Method and system for generating intelligent decision in mahjong game

Abstract

The invention provides a method and system for generating an intelligent decision in a mahjong game, equipment and a storage medium. According to the method, game playing visible information under the view angle of a target player is obtained in game playing, and all actions capable of being executed by the current player are obtained based on the current game playing visible information; and extracting and reasoning the features through a neural network based on imitation learning to obtain a reasonable target decision result. According to the method, card playing and card doing actions can be performed according to habits and thinking modes of human beings, and good anthropomorphism is achieved; meanwhile, the decision-making method is simple and efficient in calculation process and has good performance in a high-intensity concurrent environment.

Description

Technical Field

[0001] The present application relates to the field of information interaction technology, and in particular to a method and system for generating intelligent decisions in a mahjong 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 mahjong games, in order to solve problems such as players accidentally disconnecting and in the intelligent prompt function, using AI to temporarily replace players in decision-making has become a common solution. However, although the decision-making solutions in the prior art have high tactical capabilities, their action strategies are often too fixed. For example, when facing players of different skill levels or different game situations, the decision-making process always uses the same action strategy, resulting in insufficient anthropomorphism, which affects the player's user experience. Therefore, there is an urgent need for an intelligent decision-making solution that can respond flexibly to different scenarios. Summary of the invention

[0003] In view of this, the embodiments of the present application provide a method and system for generating intelligent decisions in a mahjong 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, a method for generating intelligent decisions in a mahjong game is provided, the method comprising:

[0005] Obtaining visible game information of the current player in the game, wherein the visible game information includes card features and situation status features;

[0006] After extracting the executable actions of the current player according to the visible information of the game, the actions are encoded into a first one-dimensional vector;

[0007] Performing feature inference on the card features to obtain a second one-dimensional vector;

[0008] Processing the situation state feature through multiple fully connected layers to obtain a third one-dimensional vector, and concatenating the second one-dimensional vector and the third one-dimensional vector to obtain a fourth one-dimensional vector;

[0009] The fourth one-dimensional vector is feature-converted and then multiplied with the first one-dimensional vector to obtain a decision result of the current situation.

[0010] According to a second aspect of an embodiment of the present application, a system for generating intelligent decisions in a mahjong game is provided, comprising:

[0011] An acquisition module, used to acquire visible game information of the current player in the game, wherein the visible game information includes card features and situation status features;

[0012] An extraction module, configured to extract executable actions of the current player according to the game visible information and encode the extracted actions into a first one-dimensional vector;

[0013] A first encoding module, used for performing feature inference on the card feature to obtain a second one-dimensional vector;

[0014] A second encoding module is used to process the situation state feature through multiple fully connected layers to obtain a third one-dimensional vector, and concatenate the second one-dimensional vector and the third one-dimensional vector to obtain a fourth one-dimensional vector;

[0015] A decision module is used to perform feature conversion on the fourth one-dimensional vector and then multiply the result with the first one-dimensional vector to obtain a decision result of the current situation.

[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 generating intelligent decisions in a mahjong 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 generating intelligent decisions in a mahjong game are implemented.

[0018] The method provided in the embodiment of the present application obtains the visible information of the game from the perspective of the target player, and obtains all the actions that the current player can perform based on the current visible information of the game; then, the card features and situation state features are extracted and inferred respectively through the multi-layer ResBlock and fully connected layers based on imitation learning, and a reasonable target decision result is obtained. The method can play cards and make card actions according to human habits and thinking methods, and has good anthropomorphism; at the same time, the calculation process of the above decision-making method is simple and efficient, and has good performance in a high-intensity concurrent environment. 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 generating intelligent decisions in a mahjong game provided by an embodiment of the present application;

[0021] Figure 3 It is a schematic diagram of a game of a river of blood mahjong provided in an embodiment of the present application;

[0022] Figure 4It is a structural diagram of a system for generating intelligent decisions in a mahjong 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 generating intelligent decisions in a mahjong 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 the prior art, in the intelligent decision-making process of mahjong games, the rule-based behavior tree method is usually used. However, the AI ​​strength of the behavior tree is not enough and lacks flexibility. When it is necessary to make decisions based on the situation on the field, such as the cards played by other players, there is no flexibility and reasoning ability, which makes players feel boring; and the AI ​​of the behavior tree is not human enough, and many decisions are very different from those of human players, resulting in poor user experience. Even if a model that has undergone reinforcement learning is added, the anthropomorphism of the model is still not good enough, and counterintuitive operations often occur; and the complexity of the reinforcement learning model is usually high, not only the training time is long and the training quality is difficult to guarantee, but the decision-making process in the application scenario is relatively slow, and it is difficult to cope with high-load concurrency.

[0032] For example, in the classic Mahjong game Blood Flows into a River, there are four players using 108 cards, including three suits of strips, tubes, and ten thousand; players cannot combine the card type by eating cards, but can only combine the card type by touching and bar; when winning, the cards in the player's hand must be one or two suits, and after winning, the player can continue to draw cards and win until all the cards are touched; players need to change cards in the initial stage, and they need to check the call and the flower pig after the end. These rules greatly increase the complexity of the game, and there is currently no good artificial intelligence decision-making model. Therefore, the embodiment of the present application proposes a method for generating intelligent decisions in the Mahjong game, which ensures high decision-making efficiency in the Blood Flows into a River game while simulating the real human decision-making process.

[0033] The processor 120 may execute Figure 2 The steps of a method for generating intelligent decisions in a mahjong game are shown, and the method includes steps 202 to 210.

[0034] Step 202: continuously obtain the game visible information of the current player in the game, wherein the game visible information includes card features and situation status features.

[0035] In this step, when 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 instructions, such as Figure 3 shown.

[0036] The visible information of the game includes the characteristics of the cards in the current game. Specifically, in the Blood Flows into a River Mahjong game, the card characteristics include:

[0037] The player's current hand information is used to represent the cards that the player can currently operate;

[0038] The player's remaining cards are used to represent all the cards that the player cannot see at the moment, such as the hands of the other three players and the cards in the wall;

[0039] All players' touch cards;

[0040] All players' kongs;

[0041] The N most recent cards played by all players, such as in the game of River of Blood, each player has no more than 27 cards, N≤27;

[0042] The cards that players exchange and receive during the three-card exchange process;

[0043] The cards played by the other three players in the previous round;

[0044] The player's current draw;

[0045] All players win;

[0046] 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 the game of Blood Flows into a River, the situation status characteristics include:

[0047] Banker position;

[0048] All players’ vacancy information;

[0049] Whether all players have won the hand;

[0050] Which player's turn it is currently?

[0051] Whether other players except this player admit defeat;

[0052] The remaining virtual scores of all players;

[0053] And the number of cards remaining in the wall.

[0054] 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.

[0055] Step 204: extracting the executable actions of the current player according to the game visible information and encoding them into a first one-dimensional vector.

[0056] In the embodiment of the present application, all possible action types in the Blood River Mahjong are predefined according to the rules of the game, such as: playing cards, touching cards, barring cards, winning cards and passing cards.

[0057] According to the current game visible information obtained in step 202, all executable actions of the player are obtained, such as whether the card played by other players last time can be touched, connected or win; after drawing the cards, which cards in the current hand can be played or whether the card can be won, etc. All executable actions can be exhaustively enumerated according to the current game visible information and game rules.

[0058] Furthermore, the executable actions are encoded into one-dimensional vectors. For example, which cards in the current hand can be played are represented by three one-dimensional vectors:

[0059]

[0060] Table 1

[0061]

[0062] Table 2

[0063]

[0064] Table 3

[0065] Therefore, a one-dimensional vector with a length of 3*9=27 can be used to represent the executable actions of playing a card.

[0066] Similarly, for pong and kong, a one-dimensional vector of length 27 can be used to represent the executable actions; and for hu and gua, a one-dimensional vector of length 1 can be used respectively.

[0067] Further, the one-dimensional vectors corresponding to all executable actions are merged to obtain a first one-dimensional vector. For example, in the embodiment of the blood-flowing river mahjong, all the one-dimensional vectors are merged to obtain a first one-dimensional vector with a length of 1*83, where:

[0068] Positions 1-27 represent the card selection, which represent 1-9 Wan, 1-9 Tiao, and 1-9 Tong in order;

[0069] Positions 28-54 represent the card selection, which represent 1-9 Wan, 1-9 Tiao, and 1-9 Tong in order;

[0070] Positions 55-81 represent the selection of Gang cards, which in order represent Gang 1-9 Wan, 1-9 Tiao, and 1-9 Tong respectively;

[0071] 82 represents Hupai;

[0072] 83 represents a check;

[0073] Step 206: performing feature inference on the card feature to obtain a second one-dimensional vector;

[0074] In this step, after analyzing and encoding the multiple card features obtained in step 202, an N-dimensional matrix is ​​obtained. The size of the matrix is ​​determined according to the specific information of the card features. For example, in the embodiment of the blood-flowing river mahjong, the encoding can be performed as follows:

[0075] The code corresponding to the player's current hand information is (3,4,9), that is, a 3*4*9 three-dimensional matrix, where 3 represents the three suits of mahjong tiles, 9 represents the number of mahjong tiles, and 4 represents the number of mahjong tiles. For example, if the player has 2 cards of 30,000, it can be represented as: [1][2][3]=1.

[0076] The encoding of the player's remaining cards is: (3,4,9). Similarly, this matrix can represent all the cards that the player currently cannot see.

[0077] The encoding of all players' card touches: (3,4,9), using a (3,4,9)-shaped matrix to encode the information of players' card touches;

[0078] The encoding of all players' cards: (3,4,9), using a matrix of shape (3,4,9) to encode the information of the players' cards;

[0079] Then, the codes of other card features include: the codes of the most recent N cards played by all players: (3, N1, 9), the codes of the cards played and received by the players in the process of changing three cards (3, N2, 9), the codes of the cards played by the other three players in the previous round: (3, N3, 9), the current cards of the player (3, N4, 9), the codes of all players' winning cards (3, N5, 9)..., and the obtained codes are merged to obtain a multidimensional matrix. In the Blood River Mahjong, this multidimensional matrix can be 3*N*9.

[0080] Furthermore, feature conversion is performed on the N-dimensional matrix to convert it into high-level features for subsequent policy prediction. Specifically, multi-layer ResBlock processing is performed on the N-dimensional matrix to extract high-level features.

[0081] ResBlock is a basic building block in the deep residual network (ResNet), which is used to solve the gradient vanishing problem in the use of deep neural networks. ResBlock introduces skip connections, which enables the network to directly pass input information to output, thereby maintaining the effective propagation of gradients in deep networks and preventing gradients from vanishing or exploding.

[0082] In the embodiment of the present application, several ResBlocks in the multi-layer ResBlock based on imitation learning have different divisions of labor to achieve the extraction of different features. The multi-layer deep structure can extract more abstract and complex feature representations layer by layer, enhancing the feature extraction and expression capabilities of the model. In addition, the structure of ResBlock is relatively simple, with moderate computational overhead, which is suitable for the real-time reasoning requirements of the Blood Flowing Mahjong game.

[0083] Furthermore, the processing results of the multi-layer ResBlock are expanded and flattened into a second one-dimensional vector with a length of 1*M.

[0084] Step 208: The situation state feature is converted through multiple fully connected layers to obtain a third one-dimensional vector, and the second one-dimensional vector and the third one-dimensional vector are concatenated to obtain a fourth one-dimensional vector.

[0085] In this step, the situation status features obtained in step 202 are processed by multiple fully connected layers (FC layers) to convert the input situation status 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.

[0086] Furthermore, considering the differences between different features, the distinguishing feature processing of the card features and the situation state features in step 206 and step 208 is more in line with the characteristics of each feature, so as to better extract their respective information. The above-mentioned card features are high-dimensional, highly correlated, and require complex feature extraction; while the situation state features are low-dimensional and relatively independent, and are more suitable for the fully connected layer.

[0087] Specifically, the situation status features are encoded into a one-dimensional vector. For example, a one-dimensional matrix of length B is created, the dealer position is expressed with 1 bit, the player's vacancy information is expressed with 4 bits, and so on, and all the situation status features are encoded into the one-dimensional matrix.

[0088] Furthermore, the one-dimensional matrix with a length of 1*B is input into multiple fully connected layers for processing to obtain a third one-dimensional vector with a length of 1*T. Through the processing of the fully connected layer, the original situation state features are mapped to a new feature space. Therefore, the third one-dimensional vector contains a higher-level feature representation after nonlinear transformation and feature combination, and is associated with subsequent decision results.

[0089] Furthermore, the second one-dimensional vector of length M is concatenated with the third one-dimensional vector of length T to obtain a fourth one-dimensional vector of length 1*(M+T), thereby obtaining the card type information and situation information that need to be considered simultaneously for decision-making.

[0090] In this step, the card features and scalar features are processed separately and then spliced ​​together, so that the card features and situation status features can be given equal importance in the decision-making process, so that the situation status features with smaller dimensions and the card features are processed as high-level features of equal status and will not be easily ignored by the decision-making model.

[0091] Step 210: Perform feature conversion on the fourth one-dimensional vector and multiply it with the first one-dimensional vector to obtain a decision result of the current situation.

[0092] In this step, the process of human players inducing judgments from specific information is further simulated, and the fourth one-dimensional vector is feature transformed in this step. Specifically, the fourth one-dimensional vector is input again into multiple fully connected layers based on imitation learning to simulate the reasoning process from observation to decision, and the fifth one-dimensional vector is obtained.

[0093] The length of the fifth one-dimensional vector obtained is equal to the length of the first one-dimensional vector. For example, in the game of Blood Flows into a River, the length of the fifth one-dimensional vector obtained after being processed by multiple fully connected layers is 83, which is equal to the length of the first one-dimensional vector in step 204.

[0094] Furthermore, the fifth one-dimensional vector is multiplied by the first one-dimensional vector to obtain a result vector, and the position corresponding to the maximum value in the result vector is the decision result of the current situation. The decision result is returned to the client, instructing the target player to perform corresponding actions according to the decision result.

[0095] In the above embodiments of the present application, a highly efficient decision-making method that conforms to the human thinking process is provided, which is particularly suitable for the decision-making process of games such as Mahjong. In this method, not only the visible information of the game from the perspective of the current player is obtained, but also all the actions that the current player can perform are obtained based on the current visible information of the game; and then the card features and situation state features are extracted and inferred respectively through the multi-layer ResBlock and fully connected layers based on imitation learning, so that the decision-making process is more efficient, more focused on the characteristics of each feature to obtain a reasonable target decision result, and has better anthropomorphism. At the same time, the calculation process of the above decision-making method is simple and efficient, and has better performance in a high-intensity concurrent environment.

[0096] Corresponding to the above-mentioned method embodiment for generating intelligent decision in a mahjong game, the present application also provides an embodiment of a system for generating intelligent decision in a mahjong game, such as Figure 4 As shown, the system includes:

[0097] An acquisition module, used to acquire visible game information of the current player in the game, wherein the visible game information includes card features and situation status features;

[0098] An extraction module, configured to extract executable actions of the current player according to the game visible information and encode the extracted actions into a first one-dimensional vector;

[0099] A first encoding module, used for encoding the card features into a multi-dimensional matrix, performing feature inference on the multi-dimensional matrix and flattening it into a second one-dimensional vector;

[0100] A second encoding module is used to process the situation state feature through multiple fully connected layers to obtain a third one-dimensional vector, and concatenate the second one-dimensional vector and the third one-dimensional vector to obtain a fourth one-dimensional vector;

[0101] A decision module is used to perform feature conversion on the fourth one-dimensional vector and then multiply the result with the first one-dimensional vector to obtain a decision result of the current situation.

[0102] The above is a schematic scheme of a system for generating intelligent decisions in a mahjong game match of this embodiment. It should be noted that the technical scheme of the system for generating intelligent decisions in a mahjong game match and the technical scheme of the method for generating intelligent decisions in a mahjong game match are of the same concept, and the details not described in detail in the technical scheme of the system for generating intelligent decisions in a mahjong game match can be found in the description of the technical scheme of the method for generating intelligent decisions in a mahjong game match.

[0103] 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 method for generating intelligent decisions in a mahjong game are implemented.

[0104] 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 generating intelligent decisions in the above-mentioned mahjong game 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 generating intelligent decisions in the above-mentioned mahjong game game.

[0105] 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 generating intelligent decisions in a mahjong game as described above.

[0106] 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 generating intelligent decisions in the above-mentioned mahjong game game 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 generating intelligent decisions in the above-mentioned mahjong game game.

[0107] 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.

[0108] 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.

[0109] 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.

[0110] 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.

[0111] 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 generating intelligent decisions in a mahjong game, characterized in that: include: Continuously obtain the game visible information of the current player in the game, wherein the game visible information includes card characteristics and situation status characteristics; After extracting the executable actions of the current player according to the visible information of the game, the actions are encoded into a first one-dimensional vector; Performing feature inference on the card features to obtain a second one-dimensional vector; Processing the situation state feature through multiple fully connected layers to obtain a third one-dimensional vector, and concatenating the second one-dimensional vector and the third one-dimensional vector to obtain a fourth one-dimensional vector; The fourth one-dimensional vector is feature-converted and then multiplied with the first one-dimensional vector to obtain a decision result of the current situation.

2. The method according to claim 1, wherein: The situation status feature is the key information of the current game game expressed quantitatively in numerical form.

3. The method according to claim 1, wherein: The executable actions of the current player include all actions that the current player can perform based on the current visible game information and game rules.

4. The method according to claim 1, wherein: Inferring the card features to obtain a second one-dimensional vector includes: All card features are encoded separately and merged into a multi-dimensional matrix, and the multi-dimensional matrix is ​​input into a multi-layer ResBlock based on imitation learning to extract high-level features, and then the processing results of the multi-layer ResBlock are expanded to obtain a second one-dimensional vector.

5. The method according to claim 4, wherein: The step of processing the situation state feature through a plurality of fully connected layers to obtain a third one-dimensional vector comprises: All situation state features are encoded separately and merged to obtain a one-dimensional matrix, which is then input into multiple fully connected layers based on imitation learning for processing to obtain a third one-dimensional vector.

6. The method according to claim 1, wherein: The step of performing feature conversion on the fourth one-dimensional vector and then multiplying the fourth one-dimensional vector by the first one-dimensional vector to obtain a decision result of the current situation includes: The fourth one-dimensional vector is input into multiple fully connected layers based on imitation learning to obtain a fifth one-dimensional vector, and the fifth one-dimensional vector is multiplied by the first one-dimensional vector to obtain a result vector, and the position corresponding to the maximum value in the result vector is the decision result of the current situation; wherein, the length of the fifth one-dimensional vector is equal to the length of the first one-dimensional vector.

7. The method according to claim 1, wherein: The mahjong games include river of blood mahjong and bloody battle to the end mahjong.

8. A system for generating intelligent decisions in a mahjong game, characterized in that: include: An acquisition module, used to acquire game visible information of the current player in the game, wherein the game visible information includes card features and situation status features; An extraction module, configured to extract executable actions of the current player according to the game visible information and encode the extracted actions into a first one-dimensional vector; A first encoding module, used for performing feature inference on the card feature to obtain a second one-dimensional vector; A second encoding module is used to process the situation state feature through multiple fully connected layers to obtain a third one-dimensional vector, and concatenate the second one-dimensional vector and the third one-dimensional vector to obtain a fourth one-dimensional vector; A decision module is used to perform feature conversion on the fourth one-dimensional vector and then multiply the result with the first one-dimensional vector to obtain a decision result of the current situation.

9. 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 7 are implemented.

10. 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 7 are implemented.

Images