How to calculate process parameters

Abstract

To provide a method of applying optimization technique-based algorithms and trained predictive models to deduce the unconfirmed process parameters inversely.SOLUTION: A method constructs an optimization model by treating a trained predictive model as a function of process parameters to be determined. Subject to the conditions that the process parameters to be determined must satisfy, either individually or in relation to each other. Moreover, in order to solve the constrained optimization problem, the method with penalty function and barrier function can be used to convert the constrained optimization problem into an unconstrained optimization problem. For the optimization problem, a dataset of the trained predictive model can be used to find multiple samples that meet target results as anchor points to form a multi-dimensional feasible solution space, and set or make initial solutions in this feasible solution space, and then use different numerical analysis methods to find the direction and step size, and the optimal feasible solution is an input parameter that can be used.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a method for calculating process parameters, and particularly relates to a method for inversely estimating unconfirmed process parameters by applying an algorithm based on optimization technology and a trained prediction model.

Background Art

[0002] With the progress of medical engineering technology, currently, the application of regenerative medicine in clinical treatment of diseases is becoming increasingly diversified. Regenerative medicine is mainly a medical technology that repairs damaged tissues and organs by utilizing the regenerative ability of cells, and its application scope is very wide. Also, by combining medical technologies such as tissue engineering and molecular biology, it is expected that diseases that were once considered difficult to treat, such as diabetes, neurological diseases, cardiovascular diseases, and cancer, can be improved and treated. Currently, regenerative medicine is mainly applied to organ repair, immune cell therapy, and stem cell therapy. Also, the research and application of cell therapy in regenerative medicine have been attracting increasing attention from all sectors. Cell therapy is a method in which human cells are cultured or processed outside the body and then transplanted into individual bodies for use.

[0003] In recent years, as many governments have gradually liberalized the application of cell therapy, numerous scholars both domestically and internationally have entered the field of cell therapy research. As a result, cell therapy has made remarkable progress in the treatment of many diseases, such as the treatment of skin defects with autologous fibroblasts, the treatment of knee joint cartilage defects with autologous chondrocytes, and the treatment of spinal cord injury with autologous bone marrow mesenchymal stem cells. Furthermore, the quality of cell therapy products directly impacts the safety and effectiveness of treatment. Therefore, in the cell culture process, it is necessary to strictly control the growth state of cells and to immediately monitor the culture parameters and environmental parameters of cell growth to avoid contamination or deterioration of cell quality during the culture process. In addition, previous research has shown that the variability of cells is extremely high between different cases, so the optimal culture parameters and environmental parameters for cell preparations applied to different cases are not exactly the same. For this reason, it is necessary to design and adjust each process parameter within the process for each cell preparation to achieve the desired results, and it is not possible to produce each cell preparation with fixed process parameters. Furthermore, due to the complexity and correlation of the process, conventional techniques typically only allow for the optimization of a single parameter, neglecting the overall interrelationships between parameters throughout the entire process.

[0004] Conventional techniques involve training a predictive model on a large dataset of samples using machine learning. This predictive model can generate predictive results for cell culture by inputting various process parameters, allowing users to simulate the effects of their designed cell culture process in advance. However, the predictive model trained using machine learning cannot inversely estimate the process parameters of the cell culture process from the user's desired results. In other words, when designing a cell culture process, users need to try a large number of different process parameters, resulting in significant effort and resources being spent on designing and improving the cell culture process. Furthermore, the problem of being unable to inversely estimate input parameters also exists for predictive models obtained through machine learning in fields other than cell processes.

[0005] Generally, a predictive model trained using machine learning can be represented by a function (f(x)). Furthermore, the method for inversely estimating appropriate input parameters from the expected results of the predictive model can be considered an optimization operation of that function. This optimization operation often involves using the derivative of the function; in other words, differentiating the function. However, the objective function of a predictive model trained using machine learning is usually non-symbolic, often non-differentiable, and sometimes even discontinuous. Therefore, it is difficult to calculate the derivative of this function using numerical analysis methods. In other words, it is difficult to directly inversely estimate input parameters using the objective function of a predictive model.

[0006] Therefore, in order to identify and optimize multiple parameters in a process, it is necessary to research and develop a more comprehensive and accurate method that allows for the inverse estimation of process parameters that match the desired results. [Overview of the project] [Problems that the invention aims to solve]

[0007] In view of the above, the present invention provides a method for calculating process parameters in order to solve the conventional problems described above. [Means for solving the problem]

[0008] The present invention provides a method for calculating process parameters used to verify and optimize multiple process parameters in a process. The process parameters include multiple verified input parameters and multiple unverified input parameters. The method for calculating process parameters provides a trained predictive model, which is obtained by machine learning a dataset using a machine learning method, wherein the dataset includes multiple samples, and each sample includes multiple sample parameters, and the trained predictive model is used to take multiple input parameters as input to generate prediction results corresponding to the input parameters; sets a target result corresponding to the prediction results of the trained predictive model and provides the verified input parameters among the input parameters; and compares whether the prediction results generated by inputting each of the samples in the dataset into the trained predictive model match the target result, and whether the prediction results match the target result The method includes the steps of: forming a multidimensional subspace with a plurality of anchor samples that match the result; identifying an initial sample in the multidimensional subspace; obtaining the direction of the initial sample in the multidimensional subspace; increasing the step size of the sample parameter of the initial sample along the direction; inputting the confirmed input parameter and the sample parameter with the increased step size that does not correspond to the confirmed input parameter into the trained prediction model to check whether the generated prediction result matches the target result; and, if the prediction result matches the target result, setting the sample parameter with the increased step size that does not correspond to the confirmed input parameter as the unconfirmed input parameter.

[0009] The step of identifying the initial sample in a multidimensional subspace includes the step of defining the initial sample as the midpoint of the distance between at least two of the anchor samples in the multidimensional subspace.

[0010] The step of identifying the initial sample in a multidimensional subspace includes the step of defining the initial sample as a linear combination of at least two of the anchor samples in the multidimensional subspace.

[0011] The step of obtaining the direction of the initial sample in the multidimensional subspace includes the step of calculating the second directional derivative of the initial sample along each of the anchor samples as the direction of the initial sample by the numerical analysis method.

[0012] The step of obtaining the direction of the initial sample in the multidimensional subspace includes setting the initial sample as the origin and calculating the third direction derivative of the origin toward each anchor sample as the direction of the initial sample by the numerical analysis method.

[0013] The trained prediction model is represented by an objective function. Furthermore, the step of obtaining the direction of the initial sample in the multidimensional subspace includes the step of obtaining the secant direction derivative of the initial sample toward each anchor sample by calculating the change in the function value of the objective function based on the change in the distance of the initial sample toward each anchor sample using the numerical analysis method.

[0014] The aforementioned trained predictive model can be represented by an objective function. Furthermore, the step of obtaining the direction of the initial sample in the multidimensional subspace includes the steps of sequentially increasing and adjusting the step size of the initial sample toward the directional derivative of each anchor sample; after increasing and adjusting the step size of the initial sample toward the directional derivative of each anchor sample, sequentially checking whether the function value of the objective function approaches the target result; after increasing the step size of the initial sample toward the directional derivative of any of the anchor samples, if it is confirmed that the function value of the objective function approaches the target result, the step of increasing the step size of the initial sample toward the directional derivative; and after increasing the step size of the initial sample toward the directional derivative of any of the anchor samples, if it is confirmed that the function value of the objective function does not approach the target result, the step of increasing the step size of the initial sample toward the directional derivative of another anchor sample, and then checking again whether the function value of the objective function approaches the target result.

[0015] The step of increasing the step size along the direction for the sample parameters of the initial sample further includes setting an initial size for the step size, increasing the step size along the direction for the sample parameters of the initial sample by the initial size, inputting it into the trained predictive model, and checking whether the prediction result output from the trained predictive model approaches the target result; if the prediction result output from the trained predictive model approaches the target result, determining the step size to the initial size; if the prediction result output from the trained predictive model does not approach the target result, continuing to adjust the size of the step size until the prediction result approaches the target result, and increasing the step size along the direction for the sample parameters of the initial sample by the size, inputting it into the trained predictive model, and determining the step size to the final size.

[0016] The trained predictive model includes a function of specific waiting process parameters, an objective function, and at least one constraint. The objective function minimizes the difference between the output value of the function of specific waiting process parameters and the target result, and a constrained optimization problem is modeled based on the function of specific waiting process parameters, the objective function, the at least one constraint, and the prediction result. The method for calculating the process parameters obtains the unverified input parameters by solving the optimization problem.

[0017] The method for calculating process parameters in the present invention further includes the step of adding a barrier function to the objective function in order to restrict the process parameters so that they satisfy inequality constraints.

[0018] The method for calculating process parameters in the present invention further includes the step of adding a penalty function to the objective function in order to restrict the process parameters to satisfy the constraints of the equation.

[0019] The dataset further includes a process dataset, the process dataset includes a plurality of process samples, and the sample parameters of each process sample include source parameters and culture parameters.

[0020] The source parameters for each process sample further include attribute data of the source for each process sample. The culture parameters for each process sample further include human, equipment, material, method, and environmental parameters for each process sample.

[0021] The aforementioned trained predictive model is obtained by machine learning the dataset using one of the following: Artificial Neural Networks (ANNs), Convolutional Neural Networks (CNNs), or Recurrent Neural Networks (RNNs).

[0022] The method for calculating process parameters in the present invention further includes the steps of: selecting a second sample in the sample by re-matching the verified input parameters and the sample parameters of the sample if the prediction result does not match the target result; calculating the gradient of the second sample in the objective function of the trained prediction model by the numerical analysis method; increasing the step size of the sample parameters of the second sample along the opposite direction of the gradient, inputting all the sample parameters with increased step sizes into the trained prediction model, and checking whether the generated prediction result matches the target result; and, if the prediction result matches the target result, setting the sample parameters with increased step sizes that do not correspond to the verified input parameters as unverified input parameters.

[0023] In another aspect, the present invention provides a method for calculating process parameters used to verify and optimize multiple process parameters in a process. The process parameters include multiple verified input parameters and multiple unverified input parameters. The method for calculating process parameters provides multiple trained predictive models, each of which is obtained by machine learning a dataset using a machine learning method, the dataset includes multiple samples, each of which includes multiple sample parameters, the trained models each correspond to multiple objective functions, the trained predictive models are used to take multiple input parameters as input to generate multiple prediction results corresponding to the input parameters, the prediction results each belong to different types of target results, the objective functions of the trained predictive models are combined to form a composite objective function, and the input parameters are input to the composite objective function to generate the prediction results, the method for calculating process parameters includes setting multiple target results corresponding to the prediction results of the trained predictive models and providing the verified input parameters among the input parameters, and the method for calculating process parameters The process includes: verifying whether the prediction results generated by inputting each of the samples in the input into the trained prediction model match the target result; forming a multidimensional subspace with a plurality of anchor samples for the samples whose prediction results match the target result; identifying an initial sample in the multidimensional subspace; obtaining the direction of the initial sample in the multidimensional subspace; increasing the step size of the sample parameter of the initial sample along the direction; inputting the verified input parameter and the sample parameter with the increased step size that does not correspond to the verified input parameter into the composite objective function to verify whether the generated prediction result matches the target result; and, if the prediction result matches the target result, setting the sample parameter with the increased step size that does not correspond to the verified input parameter as the unverified input parameter.

[0024] The method for calculating process parameters in the present invention further includes the step of adding a barrier function to the composite objective function so as to impose constraints such that the process parameters satisfy inequality constraint conditions.

[0025] The method for calculating process parameters in the present invention further includes the step of adding a penalty function to the composite objective function so as to impose constraints such that the process parameters satisfy equality constraint conditions.

Advantages of the Invention

[0026] Generally speaking, the method for calculating process parameters in the present invention constructs an optimization model by regarding a trained prediction model as a function of process parameters to be determined. The objective function minimizes the difference value between the function value and the target result, and is subject to the constraints of the constraint conditions that should be satisfied by each of the process parameters to be determined or among them. And, in order to obtain the solution of the above constrained optimization problem, a method including a penalty function and a barrier function is used to transform the constrained optimization problem into an unconstrained optimization problem, making it easier to obtain the solution. Regarding this optimization problem, first, a plurality of samples that match the target result are found from the dataset when the trained prediction model is trained and used as anchors to form a multi-dimensional feasible solution space. And, after setting or creating an initial solution within this feasible solution space, the direction and step size are found by different numerical analysis methods to find a better feasible solution. Then, this better feasible solution becomes the available input parameters. In addition, the method for predicting and optimizing the effect of the process in the present invention can be applied not only to the field of cell processes, but also to any other trained machine learning model for inverse estimation of process parameters and unconfirmed input parameters.

Brief Description of the Drawings

[0027] [Figure 1] FIG. 1 shows a flowchart of the steps of the method for calculating process parameters based on a specific embodiment of the present invention. [Figure 2]Figure 2 is a flowchart of further steps in a method for calculating process parameters based on several specific embodiments of the present invention. [Figure 3] Figure 3 is a flowchart of further steps in a method for calculating process parameters based on several specific embodiments of the present invention. [Figure 4] Figure 4 shows a flowchart of the direction determination step in a method for calculating process parameters based on several specific embodiments of the present invention. [Figure 5] Figure 5 shows a flowchart of the direction determination step in a method for calculating process parameters based on several specific embodiments of the present invention. [Figure 6] Figure 6 shows a flowchart of the direction determination step in a method for calculating process parameters based on another specific embodiment of the present invention. [Figure 7] Figure 7 shows a flowchart of the direction determination step in a method for calculating process parameters based on another specific embodiment of the present invention. [Figure 8] Figure 8 shows a flowchart of the steps for calculating process parameters based on another specific embodiment of the present invention. [Figure 9] Figure 9 shows a flowchart of the steps for calculating process parameters based on another specific embodiment of the present invention. [Figure 10] Figure 10 shows a flowchart of the steps for calculating process parameters based on another specific embodiment of the present invention. [Modes for carrying out the invention]

[0028] To make the advantages, spirit, and features of the present invention easier and clearer to understand, specific examples will be described and examined in detail with reference to the drawings. It should be noted that these specific examples are merely representative examples of the present invention, and the specific methods, apparatus, conditions, materials, etc., illustrated are not limiting to the present invention or the corresponding specific examples. Furthermore, the components in the drawings are used only to represent their relative positions and are not described based on actual proportions. Also, the step numbers of the present invention are merely for distinguishing different steps and do not indicate the order of the steps. The above points are explained in advance.

[0029] Refer to Figure 1. Figure 1 shows a flowchart of the steps for a process parameter calculation method based on a specific embodiment of the present invention. The process parameter calculation method in this specific embodiment is used to check and optimize multiple process parameters in a process. The process parameters include multiple checked input parameters and multiple unchecked input parameters. Checked input parameters are input parameters that have been checked by the user during operation or have actually been executed. Unchecked input parameters are input parameters that have not yet been checked by the user or have not been executed. The process parameter calculation method includes the following:

[0030] Step S10: Provide a trained predictive model. The trained predictive model is obtained by machine learning the dataset using a machine learning method. The dataset contains multiple samples, and each sample contains multiple sample parameters. The trained predictive model is used to take multiple input parameters as input and generate prediction results corresponding to those input parameters.

[0031] Step S11: Set the target result corresponding to the prediction result of the trained predictive model, and provide the verified input parameters among the input parameters.

[0032] Step S12: Each sample in the dataset is input into the trained predictive model to check whether the generated prediction matches the target result. Then, samples whose prediction matches the target result are used as multiple anchor samples to form a multidimensional subspace, and the initial sample is identified within the multidimensional subspace.

[0033] Step S13: Obtain the orientation of the initial sample in a multidimensional subspace.

[0034] Step S14: For the sample parameters of the initial sample, the step size is increased along the aforementioned direction. Then, the confirmed input parameters and other sample parameters with increased step sizes that do not correspond to the confirmed input parameters are input to the trained predictive model to check whether the generated predictive results match the target results.

[0035] Step S15: If the prediction result matches the target result, the sample parameter with an increased step size that does not correspond to the confirmed input parameter is treated as an unconfirmed input parameter.

[0036] In this specific embodiment, the trained predictive model in step S10 may be any machine learning model obtained from an open platform that has completed training, or it may be a predictive model trained by the user themselves. Furthermore, the dataset in this specific embodiment may be a set formed from any data that can be used for training, testing, and validating machine learning. In actual applications, the method for calculating process parameters in this specific embodiment can also be applied to any other trained machine learning model and used for inverse estimation of unverified input parameters. Furthermore, the trained predictive model in this specific embodiment can be obtained by machine learning the dataset using an artificial neural network (ANN), convolutional neural network (CNN), recurrent neural network (RNN), or any other machine learning algorithm or neural network algorithm. The selection of a machine learning or neural network algorithm is made according to the user's needs.

[0037] In practical applications, cell culture processes often involve multiple input parameters, some of which may have already been identified or performed. The system and method of the present invention can be used to obtain desired target results by inversely estimating unperformed parameters and then performing them. For example, a user may be attempting to perform a 21-day cell culture process, where the cell preparation to be produced must have a 95% cell viability rate, and the user has already performed the 7-day process (i.e., already has confirmed input parameters). However, they may not have confirmed how to perform the process or adjust process parameters (i.e., unconfirmed input parameters) for days 8-21 to obtain the desired target results on the predetermined 21st day. Therefore, step S11 is a step to set inverse estimation conditions for the input parameters.

[0038] In step S12, several specific samples from the dataset used to train the trained predictive model coincide with the user-defined target result (for example, the cell preparation described above must have a 95% cell viability rate). Therefore, these samples have considerable reference value in the inverse estimation of the input parameters and can be used to form a multidimensional subspace as anchor samples. When points in this multidimensional subspace (representing a vector of input parameters) are input into the objective function of the trained predictive model, function values ​​that coincide with the target result are obtained. Thus, this multidimensional subspace can become the feasible solution space of the trained predictive model's function. However, the function values ​​of points in the feasible solution space only coincide with the target result and are not the optimal function values. Therefore, in step S12, it is necessary to further identify points in the feasible solution space that will serve as initial points for finding a feasible solution. That is, it is necessary to identify initial samples.

[0039] As described in steps S13 and S14, after identifying the initial sample, it is necessary to advance from this initial sample toward a feasible solution in at least one direction and at least one step size. After determining the direction in step S13, as described in step S14, advance along that direction by at least one step size, and input the portion of the sample parameters with increased step sizes that does not correspond to the verified input parameters into the trained predictive model to check whether the output function value matches the target result. It should be noted that the input parameters input to the trained predictive model described above are the verified input parameters and some of the parameters of points in the feasible solution space. In other words, they are the input parameters after they have been combined.

[0040] If the function value obtained by inputting the combined input parameters described above into a trained predictive model matches the target result, then the point in the feasible solution space becomes a feasible solution. Therefore, as described in step S15, some of the parameters of that point (i.e., other parameters that do not correspond to the confirmed input parameters) can be designated as unconfirmed input parameters. For example, if the first 7 days of a 21-day cell culture process have been identified (confirmed input parameters), then the unconfirmed input parameters identified in step S15 are the input parameters for days 8 through 21 of the point.

[0041] In the specific embodiment described above, an initial sample is identified in a multidimensional subspace in step S12. However, the method of the present invention further includes multiple methods in the step of identifying this initial sample. Refer to Figures 2 and 3. Figures 2 and 3 are flowcharts of further steps in a method for calculating process parameters based on multiple specific embodiments of the present invention. As shown in Figure 2, this specific embodiment differs from the above specific embodiment in that step S12 further includes step S120 and substep S122. Step S120 is the step of finding multiple anchor samples from the dataset in step S12 to form a multidimensional subspace. In step S122, the midpoint of the distance between at least two anchor samples in the multidimensional subspace is used as the initial sample. Furthermore, as shown in Figure 3, this specific embodiment differs in that step S12 further includes step S120 and step S124. In step S124, a linear combination between at least two anchor samples in the multidimensional subspace is used as the initial sample. The above specific embodiment is only two of multiple methods for identifying an initial point in a feasible solution space. In practical applications, the present invention can utilize any method capable of identifying the initial point or initial sample. Furthermore, the steps corresponding to the above-described specific embodiments are the same in the specific embodiments shown in Figures 2 and 3, and therefore will not be described in detail again here.

[0042] After identifying the initial point, it is necessary to determine the direction of travel from the initial point. Refer to Figures 4 and 5. Figures 4 and 5 show flowcharts of the direction determination steps in a method for calculating process parameters based on several specific embodiments of the present invention. As shown in Figure 4, this specific embodiment differs from the above-described specific embodiment in the following respects. Specifically, the method for calculating process parameters further includes the following:

[0043] Step S132: Using numerical analysis, the second directional derivative of the initial sample along each anchor sample is calculated as the direction of the initial sample.

[0044] As shown in Figure 5, this specific embodiment differs from the specific embodiment described above in the following respects. Specifically, the method for calculating process parameters further includes the following:

[0045] Step S134: The initial sample is set as the origin, and the third directional derivative of the origin toward each anchor sample is calculated using numerical analysis as the direction of the initial sample.

[0046] Alternatively, in the specific embodiment shown in Figure 4, the tangent direction of the initial sample toward the anchor sample may be selected as the direction of propagation. In other words, the second directional derivative may be calculated using numerical analysis. It should be noted that since there are multiple anchor samples, the initial sample also has multiple second directional derivatives for these anchor samples. In practice, one second directional derivative may be selected from the initial sample, propagated by one step increment, and after combining some of the input parameters of the point after propagation with the confirmed input parameters, it may be input into the trained prediction model to check whether the output function value is better (i.e., whether the output function value is closer to the target result). If the function value is better, this second directional derivative is set as the direction of the initial sample. On the other hand, if the function value is not better, the next second directional derivative may be selected, and the same steps may be repeated until a second directional derivative that can produce a better function value is found. Furthermore, in the specific embodiment shown in Figure 5, the initial sample may first be set as the origin in the multidimensional subspace, and then the third directional derivative of the origin toward each anchor sample may be calculated using the numerical analysis method described above. Then, similar to the specific embodiment in Figure 4, a third directional derivative that may yield a better function value is selected as the direction of the initial sample. In the specific embodiment in Figure 5, after advancing one step increment from the initial sample, a new point reached in the feasible solution space may be used as a new starting point. That is, the current solution may be set back as the origin, and step S134 in Figure 5 may be repeated to advance another step increment.

[0047] In the specific embodiment shown in Figure 4, the tangential direction is defined as the direction of propagation, but as shown in the specific embodiment shown in Figure 6, the secant direction from the initial sample to the anchor sample may also be defined as the direction of propagation. Figure 6 shows a flowchart of the direction determination step in the method for calculating process parameters based on another specific embodiment of the present invention. As shown in Figure 6, this specific embodiment differs from the specific embodiment described above in the following respects. That is, the trained prediction model in this specific embodiment can be expressed as an objective function. Furthermore, it includes the following:

[0048] Step S136: Using numerical analysis, the secant derivative of the initial sample toward each anchor sample is obtained by calculating the change in the function value of the objective function based on the change in the distance of the initial sample toward each anchor sample.

[0049] In the direction determination step described above, if the initial samples proceed in the opposite direction to the feasible solution, the function value may deteriorate or deviate from expectations. Therefore, according to the method of the present invention, it is possible to sequentially search for directions in which the function value may be good. Now, refer to Figure 7. Figure 7 shows a flowchart of the direction determination step in a method for calculating process parameters based on another specific embodiment of the present invention. As shown in Figure 7, this specific embodiment differs from the specific embodiment described above in the following respects. That is, the trained predictive model in this specific embodiment can be expressed as an objective function. Furthermore, it includes the following steps.

[0050] Step S1380: The step size of the initial sample is sequentially increased and adjusted toward the directional derivative of each anchor sample.

[0051] Step S1382: After increasing and adjusting the step size of the initial samples toward the directional derivative of each anchor sample, it is sequentially checked whether the function value of the objective function is better (i.e., whether the function value of the objective function is closer to the target result).

[0052] Step S1384: After increasing the step size of the initial sample toward the directional derivative of any of the anchor samples, if it is confirmed that the function value of the objective function is better, the step size of the initial sample is increased.

[0053] Step S1386: After increasing the step size of the initial sample toward the directional derivative of any of the anchor samples, if it is confirmed that the function value of the objective function does not improve, the step size of the initial sample is increased toward the directional derivative of another anchor sample, and it is checked again whether the function value of the objective function improves.

[0054] Therefore, in this specific embodiment, first, the step size is repeatedly adjusted from the initial sample toward the first anchor sample, and it is checked whether the function value improves after increasing the step size. If an appropriate step size is found, the process can proceed along that direction with that step size. However, if an appropriate step size that improves the function value cannot be found even after adjusting the step size a certain number of times, the above adjustment is repeated for the initial sample toward the next anchor sample. As for the above step size adjustment, it may be better to first set an initial size for the step size, and then check whether the prediction result output after inputting the initial sample to the trained prediction model approaches the target result after the initial sample has progressed to this initial size. If it approaches the target result further, the process can proceed with this initial step size. On the other hand, if it does not approach the target result, the step size is repeatedly adjusted until the prediction result output from the trained prediction model approaches the target result, and the initial sample can proceed with a step size that approaches the target result. In practice, if a suitable step size that yields better function values ​​cannot be found for all anchor sample directions (for example, after adjusting the step size 10 times), the process parameter calculation is stopped. The above steps constitute a cyclical search. If the initial point has progressed to a point in the next feasible solution space, the new point may be set as the current solution, and the above steps may be repeated.

[0055] When applying the process parameter calculation method in this specific embodiment to the field of cell culture processes, the dataset in this specific embodiment may further include a cell dataset. The cell dataset includes multiple cell samples. Furthermore, the sample parameters for each cell sample include source parameters and culture parameters. In practice, the cell dataset may be data obtained from an open platform or data collected by the user themselves. In addition, the types of cell samples may include immune cells (e.g., dendritic cells (DC cells), cytokine-induced killer cells (CIK), tumor-infiltrating lymphocytes (TILs), natural killer cells (NK cells), and CAR-T cells), stem cells (e.g., peripheral blood stem cells, adipose-derived stem cells, bone marrow mesenchymal stem cells), chondrocytes, fibroblasts, etc., but are not limited to these in actual applications and should be determined according to the type of cell culture the user wants to perform. Moreover, in this specific embodiment, the source parameters for each cell sample may further include attribute data of the source of each cell sample. Each cell sample in the cell dataset may include the cell source corresponding to the cell and attribute data of the source. Attribute data may include physiological data of the source, or other source-related data such as the source's sex, age, medical history, living environment, and residential area. However, in practical applications, the source parameters of cell samples may also include other parameters that may affect cellular processes and are related to the cell source.

[0056] Furthermore, in this specific embodiment, the culture parameters for each cell sample also include human, equipment, material, method, and environmental parameters for each cell sample. The cell culture process involves many steps, and each step is related to many culture parameters. These culture parameters include, for example, human-related parameters such as the sex and age of the cell source and the experience and stability of the cell culture operator; equipment-related parameters such as the type and grade of the cell operation platform and the stability and precision of the temperature and humidity control of the cell culture apparatus; material-related parameters such as the material of the cell culture dish and the components, ratios, and formulations of the cell culture medium; method-related parameters such as the cell culture operator's technique and the cell culture process; and environmental parameters such as the ambient temperature, humidity, and carbon dioxide concentration of the cell culture environment.

[0057] Within the process, each process parameter is not arbitrarily set, but may have various constraints depending on the actual situation. The trained predictive model includes a function of specific pending process parameters, an objective function, and at least one constraint. The objective function minimizes the difference between the output value of the function of the specific pending process parameters and the target result, and a constrained optimization problem is modeled based on the function of the specific pending process parameters, the objective function, the at least one constraint, and the predictive result. The method for calculating the process parameters obtains the unconfirmed input parameters by finding a solution to the optimization problem. Now, refer to Figure 8. Figure 8 shows a flowchart of the steps of the method for calculating process parameters according to another specific embodiment of the present invention. As shown in Figure 8, this specific embodiment differs from the specific embodiment described above in the following respects. That is, the method for calculating process parameters in this specific embodiment further includes the following.

[0058] Step S100: Provide a trained predictive model. The trained predictive model is obtained by machine learning the dataset using a machine learning method. The trained predictive model includes a function of specific waiting process parameters, an objective function, and at least one constraint. The objective function includes a barrier function. The dataset includes multiple samples, and each sample includes multiple sample parameters. The trained predictive model is used to take multiple input parameters as input and generate prediction results corresponding to the input parameters.

[0059] In this specific embodiment, the barrier function is set to constrain the process parameters so that they satisfy the inequality constraint. The barrier function includes a barrier region and a barrier function value. When the process parameters satisfy the inequality constraint, they are outside the barrier region, and the corresponding barrier function value is 0. That is, it does not affect the prediction results of the trained prediction model. On the other hand, when the process parameters do not satisfy the inequality constraint, they are inside the barrier region, and the corresponding barrier function value becomes large, which has a significant impact on the prediction results of the trained prediction model. It should be noted that the other steps in this specific embodiment are the same as the corresponding steps in the specific embodiment described above, and therefore will not be described in detail again here.

[0060] See also Figure 9. Figure 9 shows a flowchart of the steps for calculating process parameters based on another specific embodiment of the present invention. As shown in Figure 9, this specific embodiment differs from the specific embodiment described above in the following respects. That is, the method for calculating process parameters in this specific embodiment further includes the following.

[0061] Step S102: Provide a trained predictive model. The trained predictive model is obtained by machine learning the dataset using a machine learning method. The trained predictive model has an objective function and at least one constraint, the objective function including a penalty function. The dataset contains multiple samples, and each sample contains multiple sample parameters. The trained predictive model is used to take multiple input parameters as input and generate prediction results corresponding to the input parameters.

[0062] In this specific embodiment, the penalty function is set to constrain the process parameters so that they satisfy the constraints of the equation. The penalty function includes a penalty region and a penalty function value. When the process parameters satisfy the constraints of the equation, they are within the penalty region, and the corresponding penalty function value is 0. On the other hand, when the process parameters do not satisfy the constraints of the equation, they are outside the penalty region, and the corresponding penalty function value becomes large, significantly affecting the prediction results of the trained prediction model. Therefore, by inputting sample parameters into a trained prediction model with a barrier function and / or penalty function, it is possible to not only check whether the generated prediction results match the target results, but also to determine whether the input parameters satisfy the actual constraints. If the input process parameters do not satisfy the conditions of the barrier function and / or penalty function, the prediction results will deviate significantly, indicating that the process parameters cannot be used because the conditions are not met. It should be noted that the other steps in this specific embodiment are the same as the corresponding steps in the specific embodiment described above, and therefore will not be described in detail again here.

[0063] The actual constraints mentioned above include, for example, that in the cell process, the combined antibiotic concentration of streptomycin and amphotericin B added to the cell culture medium on day 10 must be 200 μg / mL or less, and the sum of the serum concentration values ​​in the cell culture medium on days 8 and 10 must be 20% or less.

[0064] In practical applications, penalty functions and barrier functions are commonly used in machine learning to restrict the parameters of a machine learning model to a reasonable range to match the actual process conditions in natural environmental conditions. For example, they are used to restrict conditions such as cell culture temperature and relative humidity not being negative, and the concentration of components in the culture medium not harming the cells. By setting barrier functions and penalty functions to restrict the range of parameters, the relationship between parameters is comprehensively considered to find the combination of process parameters that comes closest to the desired target result. In addition, in practical applications, barrier functions may be applied to restrict unverified input parameters, and penalty functions may be applied to restrict verified input parameters. These can be used interchangeably depending on the user's needs. In this specific embodiment, the objective function minimizes the difference between the function value and the target result, and is constrained by the constraints that must be satisfied by each or between specific pending process parameters. Furthermore, to find the solution to the above constrained optimization problem, the method including penalty functions and barrier functions is used to convert the constrained optimization problem into an unconstrained optimization problem, making it easier to find the solution.

[0065] Furthermore, in practical applications, multiple different pre-trained predictive models can be used to predict different results for different input parameters, such as cell viability, number of proliferating cells, and culture time. Refer to Figure 10. Figure 10 shows a flowchart of the steps for a process parameter calculation method based on another specific embodiment of the present invention. The process parameter calculation method in this specific embodiment is used to verify and optimize multiple process parameters in a process. The process parameters include multiple verified input parameters and multiple unverified input parameters. The process parameter calculation method includes the following:

[0066] Step S10': Provide multiple pre-trained predictive models. Each pre-trained predictive model is obtained by machine learning a dataset using a machine learning method. The dataset contains multiple samples, and each sample contains multiple sample parameters. Each pre-trained model corresponds to a different objective function. The pre-trained predictive models are used to take multiple input parameters as input and generate multiple predictive results corresponding to those input parameters. Each predictive result belongs to a different type of objective result.

[0067] Step S11': Combine the objective functions of the trained prediction models to form a composite objective function. Inputting the input parameters into the composite objective function generates the prediction results.

[0068] Step S12': Set multiple target results corresponding to the prediction results of the trained predictive model, and provide verified input parameters among the input parameters.

[0069] Step S13': Each sample in the dataset is input into the trained predictive model to check whether the generated prediction results match the target results. Then, samples whose prediction results match the target results are used as multiple anchor samples to form a multidimensional subspace, and the initial samples are identified within the multidimensional subspace.

[0070] Step S14': Obtain the orientation of the initial sample in a multidimensional subspace.

[0071] Step S15': For the sample parameters of the initial sample, the step size is increased along the direction. Then, the verified input parameters and the sample parameters with increased step sizes that do not correspond to the verified input parameters are input into the composite objective function to check whether the generated prediction results match the target results.

[0072] Step S16': If the prediction result matches the target result, the sample parameter with an increased step size that does not correspond to the confirmed input parameter is treated as an unconfirmed input parameter.

[0073] According to the method for calculating process parameters in this specific embodiment, each trained predictive model can inversely estimate one group of input parameters, making it possible to simultaneously satisfy different desired results (for example, the cell viability of the cell product after culture is higher than 90%, and the number of cells is 1 x 10⁶). 7 (More than 100 cells and a culture time of 7 days are simultaneously satisfied). In this specific embodiment, the objective functions of each of the above-mentioned trained prediction models can be combined in any way, for example, by directly adding them together or by forming a new function. Then, by performing inverse estimation on this new function using the various steps described in this invention, a set of input parameters that can simultaneously satisfy the above conditions can be obtained. In another specific embodiment, similarly, these trained prediction models may also have barrier functions and / or penalty functions added to their combined objective functions using the above method. By adding barrier functions or penalty functions to the objective function, relatively complex constraints are transferred to the objective function, making it easier to find a solution.

[0074] In the specific embodiment described above, the objective functions of each trained prediction model are directly added together to form a new function. That is, the weight values ​​of the objective functions of each trained prediction model are all set to 1. However, in actual applications, the weight values ​​of the objective functions of each trained prediction model may be adjusted according to the priority and importance order of the desired results for each process. Furthermore, the barrier function and penalty function included in each objective function of each trained prediction model may also have their weight values ​​set and adjusted according to the priority of the desired results for each process. In addition, the construction of an optimization problem for multiple trained models can be achieved in the following two ways.

[0075] Method 1: If each trained predictive model holds a constrained optimization problem, first, the objective function of each trained predictive model is normalized (for example, converted into the form of a minimization objective function), and weight values ​​are assigned to each to form a composite objective function. Next, the constraints are considered and integrated for the composite objective function (i.e., the combination of constraints included in each trained predictive model is obtained), and the constrained optimization problem is converted into an unconstrained optimization problem. For example, an equality constraint is processed with a penalty function and an inequality constraint is processed with a barrier function to form a new function. Note that the constraints included in the objective function of each trained predictive model have different weight values ​​according to the priority of the desired results of the process. By integrating each constraint and converting it into processing by a barrier function and a penalty function, and considering the objective function of each trained predictive model, the weight values ​​of the barrier function and penalty function are derived.

[0076] Second method: When the optimization problem of each trained predictive model is transformed into an unconstrained optimization problem (i.e., when all constraints included in each trained predictive model are processed by barrier functions and / or penalty functions and added to the objective function), there is no need to adjust the constraints. Instead, the objective function of each trained predictive model is normalized, and the corresponding weight values ​​are adjusted to form a composite objective function. Then, the barrier functions and / or penalty functions included in the objective function of each trained predictive model are added to form a new function. Note that the objective function of each trained predictive model may each include its own barrier function and penalty function. Furthermore, each barrier function and penalty function can be assigned corresponding weight values ​​according to the priority of the objective function. It should be noted that the other steps in this specific embodiment are the same as the corresponding steps in the specific embodiment described above, and therefore will not be described in detail again here.

[0077] In summary, the method for calculating process parameters in the present invention constructs an optimization model by considering a trained predictive model as a function of specific pending process parameters. The objective function minimizes the difference between the function value and the target result, and is constrained by the constraints that must be satisfied by each or any of the specific pending process parameters. Furthermore, to find the solution to the above-mentioned constrained optimization problem, the constrained optimization problem is transformed into an unconstrained optimization problem using a method that includes a penalty function and a barrier function, making it easier to find the solution. For this optimization problem, first, multiple samples that match the target result are found as anchors from the dataset used to train the trained predictive model, forming a multidimensional feasible solution space. Furthermore, after setting or creating an initial solution within this feasible solution space, a better feasible solution is found by finding the direction and step size using different numerical analysis methods. This better feasible solution then becomes the usable input parameter. In addition, the method for predicting and optimizing process effects in the present invention can be applied to any other trained machine learning model, in addition to its application to the field of cellular processes, and can be used for inverse estimation of process parameters and unconfirmed input parameters.

[0078] The detailed description of the above preferred specific embodiments is intended to more clearly describe the features and spirit of the invention and does not limit the scope of the invention by the preferred specific embodiments disclosed above. Rather, it is intended that various modifications and equivalent configurations are covered within the claims of the invention. Therefore, the claims of the invention should be interpreted most broadly based on the above description to cover all possible modifications and equivalent configurations. [Explanation of Symbols]

[0079] Steps S10-S15, S120-S124, S132-S136 Steps S10'~S16', S1380~S1386

Claims

1. A method for calculating process parameters used to verify and optimize multiple process parameters in a process, wherein the process parameters include multiple verified input parameters and multiple unverified input parameters, and the method for calculating the process parameters is: A pre-trained predictive model is provided, the pre-trained predictive model is obtained by machine learning a dataset using a machine learning method, the dataset includes multiple samples, and each sample includes multiple sample parameters, and the pre-trained predictive model is used to take multiple input parameters as input and generate a prediction result corresponding to the input parameters. The steps include setting a target result corresponding to the prediction result of the trained prediction model and providing the verified input parameter among the input parameters, The steps include: checking whether the prediction result generated by inputting each sample in the dataset into the trained prediction model matches the target result; forming a multidimensional subspace with the samples whose prediction result matches the target result as a plurality of anchor samples; and identifying the initial sample within the multidimensional subspace. A step of obtaining the orientation of the initial sample in the multidimensional subspace, The steps include increasing the step size along the direction of the sample parameters of the initial sample, inputting the verified input parameters and the sample parameters with increased step sizes that do not correspond to the verified input parameters into the trained prediction model, and checking whether the generated prediction result matches the target result. If the prediction result matches the target result, the step of setting the sample parameter with an increased step size that does not correspond to the confirmed input parameter as the unconfirmed input parameter. A method that includes this.

2. The step of identifying the initial sample in the multidimensional subspace is further, A method for calculating process parameters according to claim 1, comprising the step of setting the initial sample to the midpoint of the distance between at least two of the anchor samples in the multidimensional subspace.

3. The step of identifying the initial sample in the multidimensional subspace is further, A method for calculating process parameters according to claim 1, comprising the step of using a linear combination of at least two of the anchor samples in the multidimensional subspace as the initial sample.

4. The step of obtaining the orientation of the initial sample in the multidimensional subspace is further, A method for calculating process parameters according to claim 1, comprising the step of calculating, by numerical analysis, the direction derivative representing the direction of the tangent line from the initial sample to each of the anchor samples as the direction of the initial sample.

5. The step of obtaining the orientation of the initial sample in the multidimensional subspace is further, A method for calculating process parameters according to claim 1, comprising the step of setting the initial sample as the origin and calculating, by numerical analysis, the direction derivative representing the direction from the origin toward each of the anchor samples as the direction of the initial sample.

6. The trained predictive model is represented by an objective function, and the step of obtaining the direction of the initial sample in the multidimensional subspace is further, A method for calculating process parameters according to claim 1, comprising the step of obtaining a derivative representing the secant direction from the initial sample to each of the anchor samples by calculating the change in the function value of the objective function based on the change in the distance of the initial sample toward each of the anchor samples using a numerical analysis method.

7. The trained predictive model is represented by an objective function, and the step of obtaining the direction of the initial sample in the multidimensional subspace is further, The steps include sequentially increasing and adjusting the step size of the initial sample toward the directional derivative of each anchor sample, The steps include: increasing and adjusting the step size of the initial sample toward the directional derivative of each anchor sample, and then sequentially checking whether the function value of the objective function approaches the target result; After increasing the step size of the initial sample toward the directional derivative of any of the anchor samples, if it is confirmed that the function value of the objective function approaches the target result, the step of increasing the step size of the initial sample toward the directional derivative, If, after increasing the step size of the initial sample toward the directional derivative of any of the anchor samples, it is confirmed that the function value of the objective function does not approach the target result any further, then the step size of the initial sample is increased toward the directional derivative of another anchor sample, and then it is confirmed again whether the function value of the objective function approaches the target result any further. A method for calculating process parameters according to claim 1, including the following:

8. The step of increasing the step width along the direction of the sample parameters of the initial sample is further, The steps include setting an initial size for the step size, increasing the step size of the sample parameters of the initial sample by the initial size along the direction, inputting this into the trained prediction model, and checking whether the prediction result output from the trained prediction model approaches the target result. If the prediction result output from the trained prediction model approaches the target result, the steps include determining the initial size and the step size, If the prediction result output from the trained prediction model does not approach the target result, the step continues to adjust the step size until the prediction result approaches the target result, and increases the step size of the initial sample parameter in the direction by the specified size, then inputs it into the trained prediction model to determine the final step size. A method for calculating process parameters according to claim 1, including the following:

9. The method for calculating process parameters according to claim 1, wherein the trained predictive model includes a function of specific waiting process parameters, an objective function, and at least one constraint, the objective function minimizing the difference between the output value of the function of specific waiting process parameters and the target result, a constrained optimization problem is modeled based on the function of specific waiting process parameters, the objective function, the at least one constraint, and the prediction result, and the method for calculating process parameters obtains the unconfirmed input parameters by finding a solution to the optimization problem.

10. Furthermore, the method for calculating process parameters according to claim 9, further comprising the step of adding a barrier function to the objective function in order to restrict the process parameters so that they satisfy the constraints of an inequality.

11. Furthermore, the method for calculating process parameters according to claim 9, further comprising the step of adding a penalty function to the objective function in order to restrict the process parameters to satisfy the constraints of the equation.

12. The method for calculating process parameters according to claim 1, wherein the dataset further comprises a process dataset, the process dataset comprises a plurality of process samples, and the sample parameters of each process sample include source parameters and culture parameters.

13. The method for calculating process parameters according to claim 12, wherein the source parameters of each process sample further include attribute data of the source of each process sample, and the culture parameters of each process sample further include human, equipment, material, method, and environmental parameters in each process sample.

14. The method for calculating process parameters according to claim 1, wherein the trained predictive model is obtained by machine learning the dataset using any of the following: artificial neural networks (ANN), convolutional neural networks (CNNs), and recurrent neural networks (RNNs).

15. Furthermore, If the prediction result does not match the target result, the steps include: comparing the confirmed input parameters and the sample parameters of the sample again to select a second sample from the sample; The steps include: calculating the gradient of the second sample in the objective function of the trained prediction model using numerical analysis; The steps include increasing the step size of the sample parameters of the second sample along the opposite direction of the gradient, inputting all the sample parameters with increased step sizes into the trained prediction model, and checking whether the generated prediction result matches the target result. If the prediction result matches the target result, the step of setting the sample parameter with an increased step size that does not correspond to the confirmed input parameter as the unconfirmed input parameter. A method for calculating process parameters according to claim 1, including the following:

16. A method for calculating process parameters used to verify and optimize multiple process parameters in a process, wherein the process parameters include multiple verified input parameters and multiple unverified input parameters, and the method for calculating the process parameters is: The present invention provides multiple pre-trained predictive models, each of which is obtained by machine learning a dataset using a machine learning method, the dataset containing multiple samples, each sample containing multiple sample parameters, each pre-trained predictive model corresponding to a plurality of objective functions, and the pre-trained predictive models are used to take multiple input parameters as input and generate multiple predictive results corresponding to the input parameters, each predictive result belonging to a different type of objective result. The steps include: combining the objective functions of the trained prediction models to form a composite objective function, and inputting the input parameters into the composite objective function to generate the prediction result; The steps include setting a plurality of target results corresponding to the prediction results of the trained prediction model, and providing the verified input parameters among the input parameters, The steps include: comparing whether the prediction results generated by inputting each sample in the dataset into the trained prediction model match the target results, forming a multidimensional subspace with the samples whose prediction results match the target results as a plurality of anchor samples, and identifying initial samples within the multidimensional subspace; A step of obtaining the orientation of the initial sample in the multidimensional subspace, The steps include increasing the step size along the direction for the sample parameters of the initial sample, inputting the verified input parameters and the sample parameters with increased step sizes that do not correspond to the verified input parameters into the composite objective function, and checking whether the generated prediction result matches the target result. If the prediction result matches the target result, the step of setting the sample parameter with an increased step size that does not correspond to the confirmed input parameter as the unconfirmed input parameter. A method that includes this.

17. Furthermore, the method for calculating process parameters according to claim 16, comprising the step of adding a barrier function to the composite objective function in order to restrict the process parameters to satisfy the constraints of the inequality.

18. Furthermore, the method for calculating process parameters according to claim 16, further comprising the step of adding a penalty function to the composite objective function in order to restrict the process parameters to satisfy the constraint conditions of the equation.

Images