AI graphical programming platform and method using supervised learning

Abstract

The invention discloses an AI graphical programming platform and method utilizing supervised learning. The method comprises the following steps: initializing an AI model, and configuring at least one of the following parameters for the AI model: an input layer parameter, an output layer parameter and a hidden layer parameter; obtaining a data set by using supervised learning, training an AI model by using training data in the data set, and optimizing the AI model by using a backward propagation algorithm; and obtaining a current characteristic value of the AI role, and based on the current characteristic value, automatically making a decision by using the AI model so as to control the behavior of the AI role. According to the method, the technical problems of low accuracy and training efficiency of the AI model in a graphical programming method are solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to an AI graphical programming platform and method using supervised learning. Background Art

[0002] Supervised learning is a technical method that occupies a core position in the field of machine learning. By learning and training on labeled data, the model can accurately predict or classify new data. In this process, the data set consists of input features and corresponding labels, and the training goal is to find an optimal function mapping so that when receiving new input, its label can be correctly predicted.

[0003] Existing supervised learning techniques face the problem of high computational complexity when dealing with large-scale data sets and high-dimensional features. As the amount of data and feature dimensions increase, the model training and prediction process consumes a lot of computing resources, making the algorithm implementation inefficient in a big data environment and limiting the scope of its practical application. Secondly, existing models often show poor robustness when dealing with noisy data and outliers, which easily leads to overfitting or underfitting of the model. In addition, these algorithms have poor scalability in distributed environments and are difficult to achieve efficient expansion in distributed computing frameworks. Finally, the long training time of complex models seriously affects their feasibility in real-time applications and limits the application of supervised learning techniques in application scenarios that require rapid response.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide an AI graphical programming platform and method using supervised learning, so as to at least solve the technical problems of low accuracy and low training efficiency of the AI ​​model in the graphical programming method.

[0006] According to one aspect of an embodiment of the present invention, there is provided an AI graphical programming method using supervised learning, comprising: initializing an AI model, and configuring at least one of the following parameters for the AI ​​model: input layer parameters, output layer parameters, and hidden layer parameters; acquiring a data set using supervised learning, training the AI ​​model using training data in the data set, and optimizing the AI ​​model using a backpropagation algorithm; acquiring a current feature value of an AI character, and based on the current feature value, using the AI ​​model to automatically make a decision to control the behavior of the AI ​​character.

[0007] According to another aspect of an embodiment of the present invention, there is also provided an AI graphical programming platform using supervised learning, including: an initialization module, configured to initialize an AI model, and configure at least one of the following parameters for the AI ​​model: input layer parameters, output layer parameters, and hidden layer parameters; a training module, configured to obtain a data set using supervised learning, train the AI ​​model using the training data in the data set, and optimize the AI ​​model using a backpropagation algorithm; a decision module, configured to obtain a current feature value of an AI character, and based on the current feature value, automatically make a decision using the AI ​​model to control the behavior of the AI ​​character.

[0008] In an embodiment of the present invention, the AI ​​model is initialized, and at least one of the following parameters is configured for the AI ​​model: input layer parameters, output layer parameters, and hidden layer parameters; a data set is obtained by supervised learning, the AI ​​model is trained by the training data in the data set, and the AI ​​model is optimized by a back propagation algorithm; the current feature value of the AI ​​character is obtained, and based on the current feature value, the AI ​​model is used to automatically make a decision to control the behavior of the AI ​​character. Through the above scheme, the technical problems of low accuracy and low training efficiency of the AI ​​model in the graphical programming method are solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0010] Figure 1 is a flowchart of an AI graphical programming method using supervised learning according to an embodiment of the present invention;

[0011] Figure 2 is an architecture diagram of an improved graphical programming platform for supervised learning according to an embodiment of the present invention;

[0012] Figure 3 is a workflow diagram of a graphical programming platform according to an embodiment of the present invention;

[0013] Figure 4 is a flowchart of an AI graphical programming method using supervised learning according to an embodiment of the present invention;

[0014] Figure 5 is an interface diagram of a graphical code for initializing an AI model according to an embodiment of the present invention;

[0015] Figure 6 is an interface diagram of a graphical code for setting input layer information according to an embodiment of the present invention;

[0016] Figure 7 is a schematic diagram of setting a center point according to an embodiment of the present invention;

[0017] Figure 8 This is an interface diagram that encapsulates a flight event triggered by a mouse click using a function building block according to an embodiment of the present invention;

[0018] Fig. 9 It is an interface diagram of a graphical code for setting a usage behavior building block call of an AI character according to an embodiment of the present invention;

[0019] Fig.10 is an interface diagram of a graphical code for setting a death condition of an AI character according to an embodiment of the present invention;

[0020] Fig.11 is an interface diagram of a graphical code for setting an AI character feature value according to an embodiment of the present invention;

[0021] Fig.12 is an interface diagram of a graphical code for perfecting input and output quantities according to an embodiment of the present invention;

[0022] Fig.13 is an interface diagram of a flying bird according to an embodiment of the present invention;

[0023] Fig.14 is an interface diagram of a graphical code for recording characteristic values ​​according to an embodiment of the present invention;

[0024] Fig.15 is an interface diagram of a graphical code for optimizing an AI model using a back propagation algorithm according to an embodiment of the present invention;

[0025] Fig.16 is an interface diagram of a graphical code for removing an initialization model building block according to an embodiment of the present invention;

[0026] Fig.17 is an interface diagram of a graphical code for AI to make a decision according to an embodiment of the present invention;

[0027] Fig.18 It is an architecture diagram of another improved graphical programming platform for supervised learning according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] According to an embodiment of the present invention, a method embodiment of an AI graphical programming method using supervised learning is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0031] Figure 1 is an AI graphical programming method using supervised learning according to an embodiment of the present invention, such as Figure 1 As shown, the method comprises the following steps:

[0032] Step S102, initialize the AI ​​model, and configure at least one of the following parameters for the AI ​​model: input layer parameters, output layer parameters, and hidden layer parameters.

[0033] The parameters include at least input layer parameters, output layer parameters and hidden layer parameters. The input layer parameters, output layer parameters and hidden layer parameters may include the number of layers, the number of iterations, etc. In addition, the input layer parameters may also include the number and type of input features that the AI ​​model will process, such as the Y coordinate value of the bird and the Y coordinate value of the obstacle in the game scene. The output layer parameters may also include the output content of the AI ​​model, such as the operation options of the bird, which may be "no operation" or "fly". In addition, the configuration of the hidden layer parameters is used to set the number of hidden layers and the number of neurons in each layer. By reasonably configuring these parameters, the AI ​​model will have the basis for processing input data and generating reasonable outputs.

[0034] Step S104, using supervised learning to obtain a data set, using the training data in the data set to train the AI ​​model, and using a back propagation algorithm to optimize the AI ​​model.

[0035] First, the data set is preprocessed. Specifically, the data is cleaned to detect and remove missing values ​​and outliers in the data set. Missing values ​​may lead to incomplete information during model training, while outliers may cause misjudgment or overfitting of the model. Therefore, cleaning these data helps to improve the stability and prediction accuracy of the model. Next, the cleaned data is standardized. The purpose of standardization is to scale different features in the data to the same scale to eliminate the magnitude differences between the features, thereby preventing some features with large values ​​from having unnecessary effects on the model, so that the model treats all features equally during the training process, thereby improving the training effect and the convergence speed of the model. Then, feature selection is performed based on the importance of the features. By calculating the importance index of each feature, the feature vectors that are most critical to the model prediction results can be screened out. Usually, feature vectors with an importance greater than a certain set threshold are selected as principal components, which helps to simplify the model, remove redundant features, thereby reducing the complexity of the model, and improving training efficiency and prediction accuracy. Finally, the selected principal components are used to reduce the dimensionality of the data set. Dimensionality reduction can not only reduce the dimension of the data set and reduce computational complexity, but also reduce the interference of noise, thereby improving the generalization ability of the model. Through the above steps, the most representative features in the data set are retained, while those features that contribute less to the model prediction are discarded, making the model more efficient and more robust.

[0036] Next, train the AI ​​model. The preprocessed dataset is divided into multiple subsets. This division can be achieved through random splitting, cross-validation, or stratified sampling. The purpose of dividing the dataset into subsets is to improve the generalization ability of the model, and it can also accelerate the training process through parallel processing. On each subset, the AI ​​model is trained separately to ensure that the model can learn the characteristics of different parts of the dataset. During the model training process, the gradient of the loss function relative to the model parameters must be calculated first. The loss function measures the difference between the model's predicted value and the true value, and the gradient represents the rate of change of this difference relative to the model parameters. By calculating the gradient, it is possible to determine how to adjust the model parameters to minimize the loss function so that the model can predict the results more accurately. Based on the calculated gradient, the first-order moment estimate and the second-order moment estimate are used to adjust the learning rate of the AI ​​model. The first-order moment estimate usually corresponds to the average value of the gradient, while the second-order moment estimate corresponds to the variance of the gradient. By combining these two estimates, the learning rate can be adjusted dynamically, making the model parameter update more stable and efficient. The first-order moment estimate ensures the directionality of the gradient descent, while the second-order moment estimate helps prevent the learning rate from being too large or too small, thereby avoiding model oscillation or slow convergence.

[0037] Finally, based on the calculated gradient and the adjusted learning rate, the parameters of the AI ​​model are updated. Parameter update is achieved through the back-propagation algorithm, which is an efficient optimization algorithm that passes error information layer by layer through the chain rule to accurately adjust the parameters of each layer of the model.

[0038] Specifically, the back propagation algorithm starts from the output layer of the model, calculates the gradient of the error layer by layer, and passes these gradients to the previous hidden layer and input layer through the chain rule, so that the prediction error of the model gradually decreases. This layer-by-layer transmission and update mechanism ensures that the model can gradually approach the global optimal solution through multiple iterations.

[0039] In order to ensure the effectiveness of parameter updates, this application uses the learning rate calculated previously to adjust the gradient. The adjustment of the learning rate is very important, as it determines the pace of each update. If the learning rate is too high, the model parameters may fluctuate back and forth near the optimal solution and be difficult to converge; if the learning rate is too low, the model may fall into a local optimum, resulting in too slow training speed. Therefore, by combining the first-order moment and second-order moment estimation methods, the learning rate can be dynamically adjusted to meet the needs of different training stages, thereby improving training efficiency and model performance.

[0040] With each parameter update, the model's loss function gradually decreases, and the model's predictive ability continues to increase. Through multiple rounds of iterative training, the model gradually learns and masters the patterns and features in the data until it reaches the predetermined training rounds or the model's convergence conditions. When the model's parameters are adjusted to the point where the loss function no longer decreases significantly, the training process is complete, and the AI ​​model now has good generalization capabilities and can make accurate predictions on new data.

[0041] Through this systematic training method, the AI ​​model can fully learn the patterns and regularities in the data, thereby having higher predictive ability and generalization performance.

[0042] Step S106, obtaining the current feature value of the AI ​​character, and automatically making a decision using the AI ​​model based on the current feature value to control the behavior of the AI ​​character.

[0043] The current feature value of the AI ​​character is obtained, and the current feature value is input into the AI ​​model, and the AI ​​model automatically makes a behavior decision and outputs the behavior of the AI ​​character. Then, the behavior of the AI ​​character is visually displayed in the stage area of ​​the graphical programming platform.

[0044] The present invention provides an improved graphical programming platform for supervised learning, such as Figure 2As shown, it includes: a data preprocessing module 22, a model training module 24, a prediction module 26 and a graphical programming interface 28. The data preprocessing module 22 is used to clean, standardize and select features for the input data. The model training module 24 uses an optimization algorithm to train the model and supports a variety of supervised learning algorithms. The model training module 24 uses a distributed computing framework to achieve parallel processing and improve training efficiency. The prediction module 26 is used to predict new input data and output the results. The graphical programming interface 28 is a user interface for simplifying the programming process, which usually visualizes the programming components in a drag-and-drop manner to help users build program logic more intuitively.

[0045] The workflow of the graphical programming platform is as follows Figure 3 As shown, the following steps are included:

[0046] Step S302: data preprocessing.

[0047] First, the input data is cleaned comprehensively. The cleaning process includes removing missing values ​​and outliers in the data set to ensure the integrity and consistency of the data, so as to provide a reliable basis for subsequent model training. Next, the cleaned data is standardized. The purpose of standardization is to eliminate the dimensional differences between different features so that each feature can be compared on the same scale. Specifically, the standardization process is completed by the following formula:

[0048] Xstandardized=(X-mean(X)) / std(X)

[0049] Among them, X represents the original data value, mean represents the mean (i.e., average value) of the feature, and std represents the standard deviation of the feature. Through this formula, all data are converted to a standard normal distribution with a mean of 0 and a standard deviation of 1, reducing the imbalance between features. Finally, through the feature selection process, the top k most critical features are selected based on feature importance. In this way, not only the dimension of the data is reduced, but also the generalization ability and computational efficiency of the model are improved.

[0050] Step S304: model training.

[0051] After data preprocessing, the model training phase begins. First, the module divides the processed data set into multiple subsets so that the model can be trained in parallel on different computing nodes using a distributed computing architecture. In the initial stage of training, the model parameters θ are initialized. Here θ refers to the set of parameters in the model, which determines the model's behavior and predictive ability.

[0052] Model training is optimized through the Gradient Descent algorithm. The specific process is as follows:

[0053]

[0054] Among them, J(θ) represents the loss function, which is used to measure the difference between the predicted results of the current model and the actual results. The specific form of the loss function varies depending on the supervised learning algorithm used. is the gradient of the loss function relative to the model parameter θ, indicating the direction and rate of change of the loss function under the current parameters. The learning rate α is a hyperparameter used to control the step size of parameter updates, that is, the amplitude of each parameter adjustment. In order to improve the convergence speed of the model, an adaptive learning rate mechanism is used during training. This mechanism can dynamically adjust the learning rate α according to the performance of the model during training, thereby accelerating the convergence of the model and avoiding falling into a local optimal solution.

[0055] After independent training on each subset is completed, the model parameters θ obtained from the training on each subset will be aggregated to form the updated global model parameters. This ensures that the model can effectively integrate the learning results from different subsets when processing large-scale data, thereby improving the overall performance and prediction accuracy of the model.

[0056] Step S306, making predictions.

[0057] Input new data, use the trained model to make predictions, and output the prediction results.

[0058] The present invention can realize parallel processing and model training of large-scale data by introducing a distributed computing framework, thereby greatly improving computing efficiency. At the same time, combined with data preprocessing and optimization algorithms, the impact of noise data and outliers on the model is effectively reduced, thereby significantly enhancing the robustness of the model. In addition, through adaptive learning rate and optimization algorithm, the present invention accelerates the convergence speed of the model and significantly shortens the training time. Optimized data preprocessing and algorithm selection further improve the prediction accuracy of the model. The overall process of the present invention is shown in the flowchart. By rationally configuring each module, the effect of significantly improving computing efficiency, shortening training time, enhancing model robustness and improving prediction accuracy is achieved.

[0059] The present application embodiment takes the flying bird game as an example to provide an AI graphical programming method using supervised learning, such as Figure 4 As shown, the method includes:

[0060] Step S402, initializing the model.

[0061] Select the bird as the AI ​​character, such as Figure 5 As shown, an AI model, Model 1, is initialized. The data of the input layer and the output layer are left empty, and the hidden layer can be adjusted to 4. For the flying bird game, it is recommended to set it to a single layer of 4 neurons.

[0062] Step S404, determining input and output information.

[0063] 1) Determine the information of the input layer.

[0064] like Figure 6 As shown in the figure, the two selected features are the Y coordinate values ​​of the "bird" and "under the pillar" characters. It should be noted that the center point should be set as follows Figure 7 Position shown: Middle of the bird and top center under the post.

[0065] 2) Determine the information of the output layer.

[0066] When playing the flying bird game, there are two operations: do nothing and click the mouse. Corresponding to GameAI, there are two output results: no operation and make the bird fly. Due to the setting of the mechanism, in order to call the use behavior of the AI ​​character, you must first use the function building block to encapsulate the flight event triggered by the mouse click. Figure 8 shown.

[0067] Then, set the AI ​​character's usage behavior block call, such as Fig. 9 shown.

[0068] 3) Set the character's death conditions.

[0069] like Fig.10 Set the death condition of the AI ​​character as shown, and then set the AI ​​character feature value read in real time, such as Fig.11 As shown in the figure. The AI ​​character feature values ​​are the Y coordinate of the AI ​​character and the Y coordinate under the pillar. Unlike the classifier, the input feature values ​​in GameAI change in real time.

[0070] 4) Improve the input and output quantities.

[0071] After confirming the input and output information, you need to fill in the number of input and output layers at the location of the initialization model to complete the setting of the entire model, such as Fig.12 As shown in the figure, the input layer input features are "the Y coordinate of the bird" and the Y coordinate of "under the pillar", so the input layer is set to 2; the output layer outputs "no operation" and "flying", so the output layer is also set to 2.

[0072] Step 406: training and recording data.

[0073] After setting up the model and input and output information, we need to start training the AI ​​model, i.e., Model 1. In supervised learning mode, Fig.13 As shown, you need to play the game yourself first and record the operations in the game, such as Fig.14As shown, it is used as training data. Continuously train and record, and after reaching the preset threshold, stop training and save it in model 1. After that, learning using the training data obtained above is supervised learning.

[0074] Step S408, back propagation algorithm.

[0075] When an AI character dies, Fig.15 As shown in the figure, you can use the "back propagation algorithm" to continue to optimize the selected model. After optimizing according to the recorded training data, the model will continue to act by imitating the behavior in the training data. After the training is completed, if you want to keep the previous training results, you need to remove the initialization model block and run it again. Fig.16 shown.

[0076] In some embodiments, after the AI ​​model is trained, the score of the fitness function can also be used to evaluate the strategy made. The score of the fitness function can be obtained based on the following: determining the weight of the performance indicator based on the environmental state; obtaining the weighted sum of the AI ​​strategy on different performance indicators based on the weight of the performance indicator and the performance indicator; determining the contribution of diversity to fitness based on the diversity score and diversity weight coefficient of the AI ​​strategy; determining the contribution of robustness to fitness based on the robustness score and robustness weight coefficient of the AI ​​strategy; finally, determining the score of the fitness function based on the weighted sum of the AI ​​strategy on different performance indicators, the contribution of diversity to fitness, and the contribution of robustness to fitness.

[0077] Specifically, the fitness function can be expressed as:

[0078]

[0079] Among them, D(AI) represents the diversity score of the AI ​​strategy, and γ is the weight coefficient used to adjust the influence of diversity in the fitness function. E represents the environmental state, which describes the external environment in which the AI ​​strategy is located, including dynamically changing factors such as data distribution, network conditions, input and output requirements, etc. Mi(AI) represents the i-th performance indicator. wi(E) represents the weight of the i-th performance indicator, R(AI) represents the robustness score of the strategy, and δ is the robustness weight coefficient, which is used to adjust the influence of the robustness score R(AI) on the fitness function.

[0080] This community evaluates AI strategies by introducing fitness functions, and achieves a comprehensive and dynamic evaluation of strategies after AI model training. The score of the fitness function not only takes into account the performance of the AI ​​strategy on different performance indicators, but also dynamically adjusts the weights to adapt to changes in the external environment, so that the AI ​​strategy can be optimized according to changes in the environment. In addition, the introduction of diversity scores and robustness scores ensures that the AI ​​strategy has greater stability and consistency while maintaining diversity, avoiding the risk of the strategy falling into a local optimal solution in a specific environment. By comprehensively considering different performance indicators, diversity, and robustness, this method effectively improves the adaptability and robustness of the AI ​​strategy, allowing the AI ​​model to play a better role in complex and changeable practical applications.

[0081] Step S410, using the model.

[0082] Once the model is trained, you can use the model to let the AI ​​character make decisions and play the game automatically. Place the "AI character makes decisions" block under the AI ​​character feature block. In this way, the two feature values ​​used by the AI ​​character, namely the Y coordinate of the AI ​​character obtained in real time and the Y coordinate under the column, are input into the AI ​​model as training data. The AI ​​can make decisions on the use behavior based on the current feature values, that is, "no operation" or "fly". To realize the automatic game of the AI ​​character, such as Fig.17 shown.

[0083] This application also provides an AI graphical programming platform using supervised learning, such as Fig.18 As shown, it includes: an initialization module 82 configured to initialize the AI ​​model and configure at least one of the following parameters for the AI ​​model: input layer parameters, output layer parameters, and hidden layer parameters; a training module 84 configured to obtain a data set using supervised learning, train the AI ​​model using the training data in the data set, and optimize the AI ​​model using a back propagation algorithm; a decision module 86 configured to obtain the current feature value of the AI ​​role, and based on the current feature value, automatically make a decision using the AI ​​model to control the behavior of the AI ​​role

[0084] It should be noted that the AI ​​graphical programming platform using supervised learning provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the AI ​​graphical programming platform using supervised learning provided in the above embodiment and the AI ​​graphical programming method embodiment using supervised learning belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0085] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. An AI graphical programming method using supervised learning, characterized in that: include: Initialize the AI ​​model, and configure at least one of the following parameters for the AI ​​model: input layer parameters, output layer parameters, and hidden layer parameters; Obtaining a data set using supervised learning, training an AI model using training data in the data set, and optimizing the AI ​​model using a back-propagation algorithm; The current feature value of the AI ​​character is obtained, and based on the current feature value, the AI ​​model is used to automatically make a decision to control the behavior of the AI ​​character.

2. The method according to claim 1, characterized in that Using the training data in the dataset to train the AI ​​model includes: Preprocessing the data set, dividing the preprocessed data set into multiple subsets, and training the AI ​​model on each subset; Aggregate the parameters of the AI ​​model on each subset, and update the global parameters of the AI ​​model based on the aggregated parameters.

3. The method according to claim 2, characterized in that Preprocessing the data set includes: performing at least one of the following processing on the data set: Clean the data in the data set to remove missing values ​​and outliers in the data set; The cleaned data set is subjected to data standardization, and feature vectors whose importance is greater than a threshold are selected as principal components according to feature importance, and the data in the data set are subjected to dimensionality reduction using the selected principal components.

4. The method according to claim 1, characterized in that: Training the AI ​​model includes: Calculate the gradient of the loss function with respect to the model parameters of the AI ​​model; Adjusting the learning rate of the AI ​​model based on the first-order moment estimate and the second-order moment estimate of the gradient; A model parameter of the AI ​​model is updated based on the gradient and the learning rate.

5. The method according to claim 4, characterized in that Based on the current feature value, automatically making a decision using the AI ​​model includes: inputting the current feature value into the AI ​​model, the AI ​​model automatically making a behavior decision, and outputting the behavior of the AI ​​character.

6. The method according to any one of claims 1 to 5, characterized in that The behavior of the AI ​​character is visually displayed in the stage area of ​​the graphical programming platform.

7. An AI graphical programming platform using supervised learning, characterized in that: include: An initialization module, configured to initialize the AI ​​model and configure at least one of the following parameters for the AI ​​model: an input layer parameter, an output layer parameter, and a hidden layer parameter; A training module is configured to acquire a data set using supervised learning, train an AI model using training data in the data set, and optimize the AI ​​model using a back propagation algorithm; The decision-making module is configured to obtain the current feature value of the AI ​​character, and based on the current feature value, automatically make a decision using the AI ​​model to control the behavior of the AI ​​character.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 6.

9. A computer device, characterized in that: include: Memory and processor, The memory stores a computer program; The processor is used to execute the computer program stored in the memory, and when the computer program is run, the processor is enabled to execute the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images