluating circuit performance indicators also increases. This leads to an increasing number of iterative optimizations of the circuit based on the electromagnetic simulation results, resulting in decreasing efficiency in circuit optimization. Summary of the Invention
[0004] Therefore, it is necessary to provide a circuit optimization method, apparatus, computer equipment, storage medium, and computer program product that can improve the efficiency of circuit optimization in order to address the above-mentioned technical problems.
[0005] In a first aspect, this application provides a circuit optimization method, including:
[0006] In response to the optimization command, at least one set of target optimization parameters is determined from the parameters to be optimized corresponding to the circuit to be optimized;
[0007] Using at least one set of sample data corresponding to each of the target optimization parameters, a Gaussian process model is constructed, and an acquisition function, an expected boosting function, and a probability boosting function are constructed based on the Gaussian process model;
[0008] Based on the prediction performance data corresponding to the sample data output by the Gaussian process model, multi-point prediction processing is performed through the acquisition function, the expected boosting function, and the probability boosting function to obtain optimized data;
[0009] Based on the sample data, the prediction performance data, and the optimization data, the model is reconstructed, and the objective function of the Gaussian process model is iteratively optimized.
[0010] When the iteratively optimized objective function meets the preset conditions, the iteration stops, and the final optimized objective parameters are obtained based on the iteratively optimized objective function.
[0011] In one embodiment, before constructing the Gaussian process model using at least one set of sample data corresponding to each of the target optimization parameters, the method further includes:
[0012] Collect at least one set of sample data corresponding to each of the target optimization parameters.
[0013] In one embodiment,
[0014] The sample data includes input sample data and simulation performance data corresponding to the input sample data;
[0015] The process of obtaining optimized data by performing multi-point prediction processing on the prediction performance data corresponding to the sample data output by the Gaussian process model through the acquisition function, the expected boosting function, and the probability boosting function includes:
[0016] Using the acquisition function, the expected boosting function, and the probability boosting function, new input sample data are determined based on the prediction performance data corresponding to the input sample data output by the Gaussian process model.
[0017] Based on the simulation performance data, determine the simulation performance data corresponding to the new input sample data;
[0018] The input sample data and the simulation performance data corresponding to the input sample data, the new input sample data and the simulation performance data corresponding to the new input sample data are used as the optimization data.
[0019] In one embodiment, determining the simulation performance data corresponding to the new input sample data based on the new input sample data includes:
[0020] The minimum, average, or maximum value in the simulation performance data is used as the simulation performance data corresponding to the new input sample data.
[0021] In one embodiment, after reconstructing the model based on the sample data, the prediction performance data, and the optimization data, and iteratively optimizing the objective function of the Gaussian process model, the method further includes:
[0022] Increment the number of optimizations corresponding to the objective function by one.
[0023] In one embodiment, the preset condition includes the number of optimizations reaching a preset optimization threshold;
[0024] When the iteratively optimized objective function meets a preset condition, the iteration stops, and the final optimized target parameters are obtained based on the iteratively optimized objective function, including:
[0025] When the number of optimization attempts reaches a preset threshold, the iteration stops, and the final optimized target parameters are determined based on the optimized objective function.
[0026] In one embodiment, after reconstructing the model based on the sample data, the prediction performance data, and the optimization data, and iteratively optimizing the objective function of the Gaussian process model, the method further includes:
[0027] Obtain the prediction performance parameters output by the optimized objective function;
[0028] Based on the predicted performance parameters, determine the confidence level corresponding to the optimized target parameters.
[0029] In one embodiment, the preset condition includes the confidence level reaching a preset value;
[0030] When the iteratively optimized objective function meets a preset condition, the iteration stops, and the final optimized target parameters are obtained based on the iteratively optimized objective function, including:
[0031] When the confidence level reaches the preset value, the iteration stops, and the target optimization parameters obtained according to the optimized objective function are used as the final optimized target optimization parameters.
[0032] Secondly, this application also provides a circuit optimization device, comprising:
[0033] The determination module is used to determine at least one target optimization parameter from the parameters to be optimized corresponding to the circuit to be optimized in response to the optimization command;
[0034] The construction module is used to construct a Gaussian process model using at least one set of sample data corresponding to each of the target optimization parameters, and to construct an acquisition function, an expected boosting function and a probability boosting function based on the Gaussian process model;
[0035] The prediction module is used to perform multi-point prediction processing based on the prediction performance data corresponding to the sample data output by the Gaussian process model, through the acquisition function, the expected boosting function and the probability boosting function, to obtain optimized data;
[0036] The optimization module is used to reconstruct the model based on the sample data, the prediction performance data, and the optimization data, and to iteratively optimize the objective function of the Gaussian process model.
[0037] The judgment module is used to stop the iteration when the iteratively optimized objective function meets the preset conditions, and to obtain the final optimized target parameters based on the iteratively optimized objective function.
[0038] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the circuit optimization method described in any of the above embodiments.
[0039] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the circuit optimization method described in any of the above embodiments.
[0040] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the circuit optimization method described in any of the above embodiments.
[0041] The aforementioned circuit optimization methods, devices, computer equipment, storage media, and computer program products can optimize and expand input variables without requiring new input data or re-estimation of hyperparameters. This facilitates the search for globally optimal variable design, accelerates the iterative optimization speed of the objective function, and consequently reduces the number of electromagnetic simulation iterations and shortens the circuit optimization cycle. Attached Figure Description
[0042] Figure 1 This is a diagram illustrating the application environment of the circuit optimization method in one embodiment;
[0043] Figure 2 This is a flowchart illustrating a circuit optimization method in one embodiment;
[0044] Figure 3 This is a flowchart illustrating a circuit optimization method in one embodiment;
[0045] Figure 4 This is a flowchart illustrating a circuit optimization method in one embodiment;
[0046] Figure 5 This is a flowchart illustrating a circuit optimization method in one embodiment;
[0047] Figure 6 This is a flowchart illustrating a circuit optimization method in one embodiment;
[0048] Figure 7 This is a block diagram of the circuit optimization device in one embodiment;
[0049] Figure 8 This is a schematic diagram of the circuit structure for performing actual electromagnetic simulation processing in one embodiment.
[0050] Figure 9The graph shows the changes in the actual initial saturation gain, initial saturation output power, and initial power-added efficiency under different electromagnetic wave frequency ranges when performing electromagnetic simulations for initial variables in one embodiment.
[0051] Figure 10 This diagram illustrates the number of iterations required for the objective function to converge during electromagnetic simulation, comparing variables predicted at a single point with those predicted at three points in one embodiment.
[0052] Figure 11 This is a graph showing the changes in power-added efficiency of variables after single-point prediction and variables after three-point prediction in an electromagnetic simulation under different electromagnetic wave frequency ranges in one embodiment.
[0053] Figure 12 This is a graph showing the changes in saturation gain of variables after single-point prediction and variables after three-point prediction in an electromagnetic simulation under different electromagnetic wave frequency ranges in one embodiment.
[0054] Figure 13 This is a graph showing the changes in saturated output power under different electromagnetic wave frequency ranges when performing electromagnetic simulations on variables after single-point and three-point predictions in one embodiment.
[0055] Figure 14 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0056] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0057] The circuit optimization method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network.
[0058] For example, the circuit optimization method is applied to terminal 102. When terminal 102 receives an optimization instruction from the human-computer interaction interface, it can determine at least one set of target optimization parameters from the parameters to be optimized corresponding to the circuit to be optimized. The parameters to be optimized corresponding to the circuit to be optimized can be pre-stored in the data storage system of server 104. Subsequently, terminal 102 uses at least one set of sample data corresponding to each target optimization parameter to construct a Gaussian process model, and constructs an acquisition function, an expectation boosting function, and a probability boosting function based on the Gaussian process model. Further, based on the prediction performance data corresponding to the sample data output by the Gaussian process model, multi-point prediction processing is performed through the acquisition function, expectation boosting function, and probability boosting function to obtain optimization data. Based on the sample data, prediction performance data, and optimization data, model reconstruction is performed, and the objective function of the Gaussian process model is iteratively optimized. When the iteratively optimized objective function meets the preset conditions, the iteration stops, and the final optimized target optimization parameters are obtained based on the iteratively optimized objective function. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers. Terminal 102 and server 104 can be connected directly or indirectly via wired or wireless communication, such as through a network connection.
[0059] For example, when the circuit optimization method is applied to server 104, when terminal 102 receives an optimization instruction from the human-computer interaction interface, it can send the instruction to server 104. Server 104 determines at least one set of target optimization parameters from the parameters to be optimized corresponding to the circuit to be optimized, based on the optimization instruction. The parameters to be optimized corresponding to the circuit to be optimized can be pre-stored in the data storage system of server 104. Subsequently, server 104 uses at least one set of sample data corresponding to each target optimization parameter to construct a Gaussian process model, and constructs an acquisition function, an expectation boosting function, and a probability boosting function based on the Gaussian process model. Further, based on the prediction performance data corresponding to the sample data output by the Gaussian process model, multi-point prediction processing is performed through the acquisition function, expectation boosting function, and probability boosting function to obtain optimization data. Based on the sample data, prediction performance data, and optimization data, model reconstruction is performed, and the objective function of the Gaussian process model is iteratively optimized. When the iteratively optimized objective function meets the preset conditions, the iteration stops, and the final optimized target optimization parameters are obtained based on the iteratively optimized objective function. It is understood that the data storage system can be an independent storage device, or the data storage system can be located on a server, or the data storage system can be located on another terminal.
[0060] In one embodiment, a circuit optimization method is provided. This embodiment illustrates the application of this circuit optimization method to a terminal. It is understood that this circuit optimization method can also be applied to a server, and further to a system including both a terminal and a server, and implemented through the interaction between the terminal and the server. Figure 2 As shown, the circuit optimization method includes:
[0061] Step 202: In response to the optimization instruction, determine at least one set of target optimization parameters from the parameters to be optimized corresponding to the circuit to be optimized.
[0062] An optimization instruction is an instruction that optimizes all adjustable parameters of the circuit to be optimized in order to obtain a circuit with better performance.
[0063] As an example, optimization commands can be issued by the user through the terminal's human-computer interaction interface.
[0064] According to the functional classification of circuits, the circuit to be optimized can be an amplifier circuit, a filter circuit, a comparator circuit, a clock circuit, a logic circuit, a memory circuit, a power supply circuit, a communication circuit, etc. Specifically, the circuit to be optimized can be a power amplifier circuit.
[0065] The parameters to be optimized in the circuit to be optimized include, but are not limited to: the resistance values of the resistors in the circuit to be optimized (adjusting the resistance values can change the circuit's response and performance); the values of the capacitors in the circuit to be optimized (adjusting the capacitor values can change the circuit's frequency response and energy storage capacity); the values of the inductors in the circuit to be optimized (adjusting the inductor values can change the circuit's frequency response and energy storage capacity); the gain of the amplifiers in the circuit to be optimized (adjusting the amplifier gain can change the signal amplification); and the parameters of the filters in the circuit to be optimized (adjusting the filter's cutoff frequency, bandwidth, etc., can change the filter's frequency response). The clock frequency of the circuit can be adjusted to change the circuit's operating speed; the parameters of the manager in the circuit to be optimized can be adjusted to change the circuit's power consumption and performance; the parameters of the transmission lines in the circuit to be optimized can be adjusted to change the transmission speed and impedance matching; the parameters of the switches in the circuit to be optimized can be adjusted to change the switching circuit's response and power consumption; the parameters of other circuit components in the circuit to be optimized can also be adjusted to optimize the circuit's performance and power consumption, depending on the specific circuit design.
[0066] The target optimization parameters refer to the parameters selected from the parameters to be optimized and determined to be optimized.
[0067] Step 204: Using at least one set of sample data corresponding to each objective optimization parameter, construct a Gaussian process model, and construct the acquisition function, expected boosting function and probability boosting function based on the Gaussian process model.
[0068] The sample data are the input parameters for the Gaussian process model.
[0069] In this embodiment, a circuit to be optimized may have at least one parameter to be optimized, and correspondingly, a circuit to be optimized may have at least one target optimization parameter, and each target optimization parameter has at least one sample data. As an example, when there are k target optimization parameters, k sets of sample data corresponding to the k target optimization parameters can be obtained respectively, and the sample data in each set of sample data should conform to the characteristics of a normal distribution.
[0070] Each of the k groups of sample data is taken and combined to form t variables x of the Gaussian process model.
[0071] Step 206: Based on the prediction performance data corresponding to the sample data output by the Gaussian process model, perform multi-point prediction processing through the acquisition function, expectation boosting function and probability boosting function to obtain optimized data.
[0072] As an example, a multi-point prediction strategy using the Constant Liar (CL) algorithm can be adopted, which allows for multi-point prediction of performance data without the need for new input data or re-estimation of hyperparameters, thus obtaining optimized data.
[0073] Step 208: Based on the sample data, prediction performance data, and optimization data, reconstruct the model and iteratively optimize the objective function of the Gaussian process model.
[0074] Step 210: When the iteratively optimized objective function meets the preset conditions, stop the iteration and obtain the final optimized objective parameters based on the iteratively optimized objective function.
[0075] The objective function can be any evaluable metric, such as accuracy, loss function, cross-validation score, etc.
[0076] The circuit optimization method described above uses a multi-point prediction approach, which can optimize and expand the input variables without requiring new input data or re-estimation of hyperparameters. This helps to search for the globally optimal variable design, accelerates the iterative optimization speed of the acquisition function, expectation boosting function, and probability boosting function, thereby reducing the number of electromagnetic simulation iterations and shortening the circuit optimization cycle.
[0077] like Figure 3 As shown, in some optional embodiments, before step 204, the following steps are also included:
[0078] Step 203: Collect at least one set of sample data corresponding to each target optimization parameter.
[0079] like Figure 4 As shown, in some optional embodiments, the sample data includes input sample data and simulation performance data corresponding to the input sample data;
[0080] Step 206 includes:
[0081] Step 2062: Using the acquisition function, expectation boosting function, and probability boosting function, determine new input sample data based on the prediction performance data corresponding to the input sample data output by the Gaussian process model;
[0082] Step 2064: Based on the simulation performance data, determine the simulation performance data corresponding to the new input sample data;
[0083] Step 2066: Use the input sample data and the simulation performance data corresponding to the input sample data, the new input sample data and the simulation performance data corresponding to the new sample data as optimization data.
[0084] In step 2062, the optimal variables can be determined by solving for the acquisition function in order to maximize the expected boost function and the probability boost function.
[0085] In this embodiment, the CL algorithm is used for multi-point prediction of data. First, the above-mentioned variables, i.e., the input sample data, are obtained. Predictive performance data obtained after inputting the Gaussian process model Then the predicted performance data will be used. With input sample data Rebind together as the latest input sample data Simultaneously, the simulation performance data corresponding to the input sample data will be... With a fixed value L, i.e. The simulation performance data is bound to the new input sample data. .
[0086] Latest input sample data and simulation performance data This constitutes the optimized data.
[0087] The fixed value L can be, for example, one of the minimum, average, or maximum values in the simulation performance data Y.
[0088] Based on the optimized data mentioned above, the prediction trend of the Gaussian process model can be updated.
[0089] Based on the above, the covariance matrix corresponding to each variable is calculated, and the optimized Gaussian process model is obtained based on the covariance:
[0090]
[0091] Where k represents the kernel function, which is the covariance matrix. Let represent the mean function matrix, and x represent the variables of the Gaussian process model.
[0092] Kernel functions on prior data The following can be represented as:
[0093]
[0094] Then, the Bayesian Optimization (BO) algorithm is used to determine the next point to find. Its function value , and They satisfy a joint Gaussian distribution:
[0095]
[0096] Furthermore, predict and optimize under prior conditions The probability distribution function that is satisfied is:
[0097]
[0098] in, , .
[0099] in, Represents the Gaussian process function. This represents the mean. Indicates variance.
[0100] In Boolean optimization (BO), the algorithm aims to optimize the Gaussian process model to find the minimum or maximum value given the current input variables. This requires searching for unknown variables using Acquisition Functions (AC-Fun). The Expected Improvement (EI) and Robustness of Improvement (PI) functions can balance the utilization of known variables x (for local optimization) and the exploration of unknown variables (for global optimization).
[0101] The expected improvement function is expressed as:
[0102]
[0103] in, .
[0104] The probability boosting function is expressed as:
[0105]
[0106] like Figure 5 As shown, in some optional embodiments, after step 208, the following steps are included:
[0107] Step 209a: Increment the number of optimizations corresponding to the objective function by one.
[0108] In this embodiment, the preset conditions include reaching a preset optimization threshold number of times;
[0109] Accordingly, step 210 includes: step 210a, when the number of optimizations reaches the preset number of optimizations threshold, stopping the iteration, and determining the final optimized target optimization parameters based on the optimized objective function.
[0110] By presetting the number of iterations and setting the trigger condition for the end of model iteration, when the number of optimizations of the Gaussian process model reaches the preset optimization threshold, subsequent iterations will automatically stop. Based on the prediction performance parameters of each variable output by the objective function of the Gaussian process model completed in the latest iteration, the prediction performance parameters can contain at least one numerical value. The sample data of each variable corresponding to the optimal prediction performance parameter will be used as the final optimized result of each objective optimization parameter.
[0111] like Figure 6 As shown, in some optional embodiments, after step 208, the following steps are included:
[0112] Step 209b: Obtain the prediction performance parameters of the optimized objective function output;
[0113] Step 209c: Determine the confidence level corresponding to the optimized target parameters based on the predicted performance parameters.
[0114] In this embodiment, the preset conditions include the confidence level reaching a preset value.
[0115] Accordingly, step 210 includes: step 210b, when the confidence level reaches a preset value, stop the iteration, and use the target optimization parameters obtained from the optimized objective function as the final optimized target optimization parameters.
[0116] After each iteration of optimization, based on the prediction performance parameters corresponding to each variable output by the objective function of the Gaussian process model completed in the latest iteration (each prediction performance parameter can contain at least one value), the confidence scores of each value contained in the prediction performance parameters and each value contained in the pre-set target performance parameters are calculated, and it is determined whether the confidence score of any variable reaches the preset value. If not, the next iteration continues. If so, the sample data contained in the variables whose confidence scores reach the preset value are used as the final optimized result of each target optimization parameter.
[0117] In the circuit optimization method described above, the multi-point prediction approach eliminates the need for new input data and re-estimation of hyperparameters, allowing for the optimization and expansion of input variables. This facilitates the search for the globally optimal variable design, accelerates the iterative optimization of the objective function, reduces the number of electromagnetic simulation iterations, and shortens the circuit optimization cycle. Furthermore, by pre-setting iteration thresholds or confidence levels, the iteration stopping condition of the model is established, ensuring that the final optimized target parameters better meet actual requirements. It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction for the execution of these steps; they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or stages. These steps or stages are not necessarily completed at the same time but can be executed at different times. The execution order of these steps or stages is not necessarily sequential but can be performed alternately or in turn with other steps or at least a portion of steps or stages in other steps.
[0118] Based on the same inventive concept, this application also provides a memory data access device for implementing the circuit optimization method described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more circuit optimization device embodiments provided below can be found in the limitations of the circuit optimization method described above, and will not be repeated here.
[0119] In one embodiment, such as Figure 7 As shown, a circuit optimization device 700 is provided, including: a determination module 702, a construction module 704, a prediction module 706, an optimization module 708, and a judgment module 710, wherein:
[0120] The determining module 702 is used to determine at least one set of target optimization parameters from the parameters to be optimized corresponding to the circuit to be optimized in response to the optimization command;
[0121] The construction module 704 is used to construct a Gaussian process model using at least one set of sample data corresponding to each objective optimization parameter, and to construct an acquisition function, an expected boosting function and a probability boosting function based on the Gaussian process model;
[0122] The prediction module 706 is used to perform multi-point prediction processing based on the prediction performance data corresponding to the sample data output by the Gaussian process model, through the acquisition function, expectation boosting function and probability boosting function, to obtain optimized data;
[0123] The optimization module 708 is used to reconstruct the model based on sample data, prediction performance data, and optimization data, and to iteratively optimize the objective function of the Gaussian process model.
[0124] The judgment module 710 is used to stop the iteration when the iteratively optimized objective function meets the preset conditions, and to obtain the final optimized objective parameters based on the iteratively optimized objective function.
[0125] In some optional embodiments, the construction module 704 is also configured to:
[0126] Collect at least one set of sample data corresponding to each objective optimization parameter.
[0127] In some optional embodiments, the sample data includes input sample data and simulation performance data corresponding to the input sample data;
[0128] Prediction module 706 is also configured as follows:
[0129] Using the acquisition function, expectation boosting function, and probability boosting function, new input sample data are determined based on the prediction performance data corresponding to the input sample data output by the Gaussian process model.
[0130] Based on the simulation performance data, determine the simulation performance data corresponding to the new input sample data;
[0131] The input sample data and the corresponding simulation performance data, as well as the new input sample data and the corresponding simulation performance data, are used as optimization data.
[0132] In some optional embodiments, the prediction module 706 is further configured to:
[0133] Use the minimum, average, or maximum value in the simulation performance data as the simulation performance data corresponding to the new input sample data.
[0134] In some optional embodiments, the optimization module 708 is further configured to:
[0135] Increment the number of optimizations corresponding to the objective function by one.
[0136] In some optional embodiments, the preset conditions include reaching a preset optimization threshold number of times;
[0137] The judgment module 710 is also configured as follows:
[0138] When the number of optimization attempts reaches the preset threshold, the iteration stops, and the final optimized target parameters are determined based on the optimized objective function.
[0139] In some optional embodiments, the optimization module 708 is further configured to: obtain the prediction performance parameters of the optimized objective function output;
[0140] Based on the predicted performance parameters, determine the confidence level corresponding to the optimized target parameters.
[0141] In some optional embodiments, the preset conditions include a confidence level reaching a preset value;
[0142] The judgment module 710 is also configured as follows:
[0143] When the confidence level reaches the preset value, the iteration stops, and the target optimization parameters obtained from the optimized objective function are used as the final optimized target optimization parameters.
[0144] Each module in the aforementioned circuit optimization device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.
[0145] As an example, it can be used for things like Figure 8 The circuit shown was subjected to actual electromagnetic simulation. Twenty-eight sets of initial data W1~W14 and L1~14 were randomly selected as the initial variables for the Gaussian process model. The specific initial values of the 28 sets of initial data W1~W14 and L1~14 are shown in Table 1.
[0146] Table 1
[0147]
[0148] like Figure 9 The figure shows the changes in the actual initial saturation gain, initial saturation output power, and initial power-added efficiency under different electromagnetic wave frequency ranges when performing electromagnetic simulations using the 28 sets of initial data W1~W14 and L1~14.
[0149] Subsequently, using the circuit optimization method described in any of the above embodiments, the 28 sets of data W1~W14 and L1~14 are used as target optimization parameters. After single-point prediction optimization, the target optimization parameters are obtained, namely the optimized W1~W14 and the new L1~14, which are denoted here as W1'~W14' and L1'~14'. The specific values of the 28 sets of single-point prediction optimized data W1'~W14' and L1'~14' are shown in Table 2:
[0150] Table 2
[0151]
[0152] Using the circuit optimization method described in any of the above embodiments, the 28 sets of data W1~W14 and L1~14 are used as target optimization parameters. After multi-point prediction optimization, the target optimization parameters are obtained, namely the optimized W1~W14 and the new L1~14, which are denoted here as W1''~W14'' and L1''~14''. The specific values of the 28 sets of multi-point prediction optimized data W1''~W14'' and L1''~14'' are shown in Table 3:
[0153] Table 3
[0154]
[0155] like Figure 10 The diagram shows the number of iterations required for the objective function to converge when performing electromagnetic simulations on variables predicted by 28 single points (W1'~W14' and L1'~14') and 28 multi-point predictions (W1''~W14'' and L1''~14''). It can be seen that the number of iterations required for variables predicted by three points is less than the number of iterations required for variables predicted by single points.
[0156] Furthermore, a comparison of the electromagnetic simulation time costs after single-point prediction and multi-point prediction was made. Based on the actual electromagnetic simulation results, the time for each single-point prediction iteration was: (50+1) electromagnetic simulations consumed 24.48 minutes, algorithm time consumed 1.66 minutes, and the total time for convergence (by the 23rd iteration) was 601.22 minutes (10.02 hours). The time for each three-point prediction iteration was: (50+3) electromagnetic simulations consumed 25.44 minutes, algorithm time consumed 3.08 minutes, and the total time for convergence (by the 17th iteration) was 484.84 minutes (8.081 hours). For typical electromagnetic optimization designs, this reduces electromagnetic simulation optimization time by approximately 2 hours, thus reducing circuit design time.
[0157] Table 3 shows the electromagnetic wave frequency, saturation gain, saturation output power, saturation power added efficiency, and iteration convergence time used in electromagnetic simulation of variables after single-point and multi-point prediction.
[0158] Table 4
[0159]
[0160] like Figure 11 As shown in the figure, the power-added efficiency varies across different electromagnetic wave frequency ranges during electromagnetic simulations, based on the variables obtained from single-point and multi-point predictions. The power-added efficiency is higher for the variables obtained from three-point predictions compared to those obtained from single-point predictions.
[0161] like Figure 12 As shown in the figure, the saturation gain data changes of the variables after single-point prediction and multi-point prediction in electromagnetic simulation under different electromagnetic wave frequency ranges. It can be seen that when the electromagnetic wave frequency is high, the saturation gain of the variables obtained by single-point prediction is higher than that obtained by single-point prediction.
[0162] like Figure 13 As shown in the figure, the saturated output power data changes of the variables after single-point prediction and multi-point prediction in electromagnetic simulation under different electromagnetic wave frequency ranges. It can be seen that when the electromagnetic wave frequency is high, the saturated output power corresponding to the variables obtained by single-point prediction is higher than that obtained by multi-point prediction.
[0163] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 14As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a circuit optimization method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.
[0164] Those skilled in the art will understand that Figure 14 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0165] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the circuit optimization method as described in any of the above embodiments.
[0166] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the circuit optimization method as described in any of the above embodiments.
[0167] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0168] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0169] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A circuit optimization method, characterized in that, include: In response to the optimization command, at least one set of target optimization parameters is determined from the parameters to be optimized corresponding to the circuit to be optimized; A Gaussian process model is constructed using at least one set of sample data corresponding to each of the aforementioned target optimization parameters, and an acquisition function, an expected boosting function, and a probability boosting function are constructed based on the Gaussian process model; the sample data includes input sample data and simulation performance data corresponding to the input sample data; Using the acquisition function, the expected boosting function, and the probability boosting function, new input sample data are determined based on the prediction performance data corresponding to the input sample data output by the Gaussian process model. Based on the simulation performance data and the fixed values, determine the simulation performance data corresponding to the new input sample data; The input sample data and the simulation performance data corresponding to the input sample data, the new input sample data and the simulation performance data corresponding to the new input sample data are used as optimization data; Based on the sample data, the prediction performance data, and the optimization data, the model is reconstructed, and the objective function of the Gaussian process model is iteratively optimized. When the iteratively optimized objective function meets the preset conditions, the iteration stops, and the final optimized objective parameters are obtained based on the iteratively optimized objective function.
2. The method according to claim 1, characterized in that, Before constructing the Gaussian process model using at least one set of sample data corresponding to each of the aforementioned target optimization parameters, the method further includes: Collect at least one set of sample data corresponding to each of the target optimization parameters.
3. The method according to claim 2, characterized in that, The step of determining the simulation performance data corresponding to the new input sample data based on the simulation performance data and the fixed value includes: The minimum, average, or maximum value in the simulation performance data is used as the simulation performance data corresponding to the new input sample data.
4. The method according to claim 1, characterized in that, After reconstructing the model based on the sample data, the prediction performance data, and the optimization data, and iteratively optimizing the objective function of the Gaussian process model, the process further includes: Increment the number of optimizations corresponding to the objective function by one.
5. The method according to claim 4, characterized in that, The preset conditions include the number of optimizations reaching a preset optimization threshold. When the iteratively optimized objective function meets a preset condition, the iteration stops, and the final optimized target parameters are obtained based on the iteratively optimized objective function, including: When the number of optimization attempts reaches a preset threshold, the iteration stops, and the final optimized target parameters are determined based on the optimized objective function.
6. The method according to claim 1, characterized in that, After reconstructing the model based on the sample data, the prediction performance data, and the optimization data, and iteratively optimizing the objective function of the Gaussian process model, the process further includes: Obtain the prediction performance parameters output by the optimized objective function; Based on the predicted performance parameters, determine the confidence level corresponding to the optimized target parameters.
7. The method according to claim 6, characterized in that, The preset condition includes the confidence level reaching a preset value; When the iteratively optimized objective function meets a preset condition, the iteration stops, and the final optimized target parameters are obtained based on the iteratively optimized objective function, including: When the confidence level reaches the preset value, the iteration stops, and the target optimization parameters obtained according to the optimized objective function are used as the final optimized target optimization parameters.
8. A circuit optimization device, characterized in that, include: The determination module is used to determine at least one target optimization parameter from the parameters to be optimized corresponding to the circuit to be optimized in response to the optimization command; A construction module is used to construct a Gaussian process model using at least one set of sample data corresponding to each of the target optimization parameters, and to construct an acquisition function, an expected boosting function, and a probability boosting function based on the Gaussian process model; the sample data includes input sample data and simulation performance data corresponding to the input sample data; The prediction module is used to determine new input sample data based on the prediction performance data corresponding to the input sample data output by the Gaussian process model, using the acquisition function, the expectation boosting function, and the probability boosting function. Based on the simulation performance data and fixed values, determine the simulation performance data corresponding to the new input sample data; use the input sample data and the simulation performance data corresponding to the input sample data, as well as the new input sample data and the simulation performance data corresponding to the new input sample data, as optimization data. The optimization module is used to reconstruct the model based on the sample data, the prediction performance data, and the optimization data, and to iteratively optimize the objective function of the Gaussian process model. The judgment module is used to stop the iteration when the iteratively optimized objective function meets the preset conditions, and to obtain the final optimized target parameters based on the iteratively optimized objective function.
9. The apparatus according to claim 8, characterized in that, The construction module is also used to collect at least one set of sample data corresponding to each of the target optimization parameters.
10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the circuit optimization method according to any one of claims 1 to 7.
Citation Information
Patent Citations
Multi-Testbench analog circuit optimization method based on Bayesian optimization
CN116484787A
