How to calculate the decision variable
By adding a dummy layer with artificial neurons and using an optimizer to adjust weights, the method addresses the challenge of inverse estimation in cell culture processes and other neural networks, optimizing input parameters efficiently.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- METATECH (AP) INC
- Filing Date
- 2024-07-08
- Publication Date
- 2026-06-09
AI Technical Summary
Conventional machine learning predictive models cannot inversely estimate the determinants of cell culture processes, requiring extensive trial and error to achieve desired results, and this is a common issue in other neural network training contexts as well.
A method involving a dummy layer with artificial neurons connected to a pre-trained predictive model, where the bias value of the activation function is set to 0, and an optimizer adjusts the weight values to achieve the desired output, allowing for inverse estimation of decision variables.
Enables efficient calculation of optimal input parameters by fixing model parameters and adjusting weights, reducing the need for extensive trial and error, applicable to cell culture processes and other neural network models.
Smart Images
Figure 0007872058000001 
Figure 0007872058000002 
Figure 0007872058000003
Abstract
Description
Technical Field
[0001] The present invention relates to a method for calculating decision variables, and particularly to a method for calculating decision variables for inversely estimating decision variables and process parameters by using an optimizer built in a training platform of an artificial intelligence neural network engine.
Background Art
[0002] With the progress of medical engineering technology, not only has it had a great impact on the healthcare field, but also the application of regenerative medicine in clinical practice has been increasing. Regenerative medicine is mainly a medical technology that repairs damaged tissues and organs by utilizing the regenerative ability of cells, and its application range 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 tissue engineering and regenerative therapy, organ transplantation and regeneration, tissue regeneration and repair, cancer treatment, nerve regeneration, immune cell therapy, and stem cell therapy. Among them, the research and application of cell therapy in regenerative medicine have been attracting increasing attention from all sectors. Cell therapy is a process in which human cells are cultured or processed outside the body and then transplanted into individual bodies for use.
[0003] Cell therapy offers advantages such as personalized treatment, avoidance of rejection, alternative to organ transplantation, and restoration of tissue function. Currently, as many governments gradually lift restrictions on the application of cell therapy within their respective countries, numerous scholars both domestically and internationally are entering 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. According to previous research, each cell in the human body has its own unique characteristics, and therefore the type of cell preparation applied differs depending on the disease and symptoms. Thus, it is necessary to customize cells suitable for treatment according to individual circumstances, which increases the complexity and difficulty of the cell process. Furthermore, the quality of the cell therapy product directly affects the safety and effectiveness of the treatment. Therefore, in the cell culture process, it is necessary to strictly control the growth state of the 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. However, because cellular variability is extremely high across different cases, the optimal culture and environmental parameters for cell therapies applied to different cases will not be exactly the same. Furthermore, since it is necessary to adjust each determinant variable (i.e., process parameter) within the process for each cell therapy to achieve the desired results, it is not possible to produce each cell therapy using fixed determinants.
[0004] Conventional techniques involve training a predictive model on a large dataset of samples using machine learning. The parameters of the predictive model are then adjusted using a machine learning optimizer to minimize or maximize the objective function. This improves the accuracy of the predictive model, accelerates its convergence process, and further reduces the time and computational cost required for machine learning and convergence. In addition, since this predictive model can generate cell culture prediction results by inputting various decision variables, users can simulate the effects of their designed cell culture process in advance.
[0005] However, the predictive models trained using machine learning cannot inversely estimate the determinants of the cell culture process from the user's desired results. In other words, when designing the process, users need to try a large number of different determinants, thus expending a significant amount of effort and resources on designing and improving the culture process. In addition to the need to predict parameters in the cell culture process, inverse estimation of determinants and process parameters is also required for other machine learning models used to train neural networks.
[0006] Therefore, it is necessary to develop a method to further optimize the process by enabling the inverse estimation of the process's decision variables that match the prediction results. [Overview of the Initiative] [Problems that the invention aims to solve]
[0007] In view of the above, the present invention provides a method for calculating a decision variable in order to solve the conventional problems described above. [Means for solving the problem]
[0008] The present invention provides a method for calculating a decision variable. The method provides a trained predictive model, which is obtained by machine learning using a machine learning method, which includes an input layer and an output layer, which is used to input a plurality of input parameters through the input layer and generate prediction results corresponding to the input parameters by the output layer, which sets a target result corresponding to the prediction results of the trained predictive model and provides at least one confirmed input parameter among the input parameters, which adds a dummy layer connected to the input layer of the trained predictive model, which includes a plurality of artificial neurons connected to each input terminal of the input layer, which forms a parameter prediction model with the trained predictive model and the dummy layer, which is used to make the prediction results of the trained predictive model the output of the parameter prediction model, which sets the bias value of the activation function of each artificial neuron to 0. The method includes the steps of setting the activation function so that when the input value of the activation function is 1, the output value of the activation function is 1; setting at least one first weight value for at least one first artificial neuron among the artificial neurons corresponding to at least one verified input parameter based on the at least one verified input parameter; setting the output of the parameter prediction model corresponding to the target result; training the parameter prediction model by inputting a training dataset containing at least one all-one vector into the artificial neurons of the dummy layer of the parameter prediction model; and adjusting the second weight values of the artificial neurons other than the at least one first artificial neuron based on the target result and the at least one first weight value, using the optimizer in the machine learning method that generated the trained prediction model, and setting the second weight values to a plurality of unverified input parameters among the input parameters.
[0009] The optimizer further supports Adaptive Moment Estimation (Adam) optimization, Stochastic Gradient Descent (SGD), Momentum, Nesterov Accelerated Gradient (NAG), Adaptive Gradient Algorithm (AdaGrad), Nadam (Nesterov-accelerated Adaptive Moment Estimation), RMSprop (Root Mean Square Propagation), Adadelta (Adaptive Delta), AdamW (Adam with Weight Decay), AMSGrad (Adaptive Moment Estimation with Long-term Memory), AdaBelief (Adaptive Belief), LARS (Layer-wise Adaptive Rate Scaling), Self-adaptive Hessian (AdaHessian), and RAdam (Rectified The following can be selected: Adam), Lookahead, MadGrad (Momentumized, Adaptive, and Decentralized Gradient Descent), Yogi optimizer (Yogi), and AdamMax (Adaptive Moment Estimation with Maximum).
[0010] A trained predictive model is obtained by machine learning the dataset using the aforementioned machine learning method.
[0011] The aforementioned trained predictive model includes multiple model parameters, all of which are fixed.
[0012] The trained predictive model is obtained by machine learning the aforementioned dataset using one of the following: Artificial Neural Networks (ANN), Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Recursive Neural Networks (RecNN), or Complex Neural Networks.
[0013] The present invention provides a method for calculating decision variables used in other contexts to calculate multiple decision variables. The method provides a pre-trained predictive model, which is obtained by machine learning using a machine learning method, and includes an input layer and an output layer, the pre-trained predictive model is used to input a plurality of input parameters through the input layer and generate prediction results corresponding to the input parameters by the output layer, setting a target result corresponding to the prediction result of the pre-trained predictive model, adding a dummy layer connected to the input layer of the pre-trained predictive model, the dummy layer includes a plurality of artificial neurons connected to each input terminal of the input layer, a parameter prediction model is formed by the pre-trained predictive model and the dummy layer, the prediction result of the pre-trained predictive model becomes the output of the parameter prediction model, setting the bias value of the activation function of each artificial neuron to 0, so that when the input value of the activation function is 1, the output value of the activation function becomes 1, setting the output of the parameter prediction model corresponding to the target result, training the parameter prediction model by inputting a training dataset containing at least one all-one vector to the artificial neurons of the dummy layer of the parameter prediction model, and adjusting the reference weight values of each artificial neuron based on the target result using an optimizer in the machine learning method that generated the pre-trained predictive model, and using the reference weight values as decision variables.
[0014] The optimizer further supports Adaptive Moment Estimation (Adam) optimization, Stochastic Gradient Descent (SGD), Momentum, Nesterov Accelerated Gradient (NAG), Adaptive Gradient Algorithm (AdaGrad), Nadam (Nesterov-accelerated Adaptive Moment Estimation), RMSprop (Root Mean Square Propagation), Adadelta (Adaptive Delta), AdamW (Adam with Weight Decay), AMSGrad (Adaptive Moment Estimation with Long-term Memory), AdaBelief (Adaptive Belief), LARS (Layer-wise Adaptive Rate Scaling), Self-adaptive Hessian (AdaHessian), and RAdam (Rectified The following can be selected: Adam), Lookahead, MadGrad (Momentumized, Adaptive, and Decentralized Gradient Descent), Yogi optimizer (Yogi), and AdamMax (Adaptive Moment Estimation with Maximum).
[0015] A trained predictive model is obtained by machine learning the dataset using the aforementioned machine learning method.
[0016] The aforementioned trained predictive model includes multiple model parameters, all of which are fixed.
[0017] A trained predictive model can be obtained by machine learning a dataset using one of the following: Artificial Neural Networks (ANNs), Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Recursive Neural Networks (RecNNs), or Complex Neural Networks. [Effects of the Invention]
[0018] In summary, the present invention provides a method for calculating decision variables. This method involves adding dummy layers to the input terminals of a pre-trained neural network prediction model. The dummy layers contain the same number of artificial neurons as the input terminals of the pre-trained prediction model. A new link is established between each artificial neuron and the corresponding neuron at the input terminal of the pre-trained prediction model. The input value of each artificial neuron is set to 1. If the bias value of the activation function is 0 and the input to the activation function is 1, the output is 1. Initial values for the weights of the new links are then selected and set, and these weights are considered the decision variables. The weights may have a range or other mutual constraints. The parameters of the pre-trained prediction model are frozen, and only the weights of the new links are adjusted, while an optimizer built into a common machine learning platform finds the optimal solution. The training goal is set so that the output of this parameterized prediction model becomes the desired target result. The weights of the new links obtained when training is complete become the executable input decision variables. Furthermore, the method for calculating decision variables in this invention may be applied not only to the field of cellular processes but also to other machine learning models trained by neural networks, and used for inverse estimation of decision variables and process parameters. This invention makes it possible to effectively search for optimal input parameters to achieve the desired results by utilizing a general machine learning platform and the methods built into it. [Brief explanation of the drawing]
[0019] [Figure 1A] Figure 1A shows a functional block diagram of a method for calculating decision variables based on a specific embodiment of the present invention. [Figure 1B] Figure 1B shows a functional block diagram of a method for calculating decision variables based on a specific embodiment of the present invention. [Figure 2] Figure 2 shows a flowchart of the steps for calculating the decision variable based on a specific embodiment of the present invention. [Figure 3] Figure 3 shows a flowchart of the steps for calculating the decision variable based on another specific embodiment of the present invention. [Modes for carrying out the invention]
[0020] 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.
[0021] Refer to Figures 1A, 1B, and 2. Figures 1A and 1B show functional block diagrams of the method for calculating the decision variable based on a specific embodiment of the present invention. Figure 2 shows a flowchart of the steps of the method for calculating the decision variable based on a specific embodiment of the present invention. As shown in Figure 2, in this specific embodiment, the method for calculating the decision variable includes the following.
[0022] Step S1: Provide the trained prediction model 10. The trained prediction model 10 is obtained by machine learning using a machine learning method. The trained prediction model 10 includes an input layer 11 and an output layer 12. The trained prediction model 10 is used to input a plurality of input parameters through the input layer 11 and generate a prediction result (i.e., an output result) corresponding to the input parameters by the output layer 12.
[0023] Step S2: Set a target result corresponding to the prediction result of the trained prediction model 10 and provide at least one confirmed input parameter among the input parameters.
[0024] Step S3: Add a dummy layer 20 connected to the input layer 11 of the trained prediction model 10. The dummy layer 20 includes a plurality of artificial neurons 21 respectively connected to each input terminal 111 of the input layer 11. The trained prediction model 10 and the dummy layer 20 form a parameter prediction model A, and the prediction result of the trained prediction model 10 becomes the output of the parameter prediction model A.
[0025] Step S4: Set the bias value of the activation function of each artificial neuron 21 to 0. When the input value of the activation function is 1, the output value of the activation function is 1.
[0026] Step S5: Based on at least one confirmed input parameter, set at least one first weight value of at least one first artificial neuron (not shown) among the artificial neurons 21 corresponding to the at least one confirmed input parameter respectively.
[0027] Step S6: Set the output of parameter prediction model A corresponding to the target result, and train parameter prediction model A by inputting a training dataset containing at least one all-one vector into the artificial neurons 21 of the dummy layer 20 of parameter prediction model A. Then, the optimizer (not shown) in the machine learning method that generated the trained prediction model 10 adjusts the second weight values of at least one artificial neuron 21 other than the first artificial neuron 21 based on the target result and at least one first weight value, and sets the second weight values as multiple unverified input parameters among the input parameters.
[0028] Furthermore, in the calculation method for the decision variables in this specific embodiment, the trained prediction model 10 includes multiple model parameters. All of these model parameters are fixed. Moreover, the first weight value is a fixed value, and variation and adjustment by the optimizer are not permitted. Only the second weight value of artificial neurons 21 other than the first artificial neuron can be adjusted by the optimizer.
[0029] In this specific embodiment, the trained predictive model in step S1 may be any machine learning model obtained from an open platform that has completed training, or it may be a model that the user has trained themselves. Furthermore, the dataset in this specific embodiment may be a set of data that can be used for training, testing, and validating machine learning. In actual applications, the method for calculating the decision variable in this specific embodiment can also be applied to any other trained machine learning model and used for inverse estimation of unconfirmed input parameters. In addition, the trained predictive model in this specific embodiment can be obtained by machine learning the dataset using artificial neural networks (ANNs), convolutional neural networks (CNNs), recurrent neural networks (RNNs), recursive neural networks (RecNNs), complex neural networks, 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.
[0030] In detail, the method for calculating the decision variables in this specific embodiment involves forming a parameter prediction model A by adding a dummy layer 20 connected to the trained prediction model 10 in step S3. The dummy layer 20 contains the same number of artificial neurons 21 as the input terminals 111 of the trained prediction model 10. Furthermore, a new link is established between each dummy artificial neuron 21 and the corresponding neuron at the input terminal 111 of the trained prediction model 10. In step S4, the bias value of the activation function of the artificial neurons 21 in the dummy layer 20 is set to 0. When the input value of the activation function is 1, the output value of the activation function is 1. In this specific embodiment, the activation function makes it possible to store its own numerical value in the second weight value. Subsequently, in step S6, the newly formed dummy layer 20 of the parameter prediction model A (at this point, the dummy layer 20 forms a new input layer of the parameter prediction model A) is trained by inputting a training dataset containing multiple all-one vectors. Here, each all-one vector represents data to which 1 is input for each artificial neuron 21. The aforementioned target result can become the ground truth output from parameter prediction model A. Furthermore, initial values for the weights of new links are selected and set, and the second weight value is considered the decision variable. Between the weight values, there may be other constraints, such as a range or a positive integer constraint. During the training process of parameter prediction model A described above, the optimizer adjusts the second weight value based on the target result, the training dataset, and the first weight value, enabling inverse estimation of the unconfirmed decision variable (unconfirmed input parameter).
[0031] More specifically, the optimizer is built into the training platform of the artificial intelligence engine and has the function of adjusting or fixing the parameters and weight values of the parameter prediction model A and the trained prediction model 10. Therefore, during the training process, the optimizer fixes or freezes the parameters of the trained prediction model 10, and then adjusts the weight values 30 between the artificial neuron 21 and the input terminal 111 of the trained prediction model 10, thereby minimizing the difference between the output value (prediction result) of the parameter prediction model A corresponding to each all-one vector in the training dataset and the ground truth (target result). It should be noted that the verified input parameters are fixed parameters, for example, process parameters that have already been executed within the process. Therefore, during the above training process, the optimizer also fixes or freezes the first weight values corresponding to the verified input parameters, and adjusts the other weight values (second weight values) other than the first weight values, thereby minimizing the difference between the output value (prediction result) of the parameter prediction model A and the ground truth (target result). At the end of training and upon convergence, the second weight value, adjusted by the optimizer, is combined with the fixed first weight value to form the optimal input decision variable. Parameter prediction model A has a many-to-one characteristic; that is, multiple different input parameters can correspond to the same output result. Therefore, each training data (all-one vector) has the potential to acquire one group of second weight values as the optimal input decision variable. More specifically, if the pre-trained prediction model used by the user has excellent prediction effects, it is impossible to inversely estimate the input decision variable based solely on the prediction effect and the pre-trained prediction model. However, according to the parameter prediction model in this specific embodiment, it is possible to inversely estimate the second weight value as an unconfirmed input parameter or decision variable by methods such as adding a dummy layer, fixing the bias value of the artificial neurons in the dummy layer, fixing the numerical values input during training, freezing the parameters of the pre-trained prediction model, freezing the first weight value, and minimizing the difference between the output value and the target result.
[0032] The following example illustrates the case of a cell process. A user intends to perform a 21-day process and has already completed 7 days of the process (i.e., already has confirmed input parameters). However, they may not have confirmed how to perform the process from day 8 to day 21 and how to adjust the decision variables (i.e., unconfirmed input parameters) to obtain the predetermined target result on day 21. In contrast, the above method allows setting the first weight value of the first artificial neuron among the artificial neurons corresponding to the confirmed input parameters based on the confirmed input parameters performed over 7 days, and setting the target result (for example, achieving a 95% cell viability rate in a 21-day cell culture process). Then, using the optimizer in the machine learning method that generated the trained predictive model, an all-one vector is input, and the second weight values of the other artificial neurons (excluding the first artificial neuron) are adjusted based on the difference between the target result and the output of the parameter prediction model and the first weight value, so that the second weight values become the decision variables (unconfirmed input parameters) for days 8 to 21. In other words, if the second weight value is adjusted to 2 by the optimizer, the unverified input parameter becomes 2.
[0033] In practical applications, optimizers for calculating decision variables include Adaptive Moment Estimation (Adam) optimization, Stochastic Gradient Descent (SGD), Momentum, Nesterov Accelerated Gradient (NAG), Adaptive Gradient Algorithm (AdaGrad), Nadam (Nesterov-accelerated Adaptive Moment Estimation), RMSprop (Root Mean Square Propagation), Adadelta (Adaptive Delta), AdamW (Adam with Weight Decay), AMSGrad (Adaptive Moment Estimation with Long-term Memory), AdaBelief (Adaptive Belief), LARS (Layer-wise Adaptive Rate Scaling), Self-adaptive Hessian (AdaHessian), and RAdam (Rectified You can choose from one of the following: Adam, Lookahead, MadGrad (Momentumized, Adaptive, and Decentralized Gradient Descent), Yogi Optimizer (Yogi), or AdamMax (Adaptive Moment Estimation with Maximum).
[0034] The optimizer for calculating the decision variables in this specific embodiment further includes an Adaptive Moment Estimation (Adam) optimizer (hereinafter referred to as the Adam optimizer). The Adam optimizer is also used to adjust the second weight values of the artificial neurons. The Adam optimizer is generally used to train various deep learning models, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), and transformers. The Adam optimizer can be applied to different tasks, such as image classification, natural language processing, and language translation. By adjusting the learning rate based on changes in the gradients of each parameter, the Adam optimizer allows for different learning rates in different directions. This contributes to better control over the convergence speed of machine learning training. In practical applications, the choice of optimizer is not limited to this; other optimizers with adjustable parameters and weight values can be used based on user needs and the type of training model.
[0035] The trained predictive model in this specific embodiment can be obtained by machine learning the dataset using a machine learning method. When the method for calculating the decision variables in this specific embodiment is applied to the field of cell culture processes, the dataset in this specific embodiment may further include a cell dataset. The cell dataset may include multiple cell samples. Furthermore, the sample parameters for each cell sample include source parameters and culture parameters. In practice, the cell dataset may be any data obtained from an open platform, or it may be data collected by the user themselves. In addition, the types of cell samples may be 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 in actual applications, it is not limited to these and should be determined according to the type of cell culture the user wants to perform. In addition, 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 a cell dataset may include data on the cell source and attribute data associated with that source. Attribute data may include physiological data of the source, or other data related to the source, such as the source's sex, age, medical history, living environment, and residential area. However, in practical applications, the source parameters of a cell sample may also include other parameters that may influence cellular processes and are related to the cell source. Furthermore, in practical applications, trained predictive models are selectable based on user needs and can be obtained by machine learning using different types of datasets.
[0036] 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.
[0037] The specific embodiments described above pertain to processes that include verified input parameters, i.e., processes that have already been started. However, the method for calculating decision variables in the present invention can also be applied to processes that have not yet been started and do not have verified input parameters. Refer to Figure 3 here. Figure 3 shows a flowchart of the steps of the method for calculating decision variables based on another specific embodiment of the present invention. The method for calculating decision variables in this specific embodiment is used to calculate multiple decision variables. The method includes the following:
[0038] Step S1': Provide a trained predictive model 10. The trained predictive model 10 is obtained by machine learning using a machine learning method. The trained predictive model 10 includes an input layer 11 and an output layer 12. The trained predictive model 10 is used to take multiple input parameters through the input layer 11 and generate prediction results corresponding to the input parameters through the output layer 12.
[0039] Step S2': Set a target result corresponding to the prediction result of the trained prediction model 10.
[0040] Step S3': A dummy layer 20 is added to the input layer 11 of the trained prediction model 10. The dummy layer 20 includes a plurality of artificial neurons 21, each connected to one of the input terminals 111 of the input layer 11. The trained prediction model 10 and the dummy layer 20 form a parameter prediction model A, and the prediction result of the trained prediction model 10 becomes the output of the parameter prediction model A.
[0041] Step S4': Set the bias value of the activation function of each artificial neuron 21 to 0. When the input value of the activation function is 1, the output value of the activation function becomes 1.
[0042] Step S5': The parameter prediction model A is trained by setting the output of the parameter prediction model A corresponding to the target result and inputting a training dataset containing at least one all-one vector into the artificial neurons 21 of the dummy layer 20 of the parameter prediction model A. Then, the optimizer in the machine learning method that generated the trained prediction model 10 adjusts the reference weight values of each artificial neuron 21 based on the target result, and sets these reference weight values as the decision variables.
[0043] The method for calculating the decision variable in this specific embodiment is suitable for processes that do not have confirmed input parameters, i.e., when the first weight value of the confirmed input parameters is an empty set. The optimizer adjusts the reference weight value of each artificial neuron by minimizing the difference between the output value and the target result, making inverse estimation possible. In other words, the reference weight value can be used as the decision variable (i.e., an unconfirmed decision variable). It should be noted that the other steps of the method for calculating the decision variable in this specific embodiment are almost the same as the corresponding steps in the specific embodiment described above, and therefore will not be described in detail again here.
[0044] In summary, the present invention provides a method for calculating decision variables. This method involves adding a dummy layer to the input terminals of a pre-trained neural network prediction model. The dummy layer contains the same number of artificial neurons as the input terminals of the pre-trained prediction model. A new link is established between each dummy neuron and the corresponding neuron in the original model's input terminal. The input value of each artificial neuron is set to 1. If the bias value of the activation function is 0 and the input to the activation function is 1, the output is 1. Initial values for the weights of the new links are then selected and set, and these weights are considered the decision variables. The weights may have a range or other mutual constraints. The parameters of the pre-trained prediction model are frozen, and only the weights of the new links are adjusted, while an optimizer built into a common machine learning platform finds the optimal solution. The training goal is set so that the output of this extended model becomes the desired target result. The weights of the new links obtained when training is complete become the executable input decision variables. Furthermore, the method for calculating decision variables in this invention may be applied not only to the field of cellular processes but also to other machine learning models trained by neural networks, and used for inverse estimation of decision variables and process parameters. This invention makes it possible to effectively search for optimal input parameters to achieve the desired results by utilizing a general machine learning platform and the methods built into it.
[0045] 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]
[0046] S1~S6, S1'~S5' Step A Parameter Prediction Model 10. Pre-trained predictive models 11 Input Layers 111 Input Terminals 12 Output Layers 20 Dummy Layers 21 Artificial Neurons 30 Weight values
Claims
1. A pre-trained predictive model is provided, the pre-trained predictive model is obtained by machine learning using a machine learning method, the pre-trained predictive model includes an input layer and an output layer, the pre-trained predictive model is used to input multiple input parameters through the input layer and to generate prediction results corresponding to the input parameters through the output layer, The steps include receiving a target result corresponding to the prediction result of the trained prediction model, and receiving at least one verified input parameter from among the input parameters, The steps include: adding a dummy layer connected to the input layer of the trained prediction model, the dummy layer including a plurality of artificial neurons connected to each input terminal of the input layer, forming a parameter prediction model with the trained prediction model and the dummy layer, and the prediction result of the trained prediction model becoming the output of the parameter prediction model; The steps include setting the bias value of the activation function of each artificial neuron to 0, and setting the output value of the activation function to 1 when the input value of the activation function is 1, The steps include setting at least one first weight value for at least one first artificial neuron among the artificial neurons corresponding to the at least one verified input parameter, based on the at least one verified input parameter, A method for calculating a decision variable, in which a computer performs the following steps: setting the output of the parameter prediction model corresponding to the target result; training the parameter prediction model by inputting a training dataset containing at least one all-one vector into the artificial neurons of the dummy layer of the parameter prediction model; and using an optimizer in the machine learning method that generated the trained prediction model, adjusting the second weight values of the artificial neurons other than the at least one first artificial neuron based on the target result and the at least one first weight value, and setting the second weight values to a plurality of unconfirmed input parameters among the input parameters.
2. The optimizer further supports Adaptive Moment Estimation (Adam) optimization, Stochastic Gradient Descent (SGD), Momentum, Nesterov Accelerated Gradient (NAG), Adaptive Gradient Algorithm (AdaGrad), Nadam (Nesterov-Accelerated Adaptive Moment Estimation), and RMSprop (Root Mean Square). Propagation), Adadelta (Adaptive Delta), AdamW (Adam with Weight Decay), AMSGrad (Adaptive Moment Estimation with Long-term Memory), AdaBelief (Adaptive Belief), LARS (Layer-wise Adaptive Rate) Scaling), self-adaptive Hessian (AdaHessian), RAdam (Rectified Adam), lookahead (Lookahead), MadGrad (Momentumized, Adaptive, and A method for calculating a decision variable to be executed by a computer according to claim 1, selected from any of the following: Decentralized Gradient Descent, Yogi optimizer (Yogi), and AdamMax (Adaptive Moment Estimation with Maximum).
3. The method for calculating a computer-executed decision variable according to claim 1, wherein the trained predictive model is obtained by machine learning a dataset using the machine learning method.
4. The method for calculating a decision variable performed by a computer according to claim 1, wherein the trained predictive model includes a plurality of model parameters, all of which are fixed.
5. The method for calculating a computer-executed decision variable according to claim 3, 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), Recurrent Neural Networks (RNNs), Recursive Neural Networks (RecNN), and Complex Neural Networks.
6. Used to calculate multiple decision variables, A pre-trained predictive model is provided, the pre-trained predictive model is obtained by machine learning using a machine learning method, the pre-trained predictive model includes an input layer and an output layer, the pre-trained predictive model is used to input multiple input parameters through the input layer and to generate prediction results corresponding to the input parameters through the output layer, The steps include receiving a target result corresponding to the prediction result of the aforementioned trained prediction model, The steps include: adding a dummy layer connected to the input layer of the trained prediction model, the dummy layer including a plurality of artificial neurons connected to each input terminal of the input layer, forming a parameter prediction model with the trained prediction model and the dummy layer, and the prediction result of the trained prediction model becoming the output of the parameter prediction model; The steps include setting the bias value of the activation function of each artificial neuron to 0, and setting the output value of the activation function to 1 when the input value of the activation function is 1, A method for calculating a decision variable, in which a computer performs the following steps: setting the output of the parameter prediction model corresponding to the target result; training the parameter prediction model by inputting a training dataset containing at least one all-one vector into the artificial neurons of the dummy layer of the parameter prediction model; and adjusting the reference weight value of each artificial neuron based on the target result using an optimizer in the machine learning method that generated the trained prediction model, and setting the said reference weight value as the decision variable.
7. The optimizer further supports Adaptive Moment Estimation (Adam) optimization, Stochastic Gradient Descent (SGD), Momentum, Nesterov Accelerated Gradient (NAG), Adaptive Gradient Algorithm (AdaGrad), Nadam (Nesterov-Accelerated Adaptive Moment Estimation), and RMSprop (Root Mean Square). Propagation), Adadelta (Adaptive Delta), AdamW (Adam with Weight Decay), AMSGrad (Adaptive Moment Estimation with Long-term Memory), AdaBelief (Adaptive Belief), LARS (Layer-wise Adaptive Rate) Scaling), self-adaptive Hessian (AdaHessian), RAdam (Rectified Adam), lookahead (Lookahead), MadGrad (Momentumized, Adaptive, and A method for calculating a decision variable to be performed by a computer according to claim 6, which is selected from any of the following: Decentralized Gradient Descent, Yogi optimizer (Yogi), and AdamMax (Adaptive Moment Estimation with Maximum).
8. The method for calculating a computer-executed decision variable according to claim 6, wherein the trained predictive model is obtained by machine learning the dataset using the machine learning method.
9. The method for calculating a decision variable performed by a computer according to claim 6, wherein the trained predictive model includes a plurality of model parameters, all of which are fixed.
10. The method for calculating a computer-executed decision variable according to claim 8, 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), Recurrent Neural Networks (RNNs), Recursive Neural Networks (RecNN), and Complex Neural Networks.