Method and system for intelligent decision making in freely-playing card game

Abstract

The invention provides an intelligent decision making method and system in a free card playing chess and card game, equipment and a storage medium. According to the method provided by the invention, the game visible information under the view angle of the current player and all actions which can be executed by the current player are obtained, and then the main board decision result is obtained through reasoning of the pre-trained attention module according to the information; and then inputting the main card decision result and the game visible information into a pre-trained Transform module for decoding to obtain a card leaving decision result, and further obtaining card showing selection under the current situation. Through the method provided by the invention, a more accurate and effective decision can be made for the situation under the complex game rule of free card playing; meanwhile, various variables in the game can be fully considered by combining the design of the state space and the models of different types, so that comprehensive information is considered in the decision making process, the decision making of the model in each stage is more interpretable, and the reasonability of the decision making model is further improved.

Description

Technical Field

[0001] The present application relates to the field of information interaction technology, and in particular to a method and system for making intelligent decisions in a free-play card chess and card game, a computing device, and a computer-readable storage medium. Background Art

[0002] With the development of the electronic game industry, the application of artificial intelligence technology in games is becoming increasingly popular. In card games, in order to solve problems such as accidental disconnection and in the intelligent prompt function, it has become a common solution to use AI to temporarily replace players in decision-making. In some adapted gameplays of card games, players can choose any combination of cards to play cards. The decision-making scheme in the existing technology makes decisions through predefined rule sets and behavior trees. However, this method is highly dependent on personal level and subjective decision-making, and it is difficult to cover all possible game situations; on the other hand, the rule set method lacks flexibility and adaptiveness, and cannot adjust and optimize decisions in real time. It is easy to have rigid and single behavioral decisions, which in turn affects the player experience. Therefore, there is an urgent need for an intelligent decision-making solution that can respond flexibly in free-playing chess and card games. Summary of the invention

[0003] In view of this, the embodiments of the present application provide a method and system for making intelligent decisions in a free-play card game, a computing device, and a computer-readable storage medium to address the technical defects existing in the prior art.

[0004] According to a first aspect of an embodiment of the present application, there is provided a method for making intelligent decisions in a free-play card game, the method comprising:

[0005] Continuously obtain the game visible information and executable actions of the current player in the game, wherein the game visible information includes card characteristics and situation status characteristics;

[0006] After the card features and situation status features are processed by the fully connected layer respectively, the imperfect information feature vector is obtained by splicing; the executable action is processed by the fully connected layer to obtain the action feature vector;

[0007] Inputting the action feature vector and the imperfect feature information vector into the attention module for inference to obtain a main card decision result;

[0008] The main card decision result is concatenated with the imperfect feature information vector, and the concatenated result is decoded using a Transformer model to obtain a card retention action sequence;

[0009] Remove the cards in the retaining card action sequence from the current hand to obtain a decision result.

[0010] According to a second aspect of an embodiment of the present application, a system for making intelligent decisions in a free-play card game is provided, comprising:

[0011] An extraction module, used to continuously obtain the game visible information and executable actions of the current player in the game, wherein the game visible information includes card features and situation status features;

[0012] The encoding module is used to perform full connection layer processing on the card features and situation state features respectively, and then concatenate them to obtain an imperfect feature information vector; and perform full connection layer processing on the executable action to obtain an action feature vector;

[0013] A main card decision module, used for inputting the action feature vector and the imperfect feature information vector into an attention module for inference to obtain a main card decision result;

[0014] A card retention decision module is used to concatenate the main card decision result with the imperfect feature information vector, and then use the Transformer model to decode the concatenation result to obtain a card retention action sequence;

[0015] The card-playing decision module is used to remove the cards in the card-retaining action sequence from the current hand and obtain a decision result.

[0016] According to a third aspect of an embodiment of the present application, a computing device is provided, comprising a memory, a processor, and computer instructions stored in the memory and executable on the processor, wherein when the processor executes the instructions, the steps of the above-described method for making intelligent decisions in a free-play card game are implemented.

[0017] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, which stores computer instructions, and when the instructions are executed by a processor, the steps of the above-mentioned method for making intelligent decisions in a free-play card game are implemented.

[0018] The embodiment of the present application provides a flexible, efficient and human-thinking decision-making method, which is particularly suitable for the decision-making process of card games with free play. The method obtains the visible information of the game from the perspective of the current player and all the actions that the current player can perform, and then obtains the main card decision result through the pre-trained attention module inference based on the above information, and then inputs the main card decision result and the visible information of the game into the pre-trained Transformer module for decoding to obtain the decision result of keeping the card, and then obtains the card-playing choice under the current situation. Through the method provided by the present application, more accurate and effective decisions can be made on the situation under the complex game rules of free play; at the same time, the design of the state space and the combination of different types of models can fully consider various variables in the game, so as to take into account comprehensive information in the decision-making process, and also make the decisions of the models at each stage more interpretable, further improving the rationality of the decision-making model. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a structural block diagram of a computing device provided in an embodiment of the present application;

[0020] Figure 2 It is a flowchart of a method for making intelligent decisions in a free-playing chess and card game provided by an embodiment of the present application;

[0021] Figure 3 This is a schematic diagram of a game interface of a free-playing chess and card game provided in an embodiment of the present application;

[0022] Figure 4 It is a structural diagram of a system for making intelligent decisions in a free-play card game provided by an embodiment of the present application. DETAILED DESCRIPTION

[0023] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.

[0024] The terms used in one or more embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of the present application. The singular forms of "a", "said" and "the" used in one or more embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of the present application refers to and includes any or all possible combinations of one or more associated listed items.

[0025] It should be understood that, although the terms first, second, etc. may be used to describe various information in one or more embodiments of the present application, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of the present application, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "in response to determination".

[0026] In the present application, a method and system for making intelligent decisions in a free-play card game, a computing device, and a computer-readable storage medium are provided, which are described in detail one by one in the following embodiments.

[0027] Figure 1 The structure block diagram of a computing device 100 according to an embodiment of the present application is shown. The components of the computing device 100 include but are not limited to a memory 110 and a processor 120. The processor 120 is connected to the memory 110 via a bus 130, and the database 150 is used to store data.

[0028] The computing device 100 also includes an access device 140 that enables the computing device 100 to communicate via one or more networks 160. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 140 may include one or more of any type of network interface (e.g., a network interface card (NIC)) that is wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a World Wide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, a near field communication (NFC) interface, and the like.

[0029] In one embodiment of the present application, the above components of the computing device 100 and Figure 1 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Figure 1 The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of the present application. Those skilled in the art may add or replace other components as needed.

[0030] The computing device 100 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or PC. The computing device 100 may also be a mobile or stationary server.

[0031] In some adapted card games, the game can have any number of players, and each player can also choose any combination of cards in the game interface to play cards. Since the card-playing data is an arbitrary combination and the number of players is unlimited, the logical complexity of the game is greatly increased. In the intelligent decision-making process of card games, the existing technology usually uses a behavior tree method based on a rule set, but the AI ​​using behavior trees is highly dependent on expert experience and manual writing, and it is difficult to cover all possible game situations in the above-mentioned adapted gameplay, and the effect is highly dependent on the expert level and subjective judgment. Moreover, this method lacks flexibility and adaptive capabilities, and cannot adjust and optimize decisions in real time. It is easy to have rigid and single behavioral decisions, which in turn affects the player experience.

[0032] Therefore, the embodiment of the present application first proposes a method for making intelligent decisions in a free-play card game. The processor 120 may execute Figure 2 The steps of a method for making intelligent decisions in a free-play card game include steps 202 to 210.

[0033] Step 202: Continuously obtain the visible game information and executable actions of the current player in the game, wherein the visible game information includes card features and situation status features.

[0034] In this step, after the game starts, the visible information of the game from the perspective of the target player in the game is continuously obtained according to the instruction. The game in the above step may refer to a free-playing chess and card game between two or more people. Figure 3As shown, the player client can display N hand card images in the game interface of this round of the game, and the player can select any M hand card images among the N hand card images to perform a card-playing operation. The player client can combine the hand card data of the hand card images selected by the player according to the card-playing operation of the player to obtain the player's card-playing, and then play the player's card-playing. Optionally, N is a preset value, which can be determined according to the number of game rounds. For example, in the first round of the game, N can be 6, in the second round of the game, N can be 8, in the third round of the game, N can be 10, in the fourth round of the game, N can be 10, and in the fifth round of the game, N can be 12. M is the number of hands arbitrarily selected by the player, which can be any value.

[0035] The visible information of the game includes the characteristics of the cards in the current game. Specifically, in a free-playing chess and card game, the characteristics of the cards include but are not limited to:

[0036] The primary player's current hand;

[0037] Other players’ exposed cards, such as Figure 3 The middle pair has 8 and 9, the upper pair has A and K, and the lower pair has a pair of 7;

[0038] All players’ historical card plays, i.e., the players’ historical card play records in previous rounds;

[0039] and the current round, such as Figure 3 The current round is the 3rd round in a total of 5 rounds.

[0040] Furthermore, the game visible information also includes the situation status characteristics in the current game. The situation status characteristics refer to the key information of the current game that can be quantitatively expressed in numerical form. In a free-playing chess and card game, the situation status characteristics include but are not limited to:

[0041] The current points of all players;

[0042] The historical number of cards retained by all players, that is, the number of cards a player retains in each round of the game;

[0043] The current game status of all players, such as bankruptcy;

[0044] Furthermore, the executable actions of the current player are extracted. The executable actions of the current player are an exhaustive enumeration of all the executable actions of the player under the visible information of the game, for example, Figure 3 In the game, the current player can play any single card, three strips (6669, 66655, 6669553, etc.), consecutive pairs (5566), bombs (3333), etc.

[0045] In a feasible implementation of the present application, the above-mentioned game visible information can be directly obtained by the decision server from the business logic server according to instructions.

[0046] Step 204: After fully connected layer processing is performed on the card features and situation status features respectively, they are concatenated to obtain an imperfect feature information vector; and fully connected layer processing is performed on the executable action to obtain an action feature vector.

[0047] In this step, the card features obtained in step 202 are first processed by multiple fully connected layers (FC layers) to convert the input card features into higher-level feature representations. These feature representations can better capture the essence of the current game state and provide support for subsequent decision-making.

[0048] Specifically, the current player's hand, other players' exposed cards, and each player's historical cards are encoded into multiple one-dimensional matrices and then merged to obtain a one-dimensional matrix of length 1*N. Then, the one-dimensional matrix of length 1*N is input into multiple fully connected layers for processing to obtain the first one-dimensional vector. Through the processing of the fully connected layer, the original card features are mapped to a new feature space. Therefore, the first one-dimensional vector contains a higher-level feature representation after nonlinear transformation and feature combination, and is associated with the subsequent decision results.

[0049] Furthermore, multiple fully connected layer processing is also performed on the situation state feature obtained in step 202. Specifically, the situation state feature is encoded into a one-dimensional matrix with a length of 1*M, and the one-dimensional matrix is ​​input into multiple fully connected layers for processing to obtain a second one-dimensional vector with a length of M.

[0050] Then, the first one-dimensional vector and the second one-dimensional vector are concatenated to obtain an imperfect feature information vector.

[0051] At the same time, in step 204, the executable actions of the current player are further processed by a fully connected layer to obtain action features.

[0052] Specifically, for example, the optional action information can be encoded into an S*15 matrix, where S represents the number of optional actions, which means that each optional action is represented by a vector of length 15. By processing this matrix through multiple fully connected layers, each initial multiple action features are gradually transformed, and finally a new action feature vector is obtained. This process changes the original feature space, which is more conducive to the decision-making of the subsequent model.

[0053] Step 206: Input the action feature vector and the imperfect feature information vector into the attention module for inference to obtain the main card decision result.

[0054] In this step, the action feature vector and the imperfect feature information vector are input into the attention module for reasoning to obtain the main card decision result. The attention module is a model pre-trained through a neural network architecture with an attention mechanism. The model includes the association between the visible information of the game and different executable actions. The model can be used to make decisions based on the input visible information of the game and different executable actions.

[0055] The core concept of the attention mechanism is to dynamically assign weights so that the model can focus on different parts of the input more effectively, rather than relying on a fixed context vector. Therefore, the parameters in the above attention module, such as the linear transformation matrix of Query, Key, Value, are obtained through pre-training. Those skilled in the art should know that the training process of the decision model is similar to the actual prediction process.

[0056] In a feasible implementation, the training process of the main card model (attention module) used for intelligent decision-making in a free-play card game includes:

[0057] Collect game data from high-level players;

[0058] The model is trained through supervised learning to enable it to imitate the player's card-playing strategy. During the training process, an attention mechanism is used to take the card features and situation state features as input, focusing on the most important features;

[0059] Among them, in a free-playing card game, players can select any number of hands of cards in the game interface to generate playing cards. The playing cards composed of any number of hands include a main card type. For example, the main card type of the user's playing data 444K83 is 444 (3 identical single cards), and the main card type of 8889993Q is 888999 (3 consecutive identical single cards); and the main card type of 22Q3 is 22 (pair), and the main card type of AAKKQQ9 is AAKKQQ (consecutive pairs).

[0060] Furthermore, in the learning process, in order to enable the main card model to focus on the most important main cards during training, the main card features of the user's card-playing data are extracted in the data preprocessing stage, and these features are input into the model. These features are used in the attention mechanism to determine the key actions.

[0061] Iterate and optimize the model repeatedly so that it can accurately imitate the main card playing strategy of high-level players, optimize the loss function, and freeze the parameters of the current attention module when the main card model performs well and converges on the validation set.

[0062] In step 206, the action feature vector and the imperfect feature information vector are input into the attention module for forward reasoning, which specifically includes:

[0063] The action feature vector is used as the parameter Key and Value in the attention module, and the imperfect feature information vector is used as the Query in the attention module. The attention module is forward-reasoned to obtain the attention score. After forward attention reasoning on the action feature vector and the imperfect feature information vector, the attention module can focus on the most relevant actions through the attention mechanism to output a decision. Specifically, the main card model calculates the similarity between the imperfect feature and each action feature (i.e., the attention score), and selects a reasonable main card based on the attention score.

[0064] When choosing a reasonable main card based on the attention score, you can choose different "attention score-main card" decision results according to different needs. For example, in the best decision recommendation, you can choose the main card corresponding to the maximum attention score as the main card decision result; you can also choose the main card corresponding to other attention scores according to the level of the current decision recommendation. I will not go into details here.

[0065] Step 208: Concatenate the main card decision result with the imperfect feature information vector; then use the Transformer model to decode the concatenated result to obtain a card retention action sequence.

[0066] In this step, according to the main card decision result, the main card information is processed by multiple layers of fully connected layers to obtain the main card features, and the main card features are concatenated with the imperfect feature information vector and input into the decoder of the Transformer model for decoding.

[0067] In an embodiment of the present application, a Transformer model is used to make decisions on keeping cards. Similarly, the Transformer model is a model for keeping card decisions that is pre-trained through a neural network architecture with an attention mechanism. After the main card model in step 206 converges and freezes the parameters of the current attention module, the model network parameters of the keeping card strategy continue to be trained, and a part or all of the input features are passed to the Transformer model of the neural network architecture. The Transformer model processes the input features through the encoder and decoder layers to generate a sequence of keeping cards. After repeated iterations and optimization of the model, it can accurately imitate the keeping card strategy of high-level players after the main cards are determined; similarly, when the Transformer model performs well on the validation set and converges, the parameters of the current Transformer model are frozen for the decision-making process of keeping cards.

[0068] Therefore, in the current decision-making process, the decoder of the Transformer model decodes the concatenation result of the main card feature vector and the imperfect feature information vector, and obtains the decision to keep the card based on the current game information.

[0069] Step 210: Remove the cards in the retaining action sequence from the current hand and obtain a decision result.

[0070] By comparing the current hand cards with the sequence of card-keeping actions one by one, the final decision result can be obtained, and this result will become the player's current card-playing strategy.

[0071] In a specific decision-making process embodiment, Figure 3 As shown in the figure, the current player's hand is 9666553333 and other visible information of the game. After obtaining the above visible information of the game, the decision server extracts the executable operations of the current player and sends it to the attention module for forward reasoning to obtain the attention score; then selects a reasonable corresponding main card from the attention score, such as 666; processes the selected main card information into a vector, concatenates it with the imperfect feature and inputs it into the Transformer decoder for the decision to keep the card, and obtains the card action sequence 3333, that is, the card left in this round is 3333; further, based on the current hand 9666553333 and the card action sequence 3333, it can be concluded that the current card strategy result should be 666559.

[0072] The decision server returns the decision result to the client, and the client instructs the target player to perform corresponding actions based on the decision result according to relevant instructions.

[0073] In the above embodiments of the present application, the scheme implemented based on the attention mechanism and the Transformer neural network structure has extremely high flexibility and adaptability, and provides a flexible, efficient and human thinking process-compliant decision-making method, which is particularly suitable for the decision-making process of free-play card games. The method obtains the visible information of the game from the perspective of the current player and all the actions that the current player can perform, and then obtains the main card decision result through the pre-trained attention module inference based on the above information, and then inputs the main card decision result and the visible information of the game into the pre-trained Transformer module for decoding to obtain the card retention decision result, and then obtains the card selection under the current situation. Through the method provided by the present application, more accurate and effective decisions can be made on the situation under the complex game rules of free play; at the same time, the design of the state space and the combination of different types of models can fully consider various variables in the game, so as to take into account comprehensive information in the decision-making process, and also make the decisions of the models at each stage more interpretable, further improving the rationality of the decision-making model.

[0074] Corresponding to the above-mentioned method embodiment for making intelligent decisions in a card game with free play, the present application also provides an embodiment of a system for generating intelligent decisions in a card game game, such as Figure 4 As shown, the system includes:

[0075] An extraction module, used to continuously obtain the game visible information and executable actions of the current player in the game, wherein the game visible information includes card features and situation status features;

[0076] The encoding module is used to perform full connection layer processing on the card features and situation state features respectively, and then concatenate them to obtain an imperfect feature information vector; and perform full connection layer processing on the executable action to obtain an action feature vector;

[0077] A main card decision module, used for inputting the action feature vector and the imperfect feature information vector into an attention module for inference to obtain a main card decision result;

[0078] A card retention decision module is used to concatenate the main card decision result with the imperfect feature information vector, and then use the Transformer model to decode the concatenation result to obtain a card retention action sequence;

[0079] The card-playing decision module is used to remove the cards in the card-retaining action sequence from the current hand and obtain a decision result.

[0080] The above is a schematic scheme of a system for making intelligent decisions in a card game with free cards of this embodiment. It should be noted that the technical scheme of the system for making intelligent decisions in a card game with free cards and the technical scheme of the method for making intelligent decisions in a card game with free cards belong to the same concept, and the details not described in detail in the technical scheme of the system for making intelligent decisions in a card game with free cards can all be referred to the description of the technical scheme of the method for making intelligent decisions in a card game with free cards.

[0081] In one embodiment of the present application, a computing device is also provided, including a memory, a processor, and computer instructions stored in the memory and executable on the processor, wherein when the processor executes the instructions, the steps of the above-mentioned method for making intelligent decisions in a free-play card game are implemented.

[0082] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the method for making intelligent decisions in the free-playing card game belong to the same concept, and the details not described in detail in the technical scheme of the computing device can be referred to the description of the technical scheme of the method for making intelligent decisions in the free-playing card game.

[0083] An embodiment of the present application also provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the steps of the method for making intelligent decisions in a free-play card game as described above.

[0084] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the method for making intelligent decisions in the free-playing card game mentioned above belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the technical scheme of the method for making intelligent decisions in the free-playing card game mentioned above.

[0085] The above describes specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0086] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0087] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0088] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0089] The preferred embodiments of the present application disclosed above are only used to help explain the present application. The optional embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the present application. The present application selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can understand and use the present application well. The present application is only limited by the claims and their full scope and equivalents.

Claims

1. A method for making intelligent decisions in a free-play card game, characterized in that: include: Continuously obtain the game visible information and executable actions of the current player in the game, wherein the game visible information includes card characteristics and situation status characteristics; After the card features and situation status features are processed by the fully connected layer respectively, the imperfect information feature vector is obtained by splicing; the executable action is processed by the fully connected layer to obtain the action feature vector; Inputting the action feature vector and the imperfect feature information vector into the attention module for inference to obtain a main card decision result; The main card decision result is concatenated with the imperfect feature information vector, and the concatenated result is decoded using a Transformer model to obtain a card retention action sequence; Remove the cards in the retaining card action sequence from the current hand to obtain a decision result.

2. The method according to claim 1, wherein: The situation status feature is the key information of the current game match expressed quantitatively in numerical form; the executable actions include all actions that the current player can perform based on the current visible game information and game rules.

3. The method according to claim 1, wherein: Inputting the action feature vector and the imperfect feature information vector into the attention module for inference to obtain the main card decision result includes: The action feature vector and the imperfect feature vector are input into an attention module for forward reasoning to obtain an attention score, and a main card is selected as a decision result according to the attention score.

4. The method according to claim 3, wherein: Inputting the action feature vector and the imperfect feature vector into the attention module for forward reasoning to obtain the attention score includes: The action feature vector is used as the Key and Value in the attention module, and the imperfect feature information vector is used as the Query in the attention module to perform forward reasoning of the attention module.

5. The method according to claim 1, wherein: The step of combining the main card decision result with the imperfect feature information vector includes: The main card decision result is processed by multiple layers of fully connected layers to obtain the main card feature, and the main card feature is spliced ​​with the imperfect feature information vector.

6. The method according to claim 1, wherein: The attention module is a neural network architecture including an attention mechanism, which is used to make main card decisions based on the visible information and executable actions of the game.

7. The method according to claim 6, wherein: The Transformer model is used to make a card retention decision based on the output of the attention module.

8. The method according to claim 1, wherein: The free-play card game includes a game in which any combination of cards are selected on a game interface to play cards.

9. A system for making intelligent decisions in a free-play card game, characterized in that: include: An extraction module, used to continuously obtain the game visible information and executable actions of the current player in the game, wherein the game visible information includes card features and situation status features; The encoding module is used to perform full connection layer processing on the card features and situation state features respectively, and then concatenate them to obtain an imperfect feature information vector; and perform full connection layer processing on the executable action to obtain an action feature vector; A main card decision module, used for inputting the action feature vector and the imperfect feature information vector into an attention module for inference to obtain a main card decision result; A card retention decision module is used to concatenate the main card decision result with the imperfect feature information vector, and then use the Transformer model to decode the concatenation result to obtain a card retention action sequence; The card-playing decision module is used to remove the cards in the card-retaining action sequence from the current hand and obtain a decision result.

10. A computing device comprising a memory, a processor, and computer instructions stored in the memory and executable on the processor, characterized in that: When the processor executes the instructions, the steps of the method according to any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium storing computer instructions, characterized in that: When the instruction is executed by a processor, the steps of the method described in any one of claims 1 to 8 are implemented.

Images