CN108671546A

CN108671546A - Determination method and apparatus, storage medium and the electronic device of object run

Info

Publication number: CN108671546A
Application number: CN201810502383.0A
Authority: CN
Inventors: 申俊峰; 周大军; 张力柯; 荆彦青
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2018-05-23
Filing date: 2018-05-23
Publication date: 2018-10-19

Abstract

The invention discloses determination method and apparatus, storage medium and the electronic devices of a kind of object run.Wherein, this method includes：Detect the operation requests of the first game role, operation requests are for ask in client is played for currently running one innings the first game role based on the pending object run of the first object got；In response to operation requests, target feature vector is obtained, target feature vector is at least used to indicate to be formed to obtain the probability of target object set based on the first object got；The corresponding object run of target feature vector is obtained according to trained neural network model, trained neural network model is used to indicate the mapping relations between target feature vector and object run.The present invention solves the relevant technologies and calculates operation in next step according to game player's experience, the relatively low technical problem of rate of winning of playing after causing game player to execute next operation.

Description

Determination method and apparatus, storage medium and the electronic device of object run

Technical field

The present invention relates to computer realms, are situated between in particular to a kind of determination method and apparatus of object run, storage Matter and electronic device.

Background technology

Currently, cards game client may be implemented to calculate game player's operation, such as moral in next step in the related technology It is to abandon board, or fill that state canaster client, which can calculate game player,；Random Factor Mahjong client can calculate which is got Open hands etc..But the relevant technologies are to calculate the operation of its next step according to game player's experience value, will be caused in this way This innings game is won in game player's execution rate of winning after operating in next step is relatively low.

For above-mentioned problem, currently no effective solution has been proposed.

Invention content

An embodiment of the present invention provides determination method and apparatus, storage medium and the electronic device of a kind of object run, with It at least solves the relevant technologies and operation in next step is calculated according to game player's experience, game player is caused to execute next operation Game is won the relatively low technical problem of rate afterwards.

One side according to the ... of the embodiment of the present invention, provides a kind of determination method of object run, and client is currently transported The first game role is based on the pending object run of the first object got described in one innings of capable game；In response to The operation requests obtain target feature vector, wherein the target feature vector is at least used to indicate have been obtained based on described To the first object form to obtain the probability of target object set, the target object set includes first got Object；The corresponding object run of the target feature vector is obtained according to trained neural network model, wherein described Trained neural network model is using the mapping relations between sampling feature vectors and sample operations to first nerves network The model that model is trained, when being trained to first nerves network model, the first nerves network model Input parameter is the sampling feature vectors, and the output parameter of the first nerves network model is the sample operations, described Trained neural network model is used to indicate the mapping relations between the target feature vector and the object run.

Another aspect according to the ... of the embodiment of the present invention additionally provides a kind of determining device of object run, including：Detection is single Member, the operation requests for detecting the first game role, wherein the operation requests are currently run for asking in client One innings game described in the first game role based on the pending object run of the first object got；First obtains Unit, in response to the operation requests, obtaining target feature vector, wherein the target feature vector is at least used to refer to Show and formed to obtain the probability of target object set based on first object got, the target object set includes institute State the first object got；Second acquisition unit, it is special for obtaining the target according to trained neural network model The corresponding object run of sign vector, wherein the trained neural network model is to use sampling feature vectors and sample The model that mapping relations between this operation are trained first nerves network model, to first nerves network model When being trained, the input parameter of the first nerves network model is the sampling feature vectors, the first nerves network The output parameter of model is the sample operations, and the trained neural network model is used to indicate the target feature vector With the mapping relations between the object run.

Another aspect according to the ... of the embodiment of the present invention additionally provides a kind of storage medium, is stored in the storage medium Computer program, wherein the computer program is arranged to execute any one target behaviour in the embodiment of the present invention when operation The determination method of work.

Another aspect according to the ... of the embodiment of the present invention additionally provides a kind of electronic device, including memory and processor, In, computer program is stored in the memory, the processor is arranged to execute this hair by the computer program The determination method of any one object run in bright embodiment.

In embodiments of the present invention, by after the operation requests for detecting the first game role, according to the first game angle The first object acquisition target feature vector that color has been got, wherein target feature vector, which is at least used to indicate, to be based on having obtained To the first object formed to obtain the probability of target object set, it is special then to obtain target using trained neural network model The corresponding object run of sign vector has achieved the purpose that rapidly and accurately to determine the pending object run of game role, in turn It solves the relevant technologies and operation in next step is calculated according to game player's experience, after causing game player to execute next operation Game is won the relatively low technical problem of rate, to realize the technique effect for improving the rate of winning of game role in gaming.

Description of the drawings

Attached drawing described herein is used to provide further understanding of the present invention, and is constituted part of this application, this hair Bright illustrative embodiments and their description are not constituted improper limitations of the present invention for explaining the present invention.In the accompanying drawings：

Fig. 1 is the schematic diagram of the hardware environment of the determination method of object run according to the ... of the embodiment of the present invention；

Fig. 2 is a kind of flow chart of the determination method of optional object run according to the ... of the embodiment of the present invention；

Fig. 3 is the schematic diagram of characteristic vector pickup flow according to the ... of the embodiment of the present invention；

Fig. 4 is the schematic diagram of trained neural network model according to the ... of the embodiment of the present invention；

Fig. 5 is the schematic diagram of trained neural network model deployment according to the ... of the embodiment of the present invention in the client；

Fig. 6 is the schematic diagram of trained neural network model deployment according to the ... of the embodiment of the present invention in the server；

Fig. 7 is the schematic diagram of holdem interface according to the ... of the embodiment of the present invention；

Fig. 8 is a kind of schematic diagram of the determining device of optional object run according to the ... of the embodiment of the present invention；And

Fig. 9 is a kind of structure diagram of electronic device according to the ... of the embodiment of the present invention.

Specific implementation mode

In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only The embodiment of a part of the invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people The every other embodiment that member is obtained without making creative work should all belong to the model that the present invention protects It encloses.

It should be noted that term " first " in description and claims of this specification and above-mentioned attached drawing, " Two " etc. be for distinguishing similar object, without being used to describe specific sequence or precedence.It should be appreciated that using in this way Data can be interchanged in the appropriate case, so as to the embodiment of the present invention described herein can in addition to illustrating herein or Sequence other than those of description is implemented.In addition, term " comprising " and " having " and their any deformation, it is intended that cover It includes to be not necessarily limited to for example, containing the process of series of steps or unit, method, system, product or equipment to cover non-exclusive Those of clearly list step or unit, but may include not listing clearly or for these processes, method, product Or the other steps or unit that equipment is intrinsic.

One side according to the ... of the embodiment of the present invention provides a kind of determination method of object run.

Optionally, in the present embodiment, the determination method of above-mentioned object run can be applied to as shown in Figure 1 by servicing In the hardware environment that device 102 and terminal 104 are constituted.As shown in Figure 1, server 102 is connected by network and terminal 104 It connects, above-mentioned network includes but not limited to：Wide area network, Metropolitan Area Network (MAN) or LAN, terminal 104 are not limited to PC, mobile phone, tablet electricity Brain etc..The determination method of the object run of the embodiment of the present invention can be executed by terminal 104, can also be by server 102 It is executed jointly with terminal 104.Wherein, terminal 104 execute the embodiment of the present invention object run determination method can also be by Client mounted thereto executes.

Optionally, the determination side of the object run of terminal 104 or the client executing embodiment of the present invention in terminal 104 The process of method can be described as：Client detects the operation requests of the first game role, wherein operation requests are for asking The first game role is based on the pending target behaviour of the first object got in the currently running one innings of game of client Make；Client end response obtains target feature vector in operation requests, wherein target feature vector is at least used to indicate based on The first object got forms to obtain the probability of target object set, and target object set includes the first couple got As；Client obtains the corresponding object run of target feature vector according to trained neural network model, wherein trained Neural network model is carried out to first nerves network model using the mapping relations between sampling feature vectors and sample operations The model that training obtains, when being trained to first nerves network model, the input parameter of first nerves network model is sample The output parameter of eigen vector, first nerves network model is sample operations, and trained neural network model is used to indicate Mapping relations between target feature vector and object run.

Optionally, the mesh of terminal 104 or the common execution embodiment of the present invention of client and server 102 in terminal 104 Marking the process of the determination method of operation can be described as：Client detects the operation requests of the first game role, wherein operation Request for ask client it is currently running one innings play in the first game role based on the first object institute got Pending object run；Client end response obtains target feature vector, wherein target feature vector is at least in operation requests It is used to indicate and is formed to obtain the probability of target object set based on the first object got, target object set includes having obtained The first object got；Target feature vector is sent to server 102 by client；Server 102 is according to trained nerve Network model obtains the corresponding object run of target feature vector, wherein trained neural network model is special using sample The model that mapping relations between sign vector and sample operations are trained first nerves network model, to the first god When being trained through network model, the input parameter of first nerves network model is sampling feature vectors, first nerves network mould The output parameter of type is sample operations, and trained neural network model is used to indicate between target feature vector and object run Mapping relations；Object run is sent to client by server 102.

The determination method of the object run of the embodiment of the present invention is described in detail using client as executive agent below.

Fig. 2 is a kind of flow chart of the determination method of optional object run according to the ... of the embodiment of the present invention, such as Fig. 2 institutes Show, this method may comprise steps of：

Step S202 detects the operation requests of the first game role, wherein operation requests are worked as asking in client The first game role is based on the pending object run of the first object got in one innings of game of preceding operation；

Step S204 obtains target feature vector in response to operation requests, wherein target feature vector is at least used to refer to Show and formed to obtain the probability of target object set based on the first object got, target object set includes having got First object；

Step S206 obtains the corresponding object run of target feature vector according to trained neural network model, wherein Trained neural network model is using the mapping relations between sampling feature vectors and sample operations to first nerves network The model that model is trained, when being trained to first nerves network model, the input of first nerves network model Parameter is sampling feature vectors, and the output parameter of first nerves network model is sample operations, trained neural network model The mapping relations being used to indicate between target feature vector and object run.

S202 to step S206 through the above steps, by after the operation requests for detecting the first game role, according to The first object acquisition target feature vector that first game role has been got, wherein target feature vector is at least used to indicate It is formed to obtain the probability of target object set based on the first object got, then utilizes trained neural network model The corresponding object run of target feature vector is obtained, has reached and has rapidly and accurately determined the pending object run of game role Purpose, and then solve the relevant technologies and operation in next step is calculated according to game player's experience, under causing game player to execute Game is won the relatively low technical problem of rate after one operation, to realize the skill for improving the rate of winning of game role in gaming Art effect.

In the technical solution that step S202 is provided, the currently running one innings of game of client can be asymmetry board class Game, such as holdem, mahjong etc..Herein it should be noted that symmetrically sex play refers to all game players in same a period of time Between it can be seen that same scene, show proprietary action situation.Asymmetry game refers to that some game player knows The thing of generation, and other game players do not know this part thing.Asymmetric may include that picture is asymmetric and information asymmetry, In, information asymmetry may include that data information asymmetric (such as dealing equipment promotes fighting capacity) and game data are asymmetric (such as the position of subject player is seen by a kind of equipment of purchase).May include at least one game angle in playing at one innings Color, wherein at least one game role includes the first game role, and the first game role can be at least one game role In any one.Optionally, the first game role can be controlled by game player, can also be controlled by client.

Optionally, the first game role is based on during the operation requests of the first game role can be used for asking playing at one innings The pending object run of the first object for having got.Optionally, in playing at one innings, what the first game role had been got First object can be the hands of the first game role, and the first game role is pending based on the first object got Object run may include but be not limited to abandoning board, playing a card, fill, with the operations such as entering for crossruff execution.

Client can detect the operation requests of the first game role in real time, optionally, in playing at one innings, sequence of playing a card The first game role is taken turns to, can detect the operation requests of the first game role, such as takes turns to the first game in Random Factor Mahjong When role plays a card, the operation requests of the first game role can be triggered.Optionally, at one innings play in other game roles After having executed corresponding operating, it can detect that other in the operation requests of the first game role, such as holdem game are played After role stakes on, the operation requests of the first game role can be triggered.

In the technical solution that step S204 is provided, after the operation requests for detecting the first game role, client Current first game role can be obtained first has got the first object, wherein the number of the first object can be one, Or it is multiple.Then judge whether the first object got can form to obtain target object set again, if energy Formation obtains target object set, then calculates the first object got and formed to obtain the probability of target object set, wherein Target object set includes the first object.Herein it should be noted that may include at least one target object in one innings of game Set, the first object got can form the first pair of pictograph for obtaining at least one target object set, having got It may be the same or different at the probability for obtaining each target object set.Optionally, target object set can be at one innings It wins in game.For example, in holdem game, target object set can be different board types, may include straight flush, four Item, color, suitable son etc..For another example in Random Factor Mahjong, target object set can be win the game board type, may include ten three the one, seven To board etc..

Optionally, target feature vector can be at least used to indicate forms to obtain target based on the first object got The probability of object set.Optionally, target feature vector may include multiple dimensions, and a kind of target object set can correspond to one The corresponding dimension of target object set is known as the first dimension by a dimension herein, wherein the number of the first dimension can be one It is a, or multiple.The probability for forming to obtain each target object set based on the first object got can conduct Element in first dimension.That is, step S204 acquisition target feature vectors may include：It obtains and is based on having got The first object formed to obtain the probability of target object set；Using probability as the member of the first dimension in target feature vector Element.

Optionally, target feature vector forms to obtain target object in addition to being used to indicate based on the first object got The probability of set may be used to indicate that the second game role is transferred out of based on the second object got in this innings game Resource.Herein it should be noted that in playing at this innings, the second game role and the first game role can be to play chess to close System, the second object that the second game role has been got can be the hands of the second game role, and the second game role is based on The resource that the second object got is transferred out of can be the chip that the second game role is betted based on current hands.

Optionally, can also include at least one second game role in one innings of game other than the first game role, Each second game role can correspond to a dimension in target feature vector, herein by the corresponding dimension of the second game role Referred to as the second dimension, wherein the number of the second dimension can be one, or multiple.

Optionally, the element in target feature vector in the second dimension can serve to indicate that the second game role is based on having obtained The resource that the second object got is transferred out of.That is, step S204 acquisition target feature vectors can also include：Obtain one The resource that the second game role is at least transferred out of based on the second object got in office's game；Place is normalized in resource Reason, obtains normalized value；Using normalized value as the element of the second dimension in target feature vector.It should be noted that this Resource is normalized at place so that normalized value is the numerical value in 0 to 1, can be made in target feature vector in this way The first dimension it is corresponding with the element in the second dimension.

The first dimension in target feature vector and the element in the second dimension can be got by the above process, namely Target feature vector can be got, the target feature vector in the embodiment of the present invention can be the vector of a various dimensions, mesh The dimension and the number of the second game role in one innings of game and the kind number of target object set for marking feature vector determine.

In the technical solution that step S206 is provided, after getting target feature vector, it can utilize trained Neural network model obtains object run corresponding with target feature vector, to achieve the purpose that respond operation requests, wherein instruction The neural network model perfected can serve to indicate that the mapping relations between target feature vector and object run.

Optionally, trained neural network model can be using the mapping between sampling feature vectors and sample operations The model that relationship is trained first nerves network model, wherein first nerves network model can be without any Trained model, when being trained to first nerves network model, the input parameter of first nerves network model can be sample Eigen vector, the output parameter of first nerves network model can be sample operations.

Optionally, sampling feature vectors can at least be used to indicate the sample object got based on sample game role Formation obtains the probability of target object set.Optionally, sample game role can be the game angle in different offices in the game Color, in order to ensure the accuracy of trained neural network model, the number of sample game role can be many.It is optional Ground, the sample object that sample game role has been got can be the hands of sample game role in game.Optionally, sample is special Sign vector may be used to indicate that the resource that sample game role is transferred out of based on the sample object got.It needs to illustrate It is that the format of sampling feature vectors can be identical as above-mentioned target feature vector, and difference lies in pairs that game role has been got As and the resource that is transferred out of based on the object got of game role it is different.It should be noted sampling feature vectors Acquisition methods it is identical as above-mentioned target feature vector, referring specifically to above description, details are not described herein again.

Optionally, after sample game role executes sample operations based on the sample object got, at least meet following Any one goal condition：It is formed to obtain the probability highest of target object set based on the sample object got, be based on The sample object got formed to obtain target object be integrated into one innings of game win, resource that sample game role obtains most More, the loss of sample game role resource is minimum.

Optionally, first nerves network model is carried out using the mapping relations between sampling feature vectors and sample operations Training obtains trained neural network model and may include：

First nerves network model is instructed first with the mapping relations between first eigenvector and the first operation Practice, obtains nervus opticus network model.

It should be noted that sampling feature vectors may include first eigenvector, wherein first eigenvector can be used It is formed to obtain the probability of target object set based on the third object that third game role has been got in instruction.Optionally, Three game roles can be any one in sample game role, and the third object that third game role has been got can be The hands of third game role.Third game role can meet after executing the first operation based on the third object got State goal condition.When being trained to first nerves network model, the input parameter of first nerves network model can be the The output parameter of one feature vector, first nerves network model can be the first operation.

After obtaining nervus opticus network model, following steps are repeated for nervus opticus network model, until After 4th game role executes the second operation got using trained neural network based on the 4th object got Meet goal condition：First resource is compared with Secondary resource, wherein first resource is the 5th game role to having obtained The 5th object that arrives executes obtained resource after third operation, Secondary resource be the 6th game role to got the 6th Object executes obtained resource after the 4th operation, third operation for the second feature that is got using nervus opticus network model to Corresponding operation is measured, second feature vector is used to indicate to be formed to obtain target object set based on the 5th object got Probability, the 4th operation are the corresponding operation of third feature vector that is got using nervus opticus network model, third feature to Amount is used to indicate to be formed to obtain the probability of target object set based on the 6th object got；It is more than second in first resource In the case of resource, nervus opticus network model is instructed using the mapping relations between second feature vector and third operation Practice, obtains trained neural network model；Secondary resource be more than first resource in the case of, using third feature vector with Mapping relations between 4th operation are trained nervus opticus network model, obtain trained neural network model.

It should be noted that the embodiment of the present invention first with a part of data (namely first eigenvector and first behaviour Mapping relations between work) it is trained, obtain nervus opticus network model.In order to improve trained neural network model Accuracy, then by comparing excellent between the data got using nervus opticus network model, determine more excellent Then data using the more excellent data be trained nervus opticus network model, to obtain more accurately The trained neural network model.Herein it should be noted that more excellent data can be understood as playing at one innings In the probability won is larger or obtained resource is more or the resource of loss is less.

That is, the embodiment of the present invention can be extracted from the sample object that a large amount of sample game role has been got Then sampling feature vectors carry out deep learning training using the mapping relations of sampling feature vectors and sample operations, so that The neural network model that training obtains can more accurately determine the corresponding object run of target feature vector, and then reach and carry The effect of the rate of winning of high game role in gaming.

It is corresponding obtaining target feature vector according to trained neural network model as a kind of optional embodiment After object run, the embodiment of the present invention can also control the operation of the first game role performance objective in the client.For example, In holdem game, determine that the pending object run of the first game role is to add using trained neural network model After note operation, first game role can be controlled in the client and is filled.For another example in Random Factor Mahjong, instruction is utilized The neural network model perfected determines that the pending object run of the first game role is that " four " this one card is chosen to carry out It plays a card, controls first game role in the client and get " four " this one card.

It should be noted that the alternative embodiment is suitable for the scene that the first game role is controlled by client, Ye Jike Family end can directly control the operation of the first game role performance objective.

As an alternative embodiment, corresponded to obtaining target feature vector according to trained neural network model Object run after, the embodiment of the present invention can also be used to indicate the operation information of object run to client push, wherein The operation information of the object run pushed can serve to indicate that controls the operation of the first game role performance objective in the client. For example, in holdem game, the pending object run of the first game role is determined using trained neural network It, can be to the operation information of client push " filling " this kind of prompt, to indicate that game player can control after being operated for filling First game role is filled.For another example in Random Factor Mahjong, first is determined using trained neural network model The pending object run of game role be choose " four " this one card to play a card, then in the client by " four " this One card is moved to region of playing a card, and game player is prompted to get " four " this one card.

It should be noted that the alternative embodiment is suitable for the scene that the first game role is controlled by game player, namely Game player can control the operation of the first game role performance objective according to the operation information of the object run to client push.

It for the determination method of the object run of the embodiment of the present invention, can be applied in cards game, such as Dezhou is flutterred Gram, in the asymmetric information games such as mahjong.

As a kind of optional example, the embodiment of the present invention can be applied in holdem game.

The optional example proposes a kind of feature extracting method for asymmetric information game problem, can after extraction feature It is trained to obtain trained neural network model using deep learning, trained neural network model can be deployed in Man-machine game is carried out in product.

The optional example includes mainly three parts content, respectively：The Probability Characteristics extraction that board type is set up；Nerve net The training of network model；Trained neural network model is deployed in product client, realizes man-machine chess.

Holdem game flow is summarized as follows：Each participants in a bridge game sends out 2 cards in one's hand, then sends out 5 community cards successively again, each Participants in a bridge game selects 5 and is combined into maximum board group, victory or defeat comparison is carried out with other people from 2 in hand and 5 community cards.Board Type size rule is descending to be followed successively by：

The ＞ tri- of straight flush ＞ tetra- plus a pair of ＞ colors ＞ are along tri- ＞ two of sub- ＞ to the mono- boards of ＞ a pair of ＞

Holdem gives as security altogether four alternate water injections：Bridge queen, which is opened, per human hair 2 carries out the first alternate water injection；The second wheel is given as security after sending out 3 community cards Note；Third alternate water injection is given as security after sending out the 4th community card；Fourth round note is given as security after sending out the 5th community card.All residue players after staking on It carries out than board, the maximum is won.

Feature extracting method considers the numerical value of most important three aspects：

(1) itself distributional values：Mainly consider that the board power size of itself hands, the more big then characteristic value of board power are bigger.

(2) the board type probability distribution that community card may be constituted：Based on current board face information, calculates various board types and occur Probability distribution, board type occur the more big then corresponding character numerical value of probability it is bigger.

(3) the chip situation of each players bet during the wheel is betted：The chip situation of players bet implies its board power Information, so the chip situation of each players bet is extracted as characteristic value.

In summary the feature of three aspects, may be constructed the vector characteristics of one 219 dimension, which can be used as god Input parameter through network training.Feature extraction flow can be with as shown in figure 3, the effective situation information of Efficient Characterization may include： Hands, table board and chip.Based on hands it is contemplated that at board board power, if it is straight flush, corresponding characteristic value is 1, such as Fruit is the of inferior brand of minimum, then corresponding characteristic value is 0, the dimension that can be accounted at board board power in feature vector of hands.Base Other board types are considered into board possibility in table board and hands, include mainly the probability that 213 class board types respectively occur in holdem, These probability can account for 213 dimensions of feature vector as the characteristic value of feature vector, these probability.Based on bet Chip, lower mainstream imply the board force information of other players, by the way that the chip of other players bets is normalized, obtain The normalized value arrived can be used as characteristic value, if there is other are 5 for if player, then their chip situation can account for feature to 5 dimensions of amount.Obtained one-dimensional characteristic vector length is 219.

The process of neural network model training can be with as shown in figure 4, be described in detail below：The feature that said extracted is arrived to The input parameter as neural network model is measured, regard corresponding operation (including abandoning board, with entering, filling) as neural network model Output parameter.In the training process, study can be first passed through by emulating the data generated, summarize general sex experience, then Data of new generation are generated by playing chess certainly again, newly generated data is then recycled further to learn neural network model It practises.By iterating, above-mentioned training process is until trained neural network model meets the set goal condition.

After obtaining trained neural network model, trained neural network model can be deployed in product In, to realize man-machine chess.

It is alternatively possible to trained neural network model is deployed directly into client, as shown in figure 5, client can According to board face information extraction feature, to obtain feature vector, and obtained feature vector is input to trained neural network In model, exported with obtaining corresponding action, such as abandon board, with entering, filling.Trained neural network model is deployed in In client, directly man-machine chess can be realized in client, without carrying out the communication with server.

It is alternatively possible to trained neural network model is deployed in server, by network in client and service The action of the feature and Neural Network model predictive of extraction out is transmitted between device, realizes man-machine chess.For example, as shown in fig. 6, The feature of extraction can be sent to server by each client, and the trained neural network model of server by utilizing determines feature The corresponding action of vector, and by action output to client.

In holdem game, the held hands of game player and prompt operation letter can be shown on interface Breath, including abandons board, plays a card, fills, with entering, as shown in fig. 7, the held hands of game player are 4, respectively square 1, square 2, Square 3 and square 4 can determine that this 4 hands can be formed together through the embodiment of the present invention according to 4 hands held The probability of Hua Shun is higher, therefore identified object run is filling, then " filling " this prompting frame can highlight, with reality Now prompt user carries out filling operation.Herein it should be noted that the embodiment of the present invention to the highlighted mode of prompting frame not It is specifically limited, such as is highlighted with prompting frame amplification shown in Fig. 7.

As another optional example, the embodiment of the present invention can also be applied in Random Factor Mahjong.Random Factor Mahjong and moral Different, the dimension number of effect characteristics vector that differs only in board type type and quantity of state canaster, and on how to Feature vector is extracted, how to train neural network model, and how to be determined using trained neural network model corresponding dynamic Work (such as playing a card) is similar to foregoing description process, and details are not described herein again.

The feature extracting method that above-mentioned example proposes sets up probability distribution and player chips bet situation based on board type, main To have the advantages that following：

Method has certain versatility, is applicable not only to holdem, can be applied to other chess and card games, such as fiber crops It will.

It is trained using deep learning after extraction feature, trained neural network model ability of self-teaching is with training The increase of amount is promoted.

It should be noted that for each method embodiment above-mentioned, for simple description, therefore it is all expressed as a series of Combination of actions, but those skilled in the art should understand that, the present invention is not limited by the described action sequence because According to the present invention, certain steps can be performed in other orders or simultaneously.Secondly, those skilled in the art should also know It knows, embodiment described in this description belongs to preferred embodiment, and involved action and module are not necessarily of the invention It is necessary.

Through the above description of the embodiments, those skilled in the art can be understood that according to above-mentioned implementation The method of example can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hardware, but it is very much In the case of the former be more preferably embodiment.Based on this understanding, technical scheme of the present invention is substantially in other words to existing The part that technology contributes can be expressed in the form of software products, which is stored in a storage In medium (such as ROM/RAM, magnetic disc, CD), including some instructions are used so that a station terminal equipment (can be mobile phone, calculate Machine, server or network equipment etc.) execute method described in each embodiment of the present invention.

Other side according to the ... of the embodiment of the present invention additionally provides a kind of determination side for implementing above-mentioned object run The determining device of the object run of method.Fig. 8 is a kind of determining device of optional object run according to the ... of the embodiment of the present invention Schematic diagram, as shown in figure 8, the device may include：

Detection unit 22, the operation requests for detecting the first game role, wherein operation requests are for asking in visitor The first game role is based on the pending object run of the first object got in the currently running one innings of game in family end； First acquisition unit 24, in response to operation requests, obtaining target feature vector, wherein target feature vector is at least used for Instruction forms to obtain the probability of target object set based on the first object got, and target object set includes having got The first object；Second acquisition unit 26, it is corresponding for obtaining target feature vector according to trained neural network model Object run, wherein trained neural network model is using the mapping relations between sampling feature vectors and sample operations To the model that first nerves network model is trained, when being trained to first nerves network model, first nerves The input parameter of network model is sampling feature vectors, and the output parameter of first nerves network model is sample operations, is trained Neural network model be used to indicate the mapping relations between target feature vector and object run.

It should be noted that the detection unit 22 in the embodiment can be used for executing the step in the embodiment of the present application S202, the first acquisition unit 24 in the embodiment can be used for executing the step S204 in the embodiment of the present application, the embodiment In second acquisition unit 26 can be used for execute the embodiment of the present application in step S206.

Herein it should be noted that above-mentioned module is identical as example and application scenarios that corresponding step is realized, but not It is limited to above-described embodiment disclosure of that.It should be noted that above-mentioned module as a part for device may operate in as In hardware environment shown in FIG. 1, it can also pass through hardware realization by software realization.

Optionally, first acquisition unit 24 may include：First acquisition module, for obtaining based on first got Object forms to obtain the probability of target object set；First determining module, for using probability as the in target feature vector The element of dimension, wherein target feature vector includes multiple dimensions, and multiple dimensions include the first dimension.

Optionally, first acquisition unit 24 can also include：Second acquisition module, for obtaining the second trip in one innings of game The resource that play role is at least transferred out of based on the second object got；Processing module, for place to be normalized in resource Reason, obtains normalized value；Second determining module, for using normalized value as the member of the second dimension in target feature vector Element, wherein multiple dimensions further include the second dimension.

Optionally, sampling feature vectors are at least used to indicate the sample object got based on sample game role and are formed Obtain the probability of target object set；After sample game role executes sample operations based on the sample object got, at least Meet any one following goal condition：Form to obtain the probability of target object set most based on the sample object got It is high, formed to obtain based on the sample object got target object be integrated into one innings of game win, sample game role obtains The resource that arrives at most, the resource of sample game role loss it is minimum.

Optionally, first nerves network model is carried out using the mapping relations between sampling feature vectors and sample operations Training obtains trained neural network model and may include：Utilize the mapping relations between first eigenvector and the first operation First nerves network model is trained, nervus opticus network model is obtained, wherein is carried out to first nerves network model When training, the input parameter of first nerves network model is first eigenvector, and the output parameter of first nerves network model is First operation, sampling feature vectors include first eigenvector, and first eigenvector is used to indicate based on third game role The third object got forms to obtain the probability of target object set, and third game role is based on the third object got Meet goal condition after executing the first operation；Following steps are repeated, until the 4th game role is based on the got Four objects meet goal condition after executing the second operation got using trained neural network：By first resource and second Resource is compared, wherein first resource executes institute after third operates for the 5th game role to the 5th object got Obtained resource, Secondary resource are that the 6th game role executes the money obtained after the 4th operation to the 6th object got Source, third operation are the corresponding operation of second feature vector got using nervus opticus network model, second feature vector It is used to indicate and is formed to obtain the probability of target object set based on the 5th object got, the 4th operation is to utilize the second god The corresponding operation of third feature vector got through network model, third feature vector are used to indicate based on the got Six objects form to obtain the probability of target object set；In the case where first resource is more than Secondary resource, second feature is utilized Mapping relations between vector and third operation are trained nervus opticus network model, obtain trained neural network mould Type；In the case where Secondary resource is more than first resource, the mapping relations pair between third feature vector and the 4th operation are utilized Nervus opticus network model is trained, and obtains trained neural network model.

Optionally, device can also include：Control unit, for obtaining target according to trained neural network model After the corresponding object run of feature vector, the operation of the first game role performance objective is controlled in the client.

Optionally, device can also include：Push unit, for obtaining target according to trained neural network model After the corresponding object run of feature vector, the operation information of object run is used to indicate to client push, to indicate in visitor The operation of the first game role performance objective is controlled in the end of family.

By above-mentioned module, the object got based on game role can be reached and fast and accurately determine the game angle The purpose of the pending object run of color, and then solve the relevant technologies and behaviour in next step is calculated according to game player's experience Make, the relatively low technical problem of rate of winning of playing after causing game player to execute next operation, game angle is improved to realize The technique effect of the rate of winning of color in gaming.

Another aspect according to the ... of the embodiment of the present invention additionally provides a kind of determination side for implementing above-mentioned object run The electronic device of method.

Fig. 9 is a kind of structure diagram of electronic device according to the ... of the embodiment of the present invention, as shown in figure 9, the electronic device can To include：One or more (one is only shown in figure) processors 201, memory 203, wherein can be stored in memory 203 There are computer program, processor 201 to can be set to run the computer program to execute the target of the embodiment of the present invention The determination method of operation.

Wherein, memory 203 can be used for storing computer program and module, such as the object run in the embodiment of the present invention The corresponding program instruction/module of determination method and apparatus, processor 201 is stored in calculating in memory 203 by operation Machine program and module realize the determination side of above-mentioned object run to perform various functions application and data processing Method.Memory 203 may include high speed random access memory, can also include nonvolatile memory, such as one or more magnetism Storage device, flash memory or other non-volatile solid state memories.In some instances, memory 203 can further comprise phase For the remotely located memory of processor 201, these remote memories can pass through network connection to terminal.Above-mentioned network Example includes but not limited to internet, intranet, LAN, mobile radio communication and combinations thereof.

Optionally, as shown in figure 9, the electronic device can also include：Transmitting device 205 and input-output equipment 207. Wherein, transmitting device 205 is used for via a network reception or transmission data.Above-mentioned network specific example may include wired Network and wireless network.In an example, transmitting device 205 includes a network adapter (Network Interface Controller, NIC), can be connected with other network equipments with router by cable so as to internet or LAN It is communicated.In an example, transmitting device 205 is radio frequency (Radio Frequency, RF) module, is used to pass through nothing Line mode is communicated with internet.

It will appreciated by the skilled person that structure shown in Fig. 9 is only to illustrate, electronic device can be intelligent hand Machine (such as Android phone, iOS mobile phones), tablet computer, palm PC and mobile internet device (Mobile Internet Devices, MID), the terminal devices such as PAD.Fig. 9 it does not cause to limit to the structure of above-mentioned electronic device.Example Such as, electronic device can also include more than shown in Fig. 9 or less component (such as network interface, display device), or Person has the configuration different from shown in Fig. 9.

Optionally, in the present embodiment, above-mentioned memory 203 can be used for storing computer program.

Optionally, in the present embodiment, above-mentioned processor can be set to operation computer program, to execute following step Suddenly：Detect the operation requests of the first game role, wherein operation requests are for asking in the currently running one innings of trip of client The first game role is based on the pending object run of the first object got in play；In response to operation requests, obtain Target feature vector, wherein target feature vector is at least used to indicate to be formed to obtain target based on the first object got The probability of object set, target object set include the first object got；It is obtained according to trained neural network model Take the corresponding object run of target feature vector, wherein trained neural network model is to use sampling feature vectors and sample The model that mapping relations between this operation are trained first nerves network model, to first nerves network model When being trained, the input parameter of first nerves network model is sampling feature vectors, and the output of first nerves network model is joined Number is sample operations, and trained neural network model is used to indicate the pass of the mapping between target feature vector and object run System.

Processor 201 is additionally operable to execute following step：Acquisition forms to obtain target pair based on the first object got As the probability of set；Using probability as the element of the first dimension in target feature vector, wherein target feature vector includes more A dimension, multiple dimensions include the first dimension.

Processor 201 is additionally operable to execute following step：The second game role in one innings of game is obtained at least to be based on having obtained To the resource that is transferred out of of the second object；Resource is normalized, normalized value is obtained；Using normalized value as target The element of the second dimension in feature vector, wherein multiple dimensions further include the second dimension.

Processor 201 is additionally operable to execute following step：Utilize the mapping relations between first eigenvector and the first operation First nerves network model is trained, nervus opticus network model is obtained, wherein is carried out to first nerves network model When training, the input parameter of first nerves network model is first eigenvector, and the output parameter of first nerves network model is First operation, sampling feature vectors include first eigenvector, and first eigenvector is used to indicate based on third game role The third object got forms to obtain the probability of target object set, and third game role is based on the third object got Meet goal condition after executing the first operation；Following steps are repeated, until the 4th game role is based on the got Four objects meet goal condition after executing the second operation got using trained neural network：By first resource and second Resource is compared, wherein first resource executes institute after third operates for the 5th game role to the 5th object got Obtained resource, Secondary resource are that the 6th game role executes the money obtained after the 4th operation to the 6th object got Source, third operation are the corresponding operation of second feature vector got using nervus opticus network model, second feature vector It is used to indicate and is formed to obtain the probability of target object set based on the 5th object got, the 4th operation is to utilize the second god The corresponding operation of third feature vector got through network model, third feature vector are used to indicate based on the got Six objects form to obtain the probability of target object set；In the case where first resource is more than Secondary resource, second feature is utilized Mapping relations between vector and third operation are trained nervus opticus network model, obtain trained neural network mould Type；In the case where Secondary resource is more than first resource, the mapping relations pair between third feature vector and the 4th operation are utilized Nervus opticus network model is trained, and obtains trained neural network model.

Processor 201 is additionally operable to execute following step：According to trained neural network model obtain target signature to After measuring corresponding object run, the operation of the first game role performance objective is controlled in the client.

Processor 201 is additionally operable to execute following step：According to trained neural network model obtain target signature to After measuring corresponding object run, the operation information of object run is used to indicate to client push, to indicate in the client Control the operation of the first game role performance objective.

Optionally, the specific example in the present embodiment can refer to the example described in above-described embodiment, the present embodiment Details are not described herein.

Using the embodiment of the present invention, a kind of determination scheme of object run is provided.By detecting the first game angle After the operation requests of color, the first object acquisition target feature vector for having been got according to the first game role, wherein target is special Sign vector is at least used to indicate to be formed to obtain the probability of target object set based on the first object got, then utilizes instruction The neural network model perfected obtains the corresponding object run of target feature vector, has reached and has been got based on game role Object fast and accurately determines the purpose of the pending object run of the game role, and then solves the relevant technologies according to game Player's experience calculates to be operated in next step, and the game lower technology of rate of winning is asked after causing game player to execute next operation Topic, to realize the technique effect for improving the rate of winning of game role in gaming.

Another aspect according to the ... of the embodiment of the present invention additionally provides a kind of storage medium.It is stored in the storage medium Computer program, wherein the computer program is arranged to execute the determination method of object run in above-described embodiment when operation The step of.

Optionally, in the present embodiment, storage medium can be located at multiple networks in network shown in above-described embodiment On at least one of equipment network equipment.

Optionally, in the present embodiment, storage medium is arranged to store the computer program for executing following steps：

S1 detects the operation requests of the first game role, wherein operation requests are currently run for asking in client One innings game in the first game role based on the pending object run of the first object got；

S2 obtains target feature vector, wherein target feature vector is at least used to indicate and is based in response to operation requests The first object got forms to obtain the probability of target object set, and target object set includes the first couple got As；

S3 obtains the corresponding object run of target feature vector, wherein train according to trained neural network model Neural network model be using between sampling feature vectors and sample operations mapping relations to first nerves network model into The model that row training obtains, when being trained to first nerves network model, the input parameter of first nerves network model is The output parameter of sampling feature vectors, first nerves network model is sample operations, and trained neural network model is for referring to Show the mapping relations between target feature vector and object run.

Optionally, storage medium is also configured to store the computer program for executing following steps：It obtains based on The first object got forms to obtain the probability of target object set；Using probability as the first dimension in target feature vector Element, wherein target feature vector includes multiple dimensions, and multiple dimensions include the first dimension.

Optionally, storage medium is also configured to store the computer program for executing following steps：Obtain one innings of trip The resource that the second game role is at least transferred out of based on the second object got in play；Resource is normalized, Obtain normalized value；Using normalized value as the element of the second dimension in target feature vector, wherein multiple dimensions further include Second dimension.

Optionally, storage medium is also configured to store the computer program for executing following steps：Utilize the first spy Mapping relations between sign vector and the first operation are trained first nerves network model, obtain nervus opticus network mould Type, wherein when being trained to first nerves network model, the input parameter of first nerves network model be fisrt feature to The output parameter of amount, first nerves network model is the first operation, and sampling feature vectors include first eigenvector, fisrt feature Vector is used to indicate the third object got based on third game role and is formed to obtain the probability of target object set, third Game role meets goal condition after executing the first operation based on the third object got；Following steps are repeated, directly The second operation got using trained neural network is executed based on the 4th object got to the 4th game role After meet goal condition：First resource is compared with Secondary resource, wherein first resource is the 5th game role to having obtained The 5th object got executes obtained resource after third operation, Secondary resource be the 6th game role to got the Six objects execute the resource obtained after the 4th operation, and third operation is the second feature got using nervus opticus network model The corresponding operation of vector, second feature vector is used to indicate to be formed to obtain target object set based on the 5th object got Probability, the 4th operation is the corresponding operation of third feature vector got using nervus opticus network model, third feature Vector is used to indicate to be formed to obtain the probability of target object set based on the 6th object got；It is more than the in first resource In the case of two resources, nervus opticus network model is carried out using the mapping relations between second feature vector and third operation Training, obtains trained neural network model；In the case where Secondary resource is more than first resource, third feature vector is utilized Mapping relations between the 4th operation are trained nervus opticus network model, obtain trained neural network model.

Optionally, storage medium is also configured to store the computer program for executing following steps：According to training After good neural network model obtains the corresponding object run of target feature vector, the first game role is controlled in the client Performance objective operates.

Optionally, storage medium is also configured to store the computer program for executing following steps：According to training After good neural network model obtains the corresponding object run of target feature vector, target behaviour is used to indicate to client push The operation information of work controls the operation of the first game role performance objective in the client with instruction.

Optionally, in the present embodiment, one of ordinary skill in the art will appreciate that it is complete in the method for above-described embodiment Portion or part steps are can be completed come command terminal device-dependent hardware by program, which can be stored in a meter In calculation machine readable storage medium storing program for executing, storage medium may include：Flash disk, read-only memory (Read-Only Memory, ROM), Random access device (Random Access Memory, RAM), disk or CD etc..

The embodiments of the present invention are for illustration only, can not represent the quality of embodiment.

If the integrated unit in above-described embodiment is realized in the form of SFU software functional unit and as independent product Sale in use, can be stored in the storage medium that above computer can be read.Based on this understanding, skill of the invention Substantially all or part of the part that contributes to existing technology or the technical solution can be with soft in other words for art scheme The form of part product embodies, which is stored in a storage medium, including some instructions are used so that one Platform or multiple stage computers equipment (can be personal computer, server or network equipment etc.) execute each embodiment institute of the present invention State all or part of step of method.

In the above embodiment of the present invention, all emphasizes particularly on different fields to the description of each embodiment, do not have in some embodiment The part of detailed description may refer to the associated description of other embodiment.

In several embodiments provided herein, it should be understood that disclosed client, it can be by others side Formula is realized.Wherein, the apparatus embodiments described above are merely exemplary, for example, the unit division, only one Kind of division of logic function, formula that in actual implementation, there may be another division manner, such as multiple units or component can combine or It is desirably integrated into another system, or some features can be ignored or not executed.Another point, it is shown or discussed it is mutual it Between coupling, direct-coupling or communication connection can be INDIRECT COUPLING or communication link by some interfaces, unit or module It connects, can be electrical or other forms.

The unit illustrated as separating component may or may not be physically separated, aobvious as unit The component shown may or may not be physical unit, you can be located at a place, or may be distributed over multiple In network element.Some or all of unit therein can be selected according to the actual needs to realize the mesh of this embodiment scheme 's.

In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, it can also It is that each unit physically exists alone, it can also be during two or more units be integrated in one unit.Above-mentioned integrated list The form that hardware had both may be used in member is realized, can also be realized in the form of SFU software functional unit.

The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art For member, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also answered It is considered as protection scope of the present invention.

Claims

1. a kind of determination method of object run, which is characterized in that including：

Detect the operation requests of the first game role, wherein the operation requests are currently running in client for asking The first game role is based on the pending object run of the first object got described in one innings of game；

In response to the operation requests, target feature vector is obtained, wherein the target feature vector, which is at least used to indicate, to be based on First object got forms to obtain the probability of target object set, and the target object set includes described obtained The first object got；

The corresponding object run of the target feature vector is obtained according to trained neural network model, wherein described Trained neural network model is using the mapping relations between sampling feature vectors and sample operations to first nerves network The model that model is trained, when being trained to first nerves network model, the first nerves network model Input parameter is the sampling feature vectors, and the output parameter of the first nerves network model is the sample operations, described Trained neural network model is used to indicate the mapping relations between the target feature vector and the object run.

2. according to the method described in claim 1, it is characterized in that, the acquisition target feature vector includes：

Acquisition forms to obtain the probability of the target object set based on first object got；

Using the probability as the element of the first dimension in the target feature vector, wherein the target feature vector packet Multiple dimensions are included, the multiple dimension includes first dimension.

3. according to the method described in claim 2, it is characterized in that, the acquisition target feature vector further includes：

Obtain the resource that the second game role is at least transferred out of based on the second object got described in one innings of game；

The resource is normalized, normalized value is obtained；

Using the normalized value as the element of the second dimension in the target feature vector, wherein the multiple dimension is also Including second dimension.

4. according to the method described in claim 1, it is characterized in that,

The sampling feature vectors are at least used to indicate the sample object got based on sample game role and are formed to obtain institute State the probability of target object set；

After the sample game role executes the sample operations based on the sample object got, at least meet following Any one goal condition：Form to obtain the probability of the target object set most based on the sample object got It is high, formed to obtain based on the sample object got the target object be integrated into one innings of game win, institute The resource for stating most, the described sample game role losses of the resource that sample game role obtains is minimum.

5. method according to claim 1 to 4, which is characterized in that described according to trained nerve net After network model obtains the corresponding object run of the target feature vector, the method further includes：

First game role is controlled in the client executes the object run.

6. method according to claim 1 to 4, which is characterized in that described according to trained nerve net After network model obtains the corresponding object run of the target feature vector, the method further includes：

It is used to indicate the operation information of the object run to the client push, institute is controlled in the client with instruction It states the first game role and executes the object run.

7. a kind of determining device of object run, which is characterized in that including：

Detection unit, the operation requests for detecting the first game role, wherein the operation requests are for asking in client Hold the first game role described in currently running one innings of game based on the pending target behaviour of the first object got Make；

First acquisition unit, in response to the operation requests, obtaining target feature vector, wherein the target signature to Amount is at least used to indicate to be formed to obtain the probability of target object set, the target pair based on first object got Include first object got as gathering；

Second acquisition unit, for obtaining the corresponding mesh of the target feature vector according to trained neural network model Mark operation, wherein the trained neural network model is closed using the mapping between sampling feature vectors and sample operations It is the model that is trained to first nerves network model, when being trained to first nerves network model, described the The input parameter of one neural network model is the sampling feature vectors, and the output parameter of the first nerves network model is institute Sample operations are stated, the trained neural network model is used to indicate between the target feature vector and the object run Mapping relations.

8. device according to claim 7, which is characterized in that the first acquisition unit includes：

First acquisition module forms to obtain the target object set for obtaining based on first object got Probability；

First determining module, for using the probability as the element of the first dimension in the target feature vector, wherein institute It includes multiple dimensions to state target feature vector, and the multiple dimension includes first dimension.

9. device according to claim 8, which is characterized in that the first acquisition unit further includes：

Second acquisition module, for obtaining the second game role described in one innings of game at least based on second got The resource that object is transferred out of；

Processing module obtains normalized value for the resource to be normalized；

Second determining module, for using the normalized value as the element of the second dimension in the target feature vector, In, the multiple dimension further includes second dimension.

10. device according to claim 7, which is characterized in that

11. device according to any one of claims 7 to 10, which is characterized in that described device further includes：

Control unit, for described corresponding described according to the trained neural network model acquisition target feature vector After object run, first game role is controlled in the client and executes the object run.

12. device according to any one of claims 7 to 10, which is characterized in that described device further includes：

Push unit, for described corresponding described according to the trained neural network model acquisition target feature vector After object run, the operation information of the object run is used to indicate to the client push, to indicate in the client First game role is controlled in end executes the object run.

13. a kind of storage medium, which is characterized in that be stored with computer program in the storage medium, wherein the computer Program is arranged to execute the method described in any one of claim 1 to 6 when operation.

14. a kind of electronic device, including memory and processor, which is characterized in that be stored with computer journey in the memory Sequence, the processor are arranged to run the computer program to execute the side described in any one of claim 1 to 6 Method.