Method, device and medium for optimizing layout processing model

Abstract

Embodiments of the present disclosure relate to a method, device and medium for optimizing a layout processing model. The method proposed herein includes: determining a plurality of first model samples related to the layout processing model to be optimized, wherein the first model sample among the plurality of first model samples includes at least one parameter of the layout processing model; using a machine learning model, determining a first prediction performance indicator for the layout processing model based on the plurality of first model samples, respectively, to obtain a plurality of first prediction performance indicators; in response to a target prediction performance indicator among the plurality of first prediction performance indicators being lower than a reference performance indicator, determining a plurality of second model samples, wherein the plurality of second prediction performance indicators corresponding to the plurality of second model samples are all higher than the target prediction performance indicator; and determining a target parameter for the layout processing model based at least on the plurality of second model samples. In this way, the convergence speed of the model optimization iteration can be improved.

Description

Technical Field

[0001] Example embodiments of the present disclosure generally relate to the field of integrated circuit technology, and more particularly, to a method, apparatus, and medium for optimizing a layout processing model. Background Art

[0002] In semiconductor manufacturing, photolithography is a key step in transferring circuit patterns to silicon wafers. However, light diffraction, interference, and photoresist reaction can cause pattern deformation or deviation, affecting circuit performance and yield. To solve this problem, a layout processing model was introduced to optimize and correct photolithography patterns and improve chip manufacturing accuracy.

[0003] Before the layout processing model is put into use, the parameters of the layout processing model need to be optimized. This optimization process aims to select the most appropriate combination of parameters from many potential parameters. To achieve this goal, several iterations are required, and in each iteration, the selected parameters need to be simulated and tested to accurately evaluate their performance. However, the implementation of simulation operations in this process takes up a lot of time and computing resources, which to some extent restricts the speed and efficiency of optimizing the layout processing model. Summary of the invention

[0004] In a first aspect of the present disclosure, a method for optimizing a layout processing model is provided, the method comprising: determining a plurality of first model samples related to the layout processing model to be optimized, wherein the first model samples include at least one parameter of the layout processing model; using a machine learning model, determining a first prediction performance indicator for the layout processing model based on the plurality of first model samples, respectively, to obtain a plurality of first prediction performance indicators; in response to a target prediction performance indicator among the plurality of first prediction performance indicators being lower than a reference performance indicator, determining a plurality of second model samples, wherein a plurality of second model samples among the plurality of second model samples are different from at least one first model sample among the plurality of first model samples; and determining a target parameter for the layout processing model based at least on the plurality of second model samples.

[0005] In a second aspect of the present disclosure, an electronic device is provided. The device includes at least one processing unit; and at least one memory, the at least one memory is coupled to the at least one processing unit and stores instructions for execution by the at least one processing unit. When the instructions are executed by the at least one processing unit, the device executes the method for optimizing the layout processing model of the first aspect.

[0006] In a third aspect of the present disclosure, a computer-readable storage medium is provided, wherein a computer program is stored on the computer-readable storage medium, and the computer program can be executed by a processor to implement the method for optimizing a layout processing model according to the first aspect.

[0007] In a fourth aspect of the present disclosure, a computer program product is provided, which includes computer executable instructions, which, when executed by a processor, implement the method for optimizing a layout processing model according to the first aspect of the present disclosure.

[0008] According to an embodiment of the present disclosure, in at least one iteration, a machine learning model is introduced to predict the first prediction performance index of each model sample for the layout processing model. The rapid prediction capability of the machine learning model enables the rapid identification of model samples (such as first model samples) of poor quality (for example, the first prediction performance index is lower than the reference performance index) in this iteration, and the model samples of poor quality are processed (such as screening or replacement, etc.) to obtain multiple second model samples. Compared with the model samples of poor quality, the multiple prediction performance indicators (such as multiple second prediction performance indicators) corresponding to the multiple second model samples are all higher than the target prediction performance index, thereby optimizing the quality of the model samples of this iteration. This method ensures that only relatively high-quality model samples are selected into the new generation of model samples, thereby improving the convergence speed of the entire iterative process, which is conducive to improving the optimization efficiency of the layout processing model.

[0009] It should be understood that the contents described in this content section are not intended to limit the key features or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0011] Figure 1 A schematic diagram illustrating an example environment according to some embodiments of the present disclosure is shown;

[0012] Figure 2 A flowchart showing an example process for model optimization according to some embodiments of the present disclosure is shown;

[0013] Figure 3 A schematic diagram showing an example of multiple iterations according to some embodiments of the present disclosure;

[0014] Figure 4 A schematic diagram showing an example of a cost value decreasing with the number of iterations according to an embodiment of the present disclosure; and

[0015] Figure 5 A block diagram of an electronic device is shown in which one or more embodiments of the present disclosure may be implemented. DETAILED DESCRIPTION

[0016] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0017] It should be noted that the titles of any sections / subsections provided herein are not restrictive. Various embodiments are described throughout this article, and any type of embodiment may be included under any section / subsection. In addition, the embodiments described in any section / subsection may be combined in any manner with any other embodiments described in the same section / subsection and / or different sections / subsections.

[0018] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions may be included below. The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may be included below.

[0019] As briefly described above, when optimizing the parameters of the layout processing model, several iterations are required to find the most suitable combination of parameters. For the sake of clarity, the following takes the layout processing model as the Optical Proximity Correction (OPC) model as an example for explanation.

[0020] In one solution, a genetic evolution algorithm can be used to perform this optimization. Specifically, the genetic evolution algorithm is an optimization method based on swarm intelligence, which realizes intelligent search through cooperation and competition among individuals in the group. The genetic evolution algorithm not only retains the global search strategy based on population, but also adopts real number coding and simple mutation operations based on differentials, combined with a "one-to-one" competitive survival mechanism, thereby simplifying the complexity of evolutionary computing. At the same time, the differential evolution algorithm has a unique memory function, which can track the current search progress in real time and flexibly adjust the search strategy, showing strong global convergence ability and robustness.

[0021] The execution process of the genetic evolution algorithm is as follows:

[0022] First, control parameters of the genetic evolution algorithm are set, including but not limited to population size, mutation operation factor, crossover operation factor, maximum evolutionary generations, and convergence conditions.

[0023] Next, the initial population is randomly generated. In the scenario of optimizing the parameters of the OPC model, each individual in the population represents a model sample, and each model sample includes at least one parameter of the OPC model. At this stage, the parameters of the OPC model can be randomly initialized to generate multiple model samples as the initial population. In addition, the current iteration number k is set to 1.

[0024] Subsequently, the initial population is evaluated. At this stage, the cost value (e.g., cost value) of each model sample for the OPC model can be calculated through simulation operations. With the help of the cost value, the performance index (this index can also be called simulation performance index) of the OPC model after the corresponding model sample is applied to the OPC model can be indicated.

[0025] After the evaluation is completed, it is determined whether the evaluation results meet the convergence conditions or whether the maximum evolutionary generations have been reached. If these conditions are met, the genetic evolution algorithm will stop evolving and output the best individual (a model sample) currently found as the target parameter of the OPC model. Otherwise, the genetic evolution algorithm will continue to optimize.

[0026] Specifically, mutation (eg, changing parameter values ​​in the model samples) and crossover (eg, exchanging parameter values ​​in the multiple model samples) operations are performed on the multiple model samples in the initial population, thereby generating multiple new model samples to constitute a temporary population.

[0027] Then, the temporary population is evaluated. Similar to the evaluation method described above, at this stage, the cost value of each individual (new multiple model samples) in the temporary population can be calculated through simulation operations to obtain the corresponding simulation performance indicators. In this scenario, the temporary population can also be called the offspring population of the initial population.

[0028] After the evaluation is completed, the individuals in the temporary population can be compared with the corresponding individuals in the initial population, and the better individuals can be selected to enter the next generation population (new generation model sample) according to the size of the cost value.

[0029] Finally, the iteration number k is updated so that k = k + 1. Then the new generation model samples are used as the initial population for the next iteration to perform the next iteration.

[0030] However, in the above process, calculating the cost value of each individual through simulation operation is a computationally intensive task, which is time-consuming and requires a large amount of computing resources.

[0031] In addition, in the above-mentioned iterative process, due to the limitations of the genetic evolution algorithm, each generation of model samples will produce a part of model samples with poor simulation performance indicators. These model samples may be selected into the new generation of model samples, which not only wastes computing resources, but also may affect the convergence speed of the algorithm and the final optimization effect.

[0032] In view of this, the present disclosure provides a scheme for optimizing a layout processing model. According to the scheme of the present disclosure, a plurality of first model samples related to the layout processing model to be optimized are first determined, wherein the first model samples include at least one parameter of the layout processing model. Then, using a machine learning model, a first prediction performance indicator for the layout processing model is determined based on the plurality of first model samples, respectively, to obtain a plurality of first prediction performance indicators. Next, in response to a target prediction performance indicator among the plurality of first prediction performance indicators being lower than a reference performance indicator, a plurality of second model samples are determined, wherein the plurality of second model samples correspond to a plurality of second prediction performance indicators, and the plurality of second prediction performance indicators are all higher than the target prediction performance indicator. Furthermore, at least based on the plurality of second model samples, a target parameter for the layout processing model is determined.

[0033] It will be more clearly understood through the following description that according to the scheme of the present disclosure, in at least one iteration, a machine learning model is introduced to predict the first prediction performance index of each model sample for the layout processing model. The rapid prediction capability of the machine learning model enables the rapid identification of model samples (such as first model samples) of poor quality (for example, the first prediction performance index is lower than the reference performance index) in this iteration, and the model samples of poor quality are processed (such as screening or replacement, etc.) to obtain multiple second model samples. Compared with the model samples of poor quality, the multiple prediction performance indicators (such as multiple second prediction performance indicators) corresponding to the multiple second model samples are all higher than the target prediction performance index, thereby optimizing the quality of the model samples of this iteration. This method ensures that only relatively high-quality model samples are selected into the new generation of model samples, thereby improving the convergence speed of the entire iterative process, which is conducive to improving the optimization efficiency of the layout processing model.

[0034] Various example implementations of the solution will be described in detail below in conjunction with the accompanying drawings.

[0035] Figure 1 A schematic diagram of an example environment 100 is shown according to some embodiments of the present disclosure. Figure 1 , the example environment 100 may include an electronic device 120 and a layout processing model 140 .

[0036] In the example environment 100, the electronic device 120 may be used to optimize the parameters of the layout processing model 140 (e.g., the OPC model described above). As an example, the layout processing model 140 may include any relevant processing model in the layout processing process. For example, the layout processing model may include, but is not limited to, an OPC model, a Mask Process Correction (MPC) model, and an electron beam lithography model (Ebeam).

[0037] As an example, the OPC model is mainly used in layout processing to correct imaging distortion problems caused by the optical proximity effect. The optical proximity effect is caused by the diffraction of light. When the feature size of the mask pattern is close to or smaller than the wavelength of light used by the lithography system, the imaging on the wafer will deviate from the original design pattern. This effect can cause distortion problems such as edge rounding, line end indentation, and line width deviation, which in turn affects the reliability and performance of the chip. The OPC model compensates for these optical effects by adjusting the shape and size of the mask pattern. Specifically, it may change the topology of the light-transmitting area on the mask, or add small sub-resolution auxiliary patterns to the mask to make the lithography imaging results as close as possible to the original design pattern. This correction process is critical to ensuring the reliability and accuracy of the lithography pattern.

[0038] As an example, the MPC model is mainly used in layout processing to solve problems such as electron beam blurring, photoresist diffusion, and lateral etching deviation in the mask process. The MPC model aims to predict and correct these nonlinear effects through a series of algorithms and models to ensure that the circuit layout on the mask is as consistent as possible with the mask layout after optical proximity correction (OPC). The MPC model takes into account factors such as electron beam energy distribution, chemical and physical properties of photoresist materials, and various parameters in the etching process. By simulating these processes, the MPC model can predict the impact of nonlinear effects on the pattern and generate corresponding correction data. These data are then used to adjust the mask layout or process parameters to minimize the deviation caused by nonlinear effects.

[0039] As an example, the Ebeam model is a process model related to electron beam lithography. Electron beam lithography is a technology that uses a high-energy electron beam to draw nanometer-scale patterns directly on silicon wafers or other substrates. Compared with optical lithography, electron beam lithography has higher resolution and accuracy, so it is suitable for manufacturing more advanced and smaller integrated circuits. In layout processing, the Ebeam model can be used to simulate and optimize the electron beam lithography process to ensure high-quality and high-precision patterns on silicon wafers.

[0040] For the sake of clarity, unless otherwise specified, the following description will take the layout processing model 140 being an OPC model as an example to illustrate various embodiments of the present disclosure.

[0041] As an example, the layout processing model 140 has multiple parameters, including but not limited to lithography wavelength and numerical aperture, etc. To optimize these parameters, in the example environment 100, the electronic device 120 can use a genetic evolution algorithm such as described above to find the most appropriate combination of parameters through multiple iterations.

[0042] As an example, the electronic device 120 may randomly generate a series of model samples to obtain a plurality of initial model samples 110 (the plurality of initial model samples 110 may also be referred to as an initial population of a first iteration). Exemplarily, each initial model sample includes at least one parameter of the layout processing model 140, and the parameter values ​​in different initial model samples may be different.

[0043] Subsequently, the electronic device 120 may use the multiple initial model samples 110 as a starting point to search for the target parameter 130 through multiple iterations. Different from the iterative process described above, the embodiment of the present disclosure introduces a machine learning model in at least one iteration. The embodiment of the present disclosure uses a machine learning model to determine a first prediction performance indicator of each model sample (e.g., multiple first model samples) before the simulation step. Based on the first prediction performance indicator, the electronic device 120 may process (e.g., screen out or replace, etc.) model samples with poor quality before the simulation starts to obtain a set of model samples with better quality (e.g., multiple second model samples).

[0044] Next, the electronic device 120 performs a simulation operation on the plurality of second model samples to determine a simulation performance indicator (eg, a first simulation performance indicator to be mentioned below) of each second model sample.

[0045] Subsequently, the electronic device 120 determines the convergence condition based on the simulation performance index of each second model sample. If the simulation performance indexes of the plurality of second model samples meet the convergence condition, the electronic device 120 may determine the target parameter 130 of the layout processing model 140 based on the second model sample. If the simulation performance indexes of the respective second model samples do not meet the convergence condition, the electronic device may generate a new model sample (e.g., a plurality of third model samples to be mentioned below) based on the plurality of second model samples through operations such as mutation and crossover.

[0046] After generating a plurality of third model samples, the electronic device 120 may simulate the plurality of third model samples to determine a simulation performance index of each third model sample.

[0047] Finally, the electronic device 120 selects a new generation model sample from the multiple second model samples and the multiple third model samples based on the simulation performance indicators of the multiple second model samples and the simulation performance indicators of the multiple third model samples, and uses the selected new generation model samples as input for the next iteration to perform the next iteration.

[0048] After multiple such iterations, the electronic device 120 can finally determine the target parameters 130 for the layout processing model 140 .

[0049] Once the target parameters 130 of the layout processing model 140 are found, they will be applied to the layout processing model 140, so that the layout processing model 140 can perform the corresponding tasks with higher efficiency and accuracy.

[0050] In some embodiments, the user can interact with the electronic device 120 through a terminal device. As an example, the terminal device can be any type of mobile terminal, fixed terminal or portable terminal, including a mobile phone, a desktop computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, a media computer, a multimedia tablet, a personal communication system (PCS) device, a personal navigation device, a personal digital assistant (PDA), an audio / video player, a digital camera / camcorder, a positioning device, a television receiver, a radio broadcast receiver, an e-book device, a gaming device, or any combination of the foregoing, including accessories and peripherals of these devices or any combination thereof. In some embodiments, the terminal device can also support any type of interface for the user (such as a "wearable" circuit, etc.).

[0051] The electronic device 120 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks, and big data and artificial intelligence platforms. The electronic device 120 may include, for example, a computing system / server, such as a mainframe, an edge computing node, a computing device in a cloud environment, and the like.

[0052] As an example, a communication connection may be established between the electronic device 120 and the terminal device. The communication connection may be established in a wired manner or a wireless manner. The communication connection may include, but is not limited to, a Bluetooth connection, a mobile network connection, a universal serial bus connection, a wireless fidelity connection, etc., and the embodiments of the present disclosure are not limited in this respect. In the embodiments of the present disclosure, the electronic device 120 and the terminal device may realize the interaction of data and / or signaling through the communication connection between the two.

[0053] It should be understood that the structure and function of the various elements in the environment 100 are described for exemplary purposes only and do not imply any limitation on the scope of the present disclosure.

[0054] Figure 2 A flow chart of an example process 200 for model optimization according to some embodiments of the present disclosure is shown. The process 200 may be implemented at the electronic device 120. Figure 3 A schematic diagram of an example 300 of multiple iterations according to some embodiments of the present disclosure is shown. Figure 3 The process 200 is described.

[0055] In block 210 , the electronic device 120 determines a plurality of first model samples related to a layout process model 140 to be optimized, wherein a first model sample of the plurality of first model samples includes at least one parameter of the layout process model 140 .

[0056] As described above, the electronic device 120 can use a genetic evolution algorithm, for example, to find the target parameter 130 of the layout processing model 140 through multiple iterations. Figure 3 The multiple iterations include at least the 1st iteration shown by box 310-1, the 2nd iteration shown by box 310-2, the Nth iteration shown by box 310-N, and the N+1th iteration shown by box 310-N+1, etc., where N can be any appropriate positive integer.

[0057] As an example, the process 200 may be implemented in any one of the blocks 310 - 1 to 310 -N+1. For clarity, unless otherwise specified, the following describes various embodiments of the present disclosure by taking the process 200 being implemented in block 310 -N as an example.

[0058] As an example, the first model samples may refer to Figure 3 The M model samples 320 - 1 to 320 -M are shown, where M is any appropriate positive integer. Such a first model sample may be determined by the electronic device 120 based on a model sample outputted in a previous iteration.

[0059] As an example, each first model sample includes at least one parameter of the layout process model 140 , and in different first model samples, the parameter value of the at least one parameter is different.

[0060] In box 220, the electronic device 120 uses a machine learning model to determine a first prediction performance indicator for the layout processing model 140 based on the multiple first model samples respectively to obtain multiple first prediction performance indicators.

[0061] As an example, through the machine learning model, the electronic device 120 can determine the first predicted performance indicator corresponding to each first model sample. The first predicted performance indicator can refer to a measure of the expected performance of the layout processing model 140 after applying the parameters in the first model sample. Compared with the simulation performance indicator, the first predicted performance indicator is determined by the machine learning model and therefore has a faster generation speed.

[0062] In some embodiments, the plurality of first model samples are inputs to a first iteration of a plurality of iterations related to optimizing the target parameter 130 of the layout processing model 140, and the machine learning model is trained based on a plurality of fourth model samples and a plurality of second simulation performance indicators respectively corresponding to the plurality of fourth model samples. As an example, the plurality of fourth model samples are determined based on the model samples involved in the third iteration before the first iteration, and the plurality of second simulation performance indicators are simulation performance indicators for the layout processing model 140 respectively determined based on the plurality of fourth model samples.

[0063] As an example, the first iteration may refer to Figure 3 Any iteration from the 1st iteration to the N+1th iteration. For clarity, unless otherwise specified, the following describes various embodiments of the present disclosure by taking the first iteration as the Nth iteration as an example. In this case, the third iteration may refer to one or more iterations from the 1st iteration to the N-1th iteration, and the fourth model sample may be a model sample involved in the 1st iteration to the N-1th iteration.

[0064] It should be noted that, in the embodiments of the present disclosure, "model samples involved in iteration" or similar expressions may specifically refer to model samples that have undergone at least the simulation operation described above in the iteration. Such model samples correspond to simulation performance indicators (for example, each fourth model sample corresponds to a second simulation performance indicator).

[0065] In this way, the electronic device 120 can use the simulation results from the 1st iteration to the N-1th iteration as training data for the machine learning model to help the machine learning model learn how to determine the first prediction performance indicator in a manner similar to the simulation operation.

[0066] As an example, in the N-1th iteration, it is assumed that the training data is (x, y), where x represents a model sample and y represents the corresponding simulation performance index. As an example, the simulation performance index can be represented based on the cost value described above. Exemplarily, the level of the simulation performance index is negatively correlated with the size of the cost value. For example, the larger the cost value, the lower the simulation performance index, and the worse the quality of the corresponding model sample. Conversely, the lower the cost value, the higher the simulation performance index, and the better the quality of the corresponding model sample.

[0067] Using the training data (x, y), the electronic device 120 trains the machine learning model. In this way, the machine learning model learns to predict the corresponding cost value according to the input model sample in a simulation operation manner.

[0068] In box 230, in response to a target prediction performance indicator among the multiple first prediction performance indicators being lower than a reference performance indicator, the electronic device 120 determines multiple second model samples, wherein the multiple second model samples correspond to multiple second prediction performance indicators, and the multiple second prediction performance indicators are all higher than the target prediction performance indicator.

[0069] As an example, the second predicted performance indicator may refer to a measure of the expected performance of the layout processing model 140 after applying the parameters in the second model sample. Optionally, the second predicted performance indicator may be determined by the machine learning model described above. Compared with the simulation performance indicator, the second predicted performance indicator is determined by the machine learning model and therefore has a faster generation speed.

[0070] In some embodiments, the plurality of second model samples may be determined in at least one of the following ways.

[0071] In the first way, the multiple second model samples can be pre-set candidate model samples. In this way, the multiple second model samples can be candidate samples prepared in advance. These candidate samples may be set based on previous experience, historical data or other relevant analysis. As potential alternative options, they can be directly selected from the candidate library when needed to meet specific prediction performance indicator requirements. This method is suitable for those situations where there are already rich data resources and previous experience, and can quickly select suitable second model samples from the predefined sample set.

[0072] In the second method, the plurality of second model samples may be model samples obtained by replacing the first model samples corresponding to the target prediction performance index in the plurality of first model samples. In this method, the generation of the second model samples involves a replacement operation on the first model samples. Specifically, when it is found that some first model samples perform poorly on the target prediction performance index, the electronic device 120 selectively replaces these model samples. The replacement may be based on a variety of strategies, such as new samples after adjusting parameters. In this way, the electronic device 120 can improve the model samples in a targeted manner in order to achieve or exceed the preset target prediction performance index.

[0073] In the third method, the plurality of second model samples may be the remaining model samples after screening the plurality of first model samples (excluding the first model samples corresponding to the target prediction performance index). In this method, the electronic device may screen out the first model samples that do not reach the target prediction performance index and exclude them. The remaining samples, i.e., those that perform well and meet or exceed the target prediction performance index, will be selected as the second model samples. This method ensures that the final selected second model sample set has a higher overall prediction performance by excluding samples with poor performance.

[0074] It should be noted that the selection of the above three methods can be determined according to actual needs, as long as the multiple second prediction performance indicators of the multiple second model samples are all higher than the target prediction performance indicator.

[0075] It should also be noted that, according to actual needs, multiple second model samples can also be determined in other ways, and the embodiments of the present disclosure are not limited to this.

[0076] In some embodiments, the reference performance index is determined based on a pre-set index value. These pre-set index values ​​can be determined based on comprehensive factors such as actual needs, industry standards, historical data, and expert advice. Such a reference performance index not only has a clear quantitative standard, but also can fully reflect the performance status of the evaluation object.

[0077] In some embodiments, the reference performance indicator may also be determined based on one or more iterations from the 1st iteration to the N-1th iteration.

[0078] Specifically, the electronic device 120 determines the reference performance indicator based on the indicator values ​​of the plurality of third simulation performance indicators respectively corresponding to the plurality of fifth model samples. As an example, the plurality of fifth model samples are determined based on the model samples involved in the third iteration before the first iteration, and the plurality of third simulation performance indicators are simulation performance indicators for the layout processing model 140 respectively determined based on the plurality of fifth model samples.

[0079] As an example, the electronic device 120 may divide the entire iteration process (eg, 200 iterations) into two stages, for example, the first stage is from the 1st iteration to the 50th iteration, and the second stage is from the 51st iteration to the 200th iteration.

[0080] In the first stage, the electronic device 120 collects the model samples (such as the fourth model sample or the fifth model sample) and the corresponding simulation performance indicators (such as the second simulation performance indicator or the third simulation performance indicator) after each iteration of the simulation operation. These data are collected and can be used as training data for the machine learning model. On the other hand, these data can also be used to determine the reference performance indicator.

[0081] As an example, assume that the number of parameter samples that are simulated in each iteration is 50. In the first 50 iterations, the total number of model samples that are simulated is 50×200=10,000. Using these 10,000 model samples and the corresponding 10,000 performance indicators as a basis, it can ensure that the machine learning model is fully and effectively trained, thereby improving its prediction accuracy and generalization ability. At the same time, the reference performance indicators determined based on these data will be more scientific and reasonable, thus providing a reliable reference standard for performance evaluation and optimization.

[0082] In some embodiments, the electronic device 120 sorts the plurality of third simulation performance indicators based on the indicator values ​​of the plurality of third simulation performance indicators. Then, the electronic device determines a target simulation performance indicator from the plurality of third simulation performance indicators based on the sorting results of the plurality of third simulation performance indicators. Next, the electronic device 120 determines a reference performance indicator based on the target simulation performance indicator.

[0083] In some embodiments, the electronic device 120 may sort multiple third simulation performance indicators according to the size of the indicator value. After completing the sorting, the electronic device 120 may select a target simulation performance indicator based on the sorting result. This selection process may be comprehensively judged based on actual needs, the importance of the indicator, and the specific performance of each indicator in the sorting result. As an example, the electronic device 120 may determine the reference performance indicator based on the median, average, or other representative values ​​in the sorting result.

[0084] In this way, the electronic device 120 can dynamically determine the reference performance indicator based on the simulation results in the 1st iteration to the N-1th iteration, so that the reference performance indicator is more in line with actual needs.

[0085] In the second stage, the electronic device 120 uses the trained machine learning model to perform performance prediction on the model sample to be simulated (such as the first model sample) before each iterative simulation stage to determine a first predicted performance indicator related thereto.

[0086] As an example, if the first prediction performance index is higher than (or equal to) the reference performance index, the electronic device 120 may determine that the first prediction performance index is better, and determine that the first model sample associated therewith is a high-quality model sample. If the first prediction performance index is lower than the reference performance index, the electronic device 120 may determine that the first prediction performance index is poor, and determine that the first model sample associated therewith is a low-quality model sample.

[0087] Subsequently, the electronic device 120 may replace these first model samples, for example, retaining high-quality model samples and replacing low-quality model samples with new model samples.

[0088] In some embodiments, the electronic device 120 regenerates the first model sample corresponding to the target prediction performance index among the plurality of first model samples at least once. Then, the electronic device 120 replaces the first model sample corresponding to the target prediction performance index with the regenerated first model sample until the first prediction performance index corresponding to each of the plurality of first model samples is higher than or equal to the reference performance index.

[0089] As an example, see Figure 3 For the model sample 320-1, in block 330-1, the electronic device 120 determines whether its predicted performance index is lower than the reference performance index. If the predicted performance index is higher than or equal to the reference performance index, in block 340-1, the electronic device 120 retains the model sample 320-1; if the predicted performance index is lower than the reference performance index, in block 350-1, the electronic device 120 replaces the model sample 320-1, and then returns to block 330-1 to determine again whether the predicted performance index of the replaced model sample 320-1 is lower than the reference performance index.

[0090] For the model sample 320-M, in box 330-M, the electronic device 120 determines whether its predicted performance index is lower than the reference performance index. If the predicted performance index is higher than or equal to the reference performance index, in box 340-M, the electronic device 120 retains the model sample 320-M; if the predicted performance index is lower than the reference performance index, in box 350-M, the electronic device 120 replaces the model sample 320-M, and then returns to box 330-M to again determine whether the predicted performance index of the replaced model sample 320-M is lower than the reference performance index.

[0091] By analogy, the electronic device 120 performs the above operation on each model sample in the model samples 320-1 to 320-M until the prediction performance index of each model sample in the model samples 320-1 to 320-M is higher than or equal to the reference performance index. At this time, the electronic device 120 determines that the model samples 320-1 to 320-M are all high-quality model samples.

[0092] As an example, the electronic device 120 may perform operations such as parameter variation and / or parameter crossover described above on the model samples outputted by the N-1th iteration to generate learned model samples for replacing inferior model samples.

[0093] Finally, the electronic device 120 determines these high-quality model samples as a plurality of second model samples.

[0094] Based on this approach, the electronic device 120 can quickly identify poor-quality model samples before the simulation operation, and replace them through the replacement operation, thereby avoiding unnecessary simulation of the poor-quality model samples in the subsequent simulation operation. In this way, the electronic device 120 saves the time of simulation operation to calculate poor-quality individuals. Thus, the iteration acceleration of the genetic evolution algorithm is achieved, and the final effect is optimized.

[0095] At block 240 , target parameters 130 for the layout process model 140 are determined based on at least the plurality of second model samples.

[0096] As described above, the plurality of second model samples are constructed by the electronic device 120 based on the superior model samples. Such second model samples will have better performance indicators and are more likely to reach the iterative convergence condition. In addition, even if the iterative convergence condition is not reached, these second model samples are also helpful in generating a better new generation of model samples.

[0097] In some embodiments, the electronic device 120 determines the first simulation performance index for the layout processing model 140 based on the plurality of second model samples, respectively, to obtain a plurality of first simulation performance indicators respectively corresponding to the plurality of second model samples. Then, in response to the target simulation performance indicator among the plurality of first simulation performance indicators satisfying a predetermined convergence condition, the electronic device 120 determines the target parameter 130 based on the second model sample corresponding to the target simulation performance indicator.

[0098] In some embodiments, the electronic device 120 may use the simulation operation described above to determine the simulation performance index (i.e., the first simulation performance index) of each second model sample. These first simulation performance indexes measure the simulation performance of the layout processing model 140 under the corresponding second model sample. As an example, the first simulation performance index may be represented by the simulation value described above.

[0099] Next, the electronic device 120 may check whether there is a target simulation performance indicator among these first simulation performance indicators that meets a predetermined convergence condition. As an example, the convergence condition may be a set of preset criteria for determining whether the simulation performance indicator has reached an expected level or has stabilized. When a first simulation performance indicator meets these conditions, it may be regarded as a target simulation performance indicator.

[0100] Once the target simulation performance index that meets the convergence condition is found, the electronic device 120 can determine a second model sample corresponding to the target simulation performance index. Since the second model sample performs well in the simulation operation, it can be considered as the target parameter 130 of the layout processing model 140.

[0101] In some embodiments, the plurality of first model samples are inputs to a first iteration of a plurality of iterations related to the optimization target parameter 130 of the layout processing model 140. In response to each of the plurality of first simulation performance indicators not satisfying a predetermined convergence condition, the electronic device 120 generates a plurality of third model samples based on the plurality of second model samples using a predetermined operation, wherein at least one third model sample in the plurality of third model samples is different from at least one second model sample in the plurality of second model samples. Then, the electronic device 120 determines the output of the first iteration based on the plurality of second model samples and the plurality of third model samples. Next, the electronic device 120 uses the output of the first iteration as an input to a second iteration after the first iteration to perform the second iteration.

[0102] In some embodiments, the electronic device 120 can generate multiple third model samples based on multiple second model samples by using a predetermined operation (such as deforming the parameter values ​​in the multiple model samples). In this process, at least one third model sample will be different from the original second model sample, and this difference is to introduce new search directions and possibilities, in order to find better parameter configurations in subsequent iterations.

[0103] In some embodiments, the predetermined operation includes at least one of the following: a parameter variation operation of changing parameter values ​​included in at least two second model samples, or a parameter crossover operation of exchanging parameter values ​​included in at least two second model samples.

[0104] In some embodiments, for the parameter variation operation, the electronic device 120 may select at least two second model samples and change one or more parameter values ​​therein. This change may be random or based on a specific strategy or algorithm. For example, a small disturbance may be added to the original value of a parameter, or a new value may be completely replaced. Through parameter variation, the parameter space of the model sample is expanded, which helps to discover optimal solutions that may have been overlooked.

[0105] In some embodiments, for parameter crossover operation. In this operation, the electronic device 120 can select at least two second model samples and exchange parameter values ​​between them. This exchange can be partial or complete, depending on the optimization strategy and goal. Through parameter crossover, the excellent characteristics of different model samples are combined, and it is possible to generate new samples with better performance.

[0106] It should be noted that the above two operations are not mutually exclusive, but can be used in combination flexibly. In practical applications, the electronic device 120 can dynamically select a suitable operation to execute according to factors such as the diversity of samples and computing resources.

[0107] After generating a plurality of third model samples, the electronic device 120 may comprehensively consider the second model samples and the newly generated third model samples to determine the output of the first iteration.

[0108] As an example, similar to the second model sample, the electronic device 120 can determine the simulation performance indicator for the layout processing model 140 based on multiple third model samples through simulation operations (for clarity, this simulation performance indicator is also referred to as the fourth simulation performance indicator below). In other words, the electronic device 120 uses simulation operations to determine the simulation performance indicator of each third model sample.

[0109] Next, the electronic device 120 may select a better model sample from the plurality of second model samples and the plurality of third model samples as the output of the first iteration based on the first simulation performance index of each second model sample and the fourth simulation performance index of each third model sample.

[0110] Finally, the electronic device 120 uses the output of the first iteration as the output of the second iteration (eg Figure 3 In this way, the entire parameter optimization process forms a closed loop, and through continuous iteration and adjustment, the optimal parameter configuration of the layout processing model 140 is gradually approached. In such an iterative process, each iteration is an improvement and refinement of the previous result until a simulation performance indicator that meets the predetermined convergence condition is found, thereby determining the final target parameter 130.

[0111] Figure 4 FIG. 4 is a schematic diagram showing an example 400 of a cost value decreasing with the number of iterations according to an embodiment of the present disclosure. Figure 4 , curve 410 is a curve showing the decrease of cost value with the number of iterations when the machine learning model is not introduced, and curve 420 is a curve showing the decrease of cost value with the number of iterations after the machine learning model is introduced. Figure 4 It can be seen that after the 20th iteration, curve 420 is significantly lower than curve 410, which shows that the introduction of the machine learning model enables the cost value to decrease at a faster rate (the simulation performance index can rise to the expected index at a faster rate). In other words, in order to achieve the same cost value, the number of iterations can be reduced after the introduction of the machine model, thereby greatly improving the optimization efficiency.

[0112] Based on the description of each of the above embodiments, it can be seen that according to the embodiments of the present disclosure, in at least one iteration, a machine learning model is introduced to predict the first prediction performance index of each model sample for the layout processing model 140. The rapid prediction capability of the machine learning model enables the rapid identification and exclusion of model samples with poor quality (for example, the first prediction performance index is lower than the reference performance index) in this iteration. In addition, the embodiments of the present disclosure optimize the quality of the model samples of each iteration by replacing model samples with poor quality. This replacement method ensures that only relatively high-quality model samples are selected into the new generation of model samples, thereby improving the convergence speed of the entire iterative process, which is conducive to improving the optimization efficiency of the layout processing model 140.

[0113] Figure 5 1 is a block diagram of an electronic device 500 in which one or more embodiments of the present disclosure may be implemented. The electronic device 500 may be used to implement, for example, Figure 1 The electronic device 120 shown. It should be understood that Figure 5 The electronic device 500 shown is merely exemplary and should not constitute any limitation on the functionality and scope of the embodiments described herein.

[0114] Reference Figure 5 , the electronic device 500 is in the form of a general electronic device. The components of the electronic device 500 may include, but are not limited to, one or more processors or processing units 510, a memory 520, a storage device 530, one or more communication units 540, one or more input devices 550, and one or more output devices 560. The processing unit 510 may be an actual or virtual processor and is capable of performing various processes according to a program stored in the memory 520. In a multi-processor system, multiple processing units execute computer executable instructions in parallel to improve the parallel processing capability of the electronic device 500.

[0115] The electronic device 500 typically includes a plurality of computer storage media. Such media may be any available media accessible to the electronic device 500, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 520 may be a volatile memory (e.g., registers, caches, random access memory (RAM)), a non-volatile memory (e.g., a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. The storage device 530 may be a removable or non-removable medium, and may include a machine-readable medium, such as a flash drive, a disk, or any other medium, which may be capable of being used to store information and / or data and may be accessed within the electronic device 500.

[0116] The electronic device 500 may further include additional removable / non-removable, volatile / non-volatile storage media. Figure 5 As shown in , a disk drive for reading or writing from a removable, non-volatile disk (e.g., a "floppy disk") and an optical drive for reading or writing from a removable, non-volatile optical disk may be provided. In these cases, each drive may be connected to the bus (not shown) by one or more data media interfaces. The memory 520 may include a computer program product 525 having one or more program modules that are configured to perform various methods or actions of various embodiments of the present disclosure.

[0117] The communication unit 540 implements communication with other electronic devices through a communication medium. Additionally, the functions of the components of the electronic device 500 can be implemented in a single computing cluster or multiple computing machines that can communicate through a communication connection. Therefore, the electronic device 500 can operate in a networked environment using a logical connection with one or more other servers, a network personal computer (PC), or another network node.

[0118] The input device 560 may be one or more input devices, such as a mouse, a keyboard, a tracking ball, etc. The output device 560 may be one or more output devices, such as a display, a speaker, a printer, etc. The electronic device 500 may also communicate with one or more external devices (not shown) through the communication unit 540 as needed, such as a storage device, a display device, etc., communicate with one or more devices that allow a user to interact with the electronic device 500, or communicate with any device that allows the electronic device 500 to communicate with one or more other electronic devices (e.g., a network card, a modem, etc.). Such communication may be performed via an input / output (I / O) interface (not shown).

[0119] According to an exemplary implementation of the present disclosure, a computer-readable storage medium is provided, on which computer-executable instructions are stored, wherein the computer-executable instructions are executed by a processor to implement the method described above. According to an exemplary implementation of the present disclosure, a computer program product is also provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, and the computer-executable instructions are executed by a processor to implement the method described above.

[0120] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices, equipment, and computer program products implemented according to the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0121] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0122] Computer-readable program instructions can be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, so that the instructions executed on the computer, other programmable data processing apparatus, or other device implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0123] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple implementations of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of a module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some implementations as replacements, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be realized by a special hardware-based system that performs the function or action of the specification, or can be realized by a combination of special hardware and computer instructions.

[0124] The above descriptions of various implementations of the present disclosure are exemplary, non-exhaustive, and not limited to the disclosed implementations. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The determination of the terms used herein is intended to best explain the principles of the implementations, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the various implementations disclosed herein.

Claims

1. A method for optimizing a layout processing model, characterized in that: include: Determining a plurality of first model samples related to a layout process model to be optimized, wherein the first model samples include at least one parameter of the layout process model and are inputs to a first iteration of a plurality of iterations related to a layout process model optimization target parameter; Determine, using a machine learning model, first prediction performance indicators for the layout processing model based on the plurality of first model samples, respectively, to obtain a plurality of first prediction performance indicators; In response to a target prediction performance indicator among the plurality of first prediction performance indicators being lower than a reference performance indicator, determining a plurality of second model samples, wherein the plurality of second model samples correspond to a plurality of second prediction performance indicators, and the plurality of second prediction performance indicators are all higher than the target prediction performance indicator; as well as determining target parameters for the layout processing model through simulation operations based at least on the plurality of second model samples, The reference performance indicator is determined based on the indicator values ​​of multiple third simulation performance indicators corresponding to multiple fifth model samples respectively, and the multiple fifth model samples are determined based on the model samples involved in the third iteration before the first iteration, and the multiple third simulation performance indicators are simulation performance indicators for the layout processing model determined based on the multiple fifth model samples respectively.

2. The method for optimizing a layout processing model according to claim 1, characterized in that: Determining target parameters for the layout processing model based at least on the plurality of second model samples includes: Determining first simulation performance indicators for the layout processing model based on the plurality of second model samples respectively, so as to obtain a plurality of first simulation performance indicators respectively corresponding to the plurality of second model samples; and In response to a target simulation performance indicator among the plurality of first simulation performance indicators satisfying a predetermined convergence condition, the target parameter is determined based on a second model sample corresponding to the target simulation performance indicator.

3. The method for optimizing a layout processing model according to claim 2, characterized in that: The plurality of first model samples are input to a first iteration of a plurality of iterations related to optimizing a target parameter of a layout process model, and wherein determining the target parameter for the layout process model based at least on the plurality of second model samples comprises: In response to each of the plurality of first simulation performance indicators not satisfying the predetermined convergence condition, generating a plurality of third model samples based on the plurality of second model samples by using a predetermined operation, wherein at least one third model sample of the plurality of third model samples is different from a plurality of second model samples of the plurality of second model samples; determining an output of the first iteration based on the plurality of second model samples and the plurality of third model samples; and The output of the first iteration is used as an input of a second iteration subsequent to the first iteration to perform the second iteration.

4. The method for optimizing a layout processing model according to claim 3, wherein: The predetermined operation includes at least one of the following: a parameter mutation operation that changes the parameter values ​​included in at least two second model samples, or A parameter crossover operation is performed to exchange parameter values ​​included in at least two second model samples.

5. The method for optimizing a layout processing model according to claim 3, characterized in that: The first prediction performance indicators correspond to the plurality of first model samples respectively, and in response to a target prediction performance indicator among the plurality of first prediction performance indicators being lower than a reference performance indicator, determining the plurality of second model samples comprises: The first model samples corresponding to the target prediction performance indicator among the multiple first model samples are replaced to obtain the multiple second model samples.

6. The method for optimizing a layout processing model according to claim 1, wherein: Replacing the first model sample corresponding to the target prediction performance indicator among the plurality of first model samples includes: Regenerating at least once a first model sample corresponding to the target prediction performance indicator among the plurality of first model samples; and The first model sample corresponding to the target prediction performance index is replaced by the regenerated first model sample until the first prediction performance index corresponding to each of the plurality of first model samples is higher than or equal to the reference performance index.

7. The method for optimizing a layout processing model according to claim 1, characterized in that: The plurality of first model samples are inputs to a first iteration of a plurality of iterations related to a layout processing model optimization target parameter, and the machine learning model is trained based on a plurality of fourth model samples and a plurality of second simulation performance indicators respectively corresponding to the plurality of fourth model samples; and The plurality of fourth model samples are determined based on the model samples involved in the third iteration before the first iteration, and the plurality of second simulation performance indicators are simulation performance indicators for the layout processing model determined based on the plurality of fourth model samples respectively.

8. The method for optimizing a layout processing model according to claim 1, characterized in that: The reference performance index is determined based on a preset index value.

9. The method for optimizing a layout processing model according to claim 1, characterized in that: Determining the reference performance indicator based on a plurality of third simulation performance indicators respectively corresponding to a plurality of fifth model samples comprises: sorting the plurality of third simulation performance indicators based on the indicator values ​​of the plurality of third simulation performance indicators; Based on the sorting results of the plurality of third simulation performance indicators, determining a target simulation performance indicator from the plurality of third simulation performance indicators; and Based on the target simulation performance indicator, the reference performance indicator is determined.

10. An electronic device, characterized in that: include: at least one processing unit; as well as At least one memory, the at least one memory is coupled to the at least one processing unit and stores instructions for execution by the at least one processing unit, the instructions, when executed by the at least one processing unit, cause the electronic device to perform the method for optimizing a layout processing model according to any one of claims 1 to 9.

11. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and the computer program can be executed by a processor to implement the method for optimizing a layout processing model according to any one of claims 1 to 9.

12. A computer program product, characterized in that The method comprises computer executable instructions, wherein the computer executable instructions implement the method for optimizing a layout processing model according to any one of claims 1 to 9 when being executed by a processor.

Images