CN109925718B - System and method for distributing game micro-terminal map - Google Patents
System and method for distributing game micro-terminal map Download PDFInfo
- Publication number
- CN109925718B CN109925718B CN201910032178.7A CN201910032178A CN109925718B CN 109925718 B CN109925718 B CN 109925718B CN 201910032178 A CN201910032178 A CN 201910032178A CN 109925718 B CN109925718 B CN 109925718B
- Authority
- CN
- China
- Prior art keywords
- map
- user
- game
- sequence
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The technical scheme of the invention comprises a system and a method for distributing a game micro-terminal map, which are used for realizing the following steps: acquiring user data uploaded to a server by a game client; constructing a training sample according to the user data, and further, creating a time sequence prediction model based on the training sample and the time recurrent neural network; and acquiring a map sequence which is logged in and out of the period of the game by the user, calculating a sequence number of the map to be accessed by the user by using a time sequence prediction model, and executing early loading. The beneficial effects of the invention are as follows: according to the method, historical data of a user login map is analyzed, a time sequence prediction model is trained, and a map expected to be accessed next by the user is reversely deduced according to a map login list of the user to be judged; and the map is downloaded in advance by utilizing the game micro-terminal, so that the waiting downloading time of a user is reduced, and the user experience is improved.
Description
Technical Field
The invention relates to a system, a method and a device for distributing a game micro-terminal map, belonging to the field of computer games.
Background
The miniature clients (abbreviated as micro-terminals) of the game are generally small in size, and the installation package generally only contains some configuration tables, action sound effects of the game character model, maps of novice villages, character models, action special effects and other resources. The user can download and install the game experience in a short time, and the micro-terminal can start downloading of other resources (mainly map resources) which are not downloaded in the background in the experience process, so that the user can play and download without perception. The user experience of this process is better, eliminating the lengthy wait process of downloading all of the resources of the game. The method can effectively solve the problems that most network game clients are large in size, long in user downloading and installing time, low in logged-in users and low in game user survival rate.
Considering that the map resource packages are more and larger, the micro-terminals have sequence and time for downloading when downloading the resource packages in the background. When a user is about to use and access resources in the micro-terminal installation package that are not downloaded at the time, the micro-terminal downloads the map to the server and loads the map into the game. This process can cause the user to interrupt game operations and wait for the download to complete, which affects the large user experience. How to accurately predict the map resource expected to be used by each user and enable the micro-terminal to download in the background in advance to improve the user experience is an industrial technical problem. According to the known literature, the industry does not have mature technology for accurately predicting the access sequence of each user to the map. Conventional prediction algorithms are generally based on popular predictions or current map association classifications, i.e., determining the downloading priority according to the popularity and frequency of use of the map, or counting and calculating the association frequency of the map according to the current map to determine the downloading priority. The method does not consider the time sequence correlation and the user attribute of the user login map, has poor effect in practical application and has low accuracy. Generally, a user logs into a map for some purpose, such as monster, copy, collect, or perform a specified task; the user has an association to jump from the last map to the next map, such as receiving a new task to a new map to operate. The association between maps is affected by a number of factors, such as the design of the game scenario, the activity tasks, the status of the player's character, etc. It is difficult for the conventional method to find the timing rule from such many factors.
According to research, in the field of game micro-terminals, research in the industry mainly focuses on the realization of a mechanism for downloading and playing at the same time after micro-terminal installation. For example, CN104503784a, a method and system for controlling downloading of a micro-terminal by using script, proposes a downloading rule of the micro-terminal, which includes establishing a downloading level mechanism, analyzing and optimizing a downloading sequence of resources, prejudging resources required by a subsequent task according to a current task, downloading in advance, merging downloads, and the like. CN105988809a, an online loading method of game resources and a micro-end engine, describes an online loading method and a plurality of modules, including a game resource pool, an updating module, a resource information list file, a resource storage module, a receiving module, a multithreaded downloading module and a pushing module. CN105847429A, a method for realizing micro-terminal downloading resource, provides a micro-terminal design method for multi-account sharing interaction.
In map resource prediction, conventional prediction algorithms are generally based on popular predictions or current map association classifications. The hot prediction is to count the frequency of logging in the map by the user and determine the downloading priority sequence of the map according to the frequency, namely the hot map is downloaded preferentially. The association classification is to count association frequency between every two maps in a log of a login map of a user; and when the map where the user to be judged is currently located is given, finding out all maps associated with the map, and sorting according to the association frequency, and downloading the map in the priority mode before sorting. The two methods do not consider the time sequence correlation of the user login map, and have poor effect and low accuracy in practical application. In order to solve the problems, the invention finds out the priority order rule of jump access among maps by utilizing a time sequence mining algorithm by mining the map login log of the user, and accurately predicts the map to be logged by each user by utilizing the rule. The method has strong universality in the field of the invention, and can be used in the field of the access sequence of articles.
Most of the traditional machine learning methods (such as SVM, logistic regression, forward neural network, etc.) are based on the premise of independence assumption, and have no explicit modeling of time. This writing method can only handle independent inputs, i.e. the previous and the next input are completely irrelevant. But in the time-series data, the former input and the latter input are related. For example, when understanding the meaning of a sentence, it is not enough to understand each word of the sentence in isolation, and it is necessary to process the whole sequence in which the words are connected, which has semantics. A small number of methods (e.g., deep belief networks) connect nodes and temporally predecessor nodes with successor nodes, with temporal implicit modeling through a sliding window of contexts. However, the models ignore the actual situation of long-term dependence in time sequence data, for example, a model trained by using a time window with a size of 5 cannot answer a problem requiring 6 time inputs; resulting in poor performance.
Yet another part of the methods (e.g., markov chain models) can explicitly express time dependence but are difficult to apply in practice due to excessive computational complexity. Specifically, the Markov chain model uses a dynamic programming algorithm, which is the square of the state space size in terms of computational complexity. Secondly, the size of the model transition table (the probability of the state transition trajectory at two arbitrary time points) is also the square of the state space size; if the number of hidden states is large, the operation complexity increases in square number. In addition, the increase in hidden state space grows exponentially with the size of the window; resulting in excessive computational complexity.
Disclosure of Invention
The invention provides a system and a method for distributing a game micro-terminal map, which adopt a time sequence recurrent neural network model to capture long-time dependent information, and compared with a Markov chain model, the current state of the model depends on the current input and the network state of the last time step, and a hidden layer contains network state information of any time point and almost corresponds to the time context of any length. And the number of different states may increase exponentially as the number of nodes of the hidden layer increases. For example, each node is represented by a value of 2, and the N hidden layer nodes can also represent states of the power of 2, so that the expressive force is super strong. When the potential expression capacity grows exponentially along with the growth of the node number, the complexity of deduction and training grows quadratically, the calculation time complexity is low, and the practical application capacity is achieved. From the experimental results, the scheme of the invention can capture the time sequence information of high distinction degree among the maps and accurately predict the map to be registered by the player.
The technical scheme of the invention comprises a system for distributing a game micro-terminal map, which comprises a data access unit, a time sequence training unit and a map prediction unit, and is characterized in that: the data access unit is used for acquiring user data uploaded to the game server by the game client; the time sequence training unit is used for constructing training samples according to the user data, and further, creating a time sequence prediction model based on the training samples and the time recurrent neural network; and the prediction unit is used for acquiring the map sequence which is logged in the period of the user logging in and out of the game, calculating the sequence number of the map which is about to be accessed by the user by using the time sequence prediction model, and executing the loading in advance.
The system for distributing a game micro-terminal map according to claim, wherein the data access unit obtains user data uploaded to the game server by the game client, wherein the user data comprises user attribute data and user behavior data, and the user attribute data comprises: the user attribute data are superior attribute data to which the game roles belong, wherein each micro-terminal game comprises a plurality of superior attribute data, and each superior game attribute comprises a plurality of game roles; the user behavior data comprises map id lists accessed in the process of logging in and logging out of the game once, wherein the map id lists are arranged in the order of access.
The system for distributing game micro-terminal maps according to claim, wherein the time sequence training unit specifically comprises: the training data module is used for carrying out grouping processing on a plurality of users in the game according to at least two superior attributes to obtain a plurality of user groups, wherein each user group comprises a plurality of users and a map access sequence corresponding to the plurality of users; and the time sequence model module is used for inputting the map sequence accessed by each user, and training the input map sequence by using the time sequence recurrent neural network to obtain a time sequence model for predicting the map to be accessed by the user.
The system for distributing a game micro-end map according to, wherein the timing model module is further configured to perform the steps of: taking the map sequence accessed by each user as input and recording as { x } 1 ,x 2 ,...,x t }，x t Id number representing map; constructing a hidden layer state with time sequence information and recording as { h } 1 ,h 2 ,...,h t -a }; based on hidden layer state h t And input x t And the time sequence caused by the state of the upper hidden layer depends on time for a long time, a mapping function is constructed to output an estimated map value, and the estimated map id value is obtained and recorded as { y } 1 ,y 2 ,...,y t }。
The system for distributing a game micro-end map according to claim, wherein the mapping function further comprises: the hidden layer state h t By a mapping function f w According to the hidden layer state h t-1 And x of input t Generating a calculation formula h t ＝f w (h t-1 ,x t ) Transport and deliverGo out y t From the mapping function parameter W and the current hidden layer state h t Generating, the calculation formula is y t ＝W hy h t The method comprises the steps of carrying out a first treatment on the surface of the The mapping function f w The method comprises the following steps: h is a t ＝o t ⊙tanh(f t ⊙c t-1 +i t ⊙tanh(W xc x t +W hc h t-1 +b c ) For mapping function f) w Wherein o t As output unit, f t ⊙c t-1 +i t As input element, wt.) xc x t +W hc h t-1 +b c Is a forgetting unit.
The system for distributing a game micro-terminal map according to claim, wherein the output unit specifically comprises: for conditionally performing corresponding output according to the input signal and hidden layer state, with a calculation formula of h t ＝o t ⊙tanh(c t ) Specifically: building c using Sigmoid layer t Information o to be output in (a) t O is equal to t Mapping between (0, 1); using tanh layer to handle c t Regularized to between-1 and 1; and multiplying the two types of information weights in the steps, and outputting a result.
The system for distributing a game micro-terminal map according to claim, wherein the input unit specifically comprises: conditionally determining the value of the state of the inner hidden layer to be updated from the input signal, in particular by the tanh layer for generating a new candidate value c t (-1, 1) with a formula of
Concealing the state of not updated layer c t-1 And f t Multiplying and deleting redundant information; adding i t *c t The new content to be added is updated as a unit, and its calculation formula is +>
The system for distributing a game micro-terminal map according to claim, wherein the forgetting unit specifically comprises: for conditional screening of input and/or output information, withBody ground, according to the output h of the last moment t-1 And current input x t To generate a vector f of 0 to 1 t And take f t Determining whether to let last learned information c t-1 A pass or partial pass, wherein 0 indicates that no information is allowed to pass, 1 indicates that all information is allowed to pass, and the calculation formula is: f (f) t ＝σ(W xc x t +W hc h t-1 +b c )。
The system for distributing a game micro-end map according to, wherein the time sequence training unit comprises a model training module, and is characterized in that: the method is used for training a model by using a Back Propagation (BPTT) algorithm with time, wherein the BPTT algorithm is a training algorithm aiming at a circulating layer, and a mapping function f can be obtained after training w Is a hidden layer state parameter { h } 1 ,h 2 ,...,h t }。
The system for distributing a game micro-terminal map according to claim, wherein the prediction unit specifically comprises: the model selection module is used for judging the basic attributes of the users accessing the micro-terminals, judging the superior attributes of the users according to the basic attributes of the users, and further selecting the corresponding prediction models; model prediction module for inputting map sequence { x } 1 ,x 2 ,...,x t Time sequence model according to y t ＝W hy h t Computing each map candidate y for the next possible access t And selecting the map id with the highest probability as a prediction result.
The technical scheme of the invention also comprises a method for distributing the game micro-terminal map, which comprises the following steps: acquiring user data uploaded to a server by a game client; constructing a training sample according to the user data, and further, creating a time sequence prediction model based on the training sample and the time recurrent neural network; and acquiring a map sequence which is logged in and out of the period of the game by the user, calculating a sequence number of the map to be accessed by the user by using a time sequence prediction model, and executing early loading.
The beneficial effects of the invention are as follows: according to the method, historical data of a user login map is analyzed, a time sequence prediction model is trained, and a map expected to be accessed next by the user is reversely deduced according to a map login list of the user to be judged; and the map is downloaded in advance by utilizing the game micro-terminal, so that the waiting downloading time of a user is reduced, and the user experience is improved.
Drawings
FIG. 1 is a diagram of a system framework according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a map access list sample according to an embodiment of the present invention;
FIG. 3 is a timing model training flow diagram according to an embodiment of the present invention;
FIG. 4a is a diagram of recurrent neural network training logic;
FIG. 4b shows a neural network training sample according to an embodiment of the present invention;
FIG. 5a shows a data output y according to an embodiment of the invention t A schematic diagram;
FIG. 5b shows a mapping function f according to an embodiment of the invention w A schematic diagram;
fig. 6 is a flow chart of a predictive map according to an embodiment of the invention.
Detailed Description
The conception, specific structure, and technical effects produced by the present invention will be clearly and completely described below with reference to the embodiments and the drawings to fully understand the objects, aspects, and effects of the present invention.
It should be noted that, unless otherwise specified, when a feature is referred to as being "fixed" or "connected" to another feature, it may be directly or indirectly fixed or connected to the other feature. Further, the descriptions of the upper, lower, left, right, etc. used in this disclosure are merely with respect to the mutual positional relationship of the various components of this disclosure in the drawings. As used in this disclosure, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. Furthermore, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in the description presented herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and/or" as used herein includes any combination of one or more of the associated listed items.
It should be understood that although the terms first, second, third, etc. may be used in this disclosure to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element of the same type from another. For example, a first element could also be termed a second element, and, similarly, a second element could also be termed a first element, without departing from the scope of the present disclosure. The use of any and all examples, or exemplary language (e.g., "such as") provided herein, is intended merely to better illuminate embodiments of the invention and does not pose a limitation on the scope of the invention unless otherwise claimed.
Fig. 1 is a system frame diagram according to an embodiment of the present invention. The system mainly comprises three parts, namely a behavior data access unit 101, a time sequence training unit 102 and a map prediction unit 103. In particular, the method comprises the steps of,
the data access unit is responsible for accessing two types of data uploaded from the client to the server, and comprises the following steps:
user attributes: a matrix and a genre (game upper level attribute) of a user character in a game, and the like; the game map scene experienced by the user is different in different camps and genres, and personalized map prediction is needed according to the information.
User behavior: a list of map ids that the user has accessed during a log-in to log-out of the game. The user can be regarded as one experience period from the game login to the game login, the map list logged in the period is collected, the list stores the login sequence of the maps, and the sequence contains the association rule. The example can refer to fig. 2, which is a basic information number source predicting that a user will access a map.
The time sequence training unit is used for constructing a training sample based on the attribute and the behavior log of the user and generating a time sequence prediction model.
The map prediction unit is responsible for predicting the next map to be accessed by a given map sequence which a user has logged in during a login and logout experience period.
FIG. 3 is a flow chart of training a time series model according to an embodiment of the present invention. Based on the attributes and behavior logs of the user, training samples are constructed for generating a time sequence prediction model. The training process is described with reference to fig. 3. The method comprises two parts, including generating training data and training time sequence model in groups. In particular, the method comprises the steps of,
training data: considering that the scenes seen by users of different genres and camps are different, the map resources required to be accessed are also different. In order to more accurately find the access association rule of various users to map resources, the invention firstly groups the users according to the information of camping and genres. And then customizing and constructing training data aiming at each group for training a time sequence model to effectively describe the sequence rule of each group of users accessing map resources. In the experiment, the game had 4 camps and 13 house groups, split into 4 x 13 user groups. The training data is an access order list of the users in each group to the map in the login and logout experience period, as shown in fig. 2.
Time sequence model: the invention adopts a time recurrent neural network with long and short term memory to model the map access sequence and excavate the sequence rule. The network captures sequence information in the serialized data through periodical connection of hidden layer nodes, and can predict the serialized data. Unlike other forward neural networks, the time-series recurrent neural network can save the state of a context, and can even store, learn and express related information in context windows of any length. For example, in the text field, the front and back words in a sentence are not independent; when predicting what the next word of a sentence is, it is generally necessary to know the previous word. The time sequence recurrent neural network can effectively capture context front and back information (namely instant information) among words, is widely applied to the fields of video understanding, voice recognition, text processing and the like, and has better performance. The method is applied to a new field, namely map prediction of game micro-terminals.
Fig. 4a is a diagram of recurrent neural network training logic. Specifically, the map sequence accessed by each user is used as input and is recorded as { x } 1 ,x 2 ,...,x t }，x t ID number representing map, build hidden layer state with time sequence information, record as { h } 0 ,h 1 ,...,h t -a }; based on hidden layer state ht and input x t And the time sequence caused by the state of the upper hidden layer depends on a long time, a mapping function is constructed to output an estimated map value, and the estimated map value is recorded as { y } 1 ,y 2 ,...,y t }. The logic diagram refers to fig. 4a. The training process is to reversely deduce the parameters of the mapping function relation and the hidden layer state parameters { h }, through training data 0 ,h 1 ,...,h t }。
Fig. 4b shows a neural network training sample according to an embodiment of the present invention. The time series recurrent neural network builds a functional mapping relationship using the user's access map time series data as training sources, such as ' 283, 293, 102,2, 73 ', see fig. 5a. Specifically, when the user is on the map with the ID 283, the map that he expects to go next is 293; the neural network constructs a mapping function to output 293 based on input 283
FIG. 5a shows a data output y according to an embodiment of the invention t Schematic diagram. Wherein the current hidden layer state h t By a mapping function f W (.) according to hidden layer state h t-1 And input x t Generating, wherein the specific formula refers to formula 1-1; output y t From the mapping function parameter W and the current hidden layer state h t Specific formulas are generated with reference to formulas 1-2.
h t ＝f w (h t-1 ,x t ) 1-1
y t ＝W hy h t 1-2
FIG. 5b shows a mapping function f according to an embodiment of the invention w Schematic diagram. In order to solve the long-term dependence problem, the present invention uses equations 1-3 as the mapping function f W ()。
h t ＝o t ⊙tanh(f t ⊙c t-1 +i t ⊙tanh(W xc x t +W hc h t-1 +b c ) Formula 1-3
This mapping function adds a unit to determine that information is useful. Only the information conforming to the algorithm authentication is left, and the information not conforming to the algorithm authentication is forgotten through a forgetting door. The forgetting mechanism can effectively solve the problem of long-term dependence in time sequence, in particular to the problem of computational complexity. This mapping function consists of three parts, including an input unit, a forgetting unit, and an output unit.
Forgetting unit: the unit is responsible for conditionally deciding what information to throw away from the block to promote the generalization ability of the model. According to the output h of the last moment t-1 And current input x t To generate a vector f of 0 to 1 t And take f t Determining whether to let last learned information c t-1 Pass through or partially pass through. Where 0 means "do not let any information pass" and 1 means "let all information pass". Specific reference is made to FIGS. 1-4
f t ＝σ(W xc x t +W hc h t-1 +b c ) 1-4
An input unit: the unit is responsible for conditionally determining the value of the update of the inner hidden layer state from the input signal. First, a new candidate value-c is generated through the tanh layer t The tan h layer has the main function of classifying the numerical values between-1 and 1, and is shown in the reference formulas 1-5;
then multiplying old states ct-1 and ft, forgetting some unwanted information; then, if×ct is added as new content to be added to perform cell update, referring to equations 1 to 6.
An output unit: the unit conditionally determining the output based on the input signal and the internal hidden layer state of the blockWhat is going out. First build c using sigmoid layer t Information o to be output in (a) t Where sigmoid is often used as a threshold function of a neural network, mapping variables between 0, 1; then c is carried out by using the tanh layer t Regularized to between-1 and 1; finally, the two kinds of information weights are multiplied, the result is output, and the participation formulas 1-7 are adopted.
h t ＝o t ⊙tanh(c t ) 1-7
The invention trains the model using the back propagation BPTT algorithm over time. The BPTT algorithm is a training algorithm for the loop layer, comprising three steps:
1. forward calculating an output value of each neuron;
2. reversely calculating an error term value of each neuron, which is the partial derivative of the error function E on the weighted input of the neuron j;
3. and calculating the gradient of each weight, and finally updating the weight by using a random gradient descent algorithm.
After training, the mapping function f can be obtained W (-) parameter, hidden layer state parameter { h } 0 ,h 1 ,...,h t }。
Fig. 6 is a flow chart of a predictive map according to an embodiment of the invention. Given a sequence of maps that the user has logged in during a log-in and log-out experience period, the unit is responsible for predicting the next map he will access, flow is referred to in fig. 6. For the user to be determined, a suitable time sequence model is first selected according to the attributes (including the genres and camps) of the unit 401, and the model is used for prediction, referring to the unit 402.
Model selection: considering that the map scenes seen by users of different genres and camps are different, the unit 3 trains a corresponding model according to the group users in a one-to-one correspondence. In the experiment, the game has 4 camps and 13 genres, and is divided into 4×13 user groups; each user group trains a corresponding time sequence model, namely 4×13 models. Firstly, selecting a corresponding time sequence model from 4 multiplied by 13 models according to the attribute of a user to be judged, namely the affiliated house group and the camping.
Model prediction: map based on inputSequence { x 1 ,x 2 ,...,x t The timing model calculates each map candidate y that is next likely to be accessed according to equations 1-2 t The map ID with the highest probability is selected as the prediction result.
Specifically, the timing model considers how many consecutive inputs of data before each input data is linked. For example, the map sequence frequently occurring in training data is "… abcdbedf …"; when the user accesses the map "D", the time sequence model calculation finds that when the user has accessed the maps "B" and "C" before accessing the map "D", then the probability that the prediction output at the moment is B is higher; if the data received by the user before accessing "D" are "C" and "E", the probability that the predicted output at this time is F is greater. When the map sequence accessed by the user to be determined is 'BCD', the timing model predicts that the map to be accessed by the user is B by calculating all map candidates which are possible to be accessed next and finding the probability of B to be the largest.
The technical scheme of the invention further provides effect verification of the system. The unit verifies the effect of the project, and the configuration and results of the experiment are as follows:
verification configuration: the personality prediction result is applied to game micro-terminal map distribution; users are equally divided into two groups by AB testing, where group a applies timing predictions and group B applies old download mechanisms (including popular and associative maps). When the predicted result is matched with a map on which the character is to be registered, a hit is calculated once. And verifying the usability and user experience conditions of different mechanisms by counting the map access hit rate, wherein the hit rate is equal to the hit times divided by the request times. Experience actual data verifies that the hit rate of the algorithm is improved to 32.5% compared with that of a popular map method and is improved to 23.4% compared with that of the popular map method, and the algorithm is obviously superior to that of the traditional method.
The applicant of the invention considers that the method trains a time sequence prediction model by analyzing the historical data of the user logging on the map, and pushes back the map expected to be accessed by the user in the next step according to the map logging list of the user to be judged. Further, the map is downloaded in advance by utilizing the game micro-terminal, so that waiting downloading time of a user is reduced, and user experience is improved. According to known documents, there is no mature method for accurately predicting the access order of a map. Conventional prediction algorithms are generally based on hot prediction or map association analysis, and the methods do not consider time sequence correlation of a user login map and user attributes, so that the prediction accuracy is not high. Unlike traditional method, the invention makes full use of the sequence rule of the access between maps, and improves the prediction precision by using the sequence rule, and obviously improves the recognition accuracy compared with the traditional method. In experiments, the invention is tested on historical data of map access, and the predicted hit rate is 86.5 percent on average. The method and the system apply the predicted result to the individual distribution of the micro-terminal map of the game, predict the map to be accessed for each user individual, and improve the hit rate to 32.5% compared with a popular map method and 23.4% compared with a popular map method, thereby having huge service value.
It should be appreciated that embodiments of the invention may be implemented or realized by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer readable memory. The methods may be implemented in a computer program using standard programming techniques, including a non-transitory computer readable storage medium configured with a computer program, where the storage medium so configured causes a computer to operate in a specific and predefined manner, in accordance with the methods and drawings described in the specific embodiments. Each program may be implemented in a high level procedural or object oriented programming language to communicate with a computer system. However, the program(s) can be implemented in assembly or machine language, if desired. In any case, the language may be a compiled or interpreted language. Furthermore, the program can be run on a programmed application specific integrated circuit for this purpose.
Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes (or variations and/or combinations thereof) described herein may be performed under control of one or more computer systems configured with executable instructions, and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications), by hardware, or combinations thereof, collectively executing on one or more processors. The computer program includes a plurality of instructions executable by one or more processors.
Further, the method may be implemented in any type of computing platform operatively connected to a suitable computing platform, including, but not limited to, a personal computer, mini-computer, mainframe, workstation, network or distributed computing environment, separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, and so forth. Aspects of the invention may be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, optical read and/or write storage medium, RAM, ROM, etc., such that it is readable by a programmable computer, which when read by a computer, is operable to configure and operate the computer to perform the processes described herein. Further, the machine readable code, or portions thereof, may be transmitted over a wired or wireless network. When such media includes instructions or programs that, in conjunction with a microprocessor or other data processor, implement the above steps, the invention herein includes these and other different types of non-transitory computer-readable storage media. The invention also includes the computer itself when programmed according to the methods and techniques of the invention.
A computer program can be applied to the input data to perform the functions herein to convert the input data to generate output data that is stored to the non-volatile memory. The output information may also be applied to one or more output devices such as a display. In a preferred embodiment of the invention, the transformed data represents physical and tangible objects, including specific visual depictions of physical and tangible objects produced on a display.
The present invention is not limited to the above embodiments, but can be modified, equivalent, improved, etc. by the same means to achieve the technical effects of the present invention without departing from the spirit and principle of the present invention. Various modifications and variations are possible in the technical solution and/or in the embodiments within the scope of the invention.
Claims (10)
1. The system for distributing the game micro-terminal map comprises a data access unit, a time sequence training unit and a map prediction unit, and is characterized in that:
the data access unit is used for acquiring user data uploaded to the game server by the game client; the data access unit obtains user data uploaded to the game server by the game client, wherein the user data comprises user attribute data and user behavior data, and the user attribute data and the user behavior data are: the user attribute data are superior attribute data to which the game roles belong, wherein each micro-terminal game comprises a plurality of superior attribute data, and each superior game attribute comprises a plurality of game roles; the user behavior data comprises a map id list accessed in the process of logging in and logging out of the game once, wherein the map id list is arranged in the access sequence;
the time sequence training unit is used for constructing training samples according to the user data, and further, creating a time sequence prediction model based on the training samples and the time recurrent neural network;
and the prediction unit is used for acquiring the map sequence which is logged in and out of the period of the game by the user, calculating the sequence number of the map which is about to be accessed by the user by using the time sequence prediction model, and executing the loading in advance.
2. The system for distributing game micro-end maps according to claim 1, wherein the time sequence training unit specifically comprises:
the training data module is used for carrying out grouping processing on a plurality of users in the game according to at least two superior attributes to obtain a plurality of user groups, wherein each user group comprises a plurality of users and a map access sequence corresponding to the plurality of users;
and the time sequence model module is used for inputting the map sequence accessed by each user, and training the input map sequence by using the time sequence recurrent neural network to obtain a time sequence model for predicting the map to be accessed by the user.
3. The system for distributing a game micro-end map according to claim 2, wherein the timing model module is further configured to perform the steps of:
taking the map sequence accessed by each user as input and recording as
，/>
Id number representing map;
constructing hidden layer state with time sequence information and recording as
；
Based on hidden layer state
And input->
And the time sequence is dependent for a long time due to the state of the upper hidden layer, a mapping function is constructed to output an estimated map value, and the estimated map id value is obtained and recorded as ++>
。
4. The system for distributing a game micro-end map of claim 3, wherein the mapping function further comprises:
the hidden layer state
By mapping function->
According to hidden layer status->
And input +.>
Generating the calculation formula of
Output->
By mapping function parameters->
And the current hidden layer state->
Generating, the calculation formula is->
；
The mapping function
The method comprises the following steps: />
For mapping function->
Wherein->
For the output unit->
For input unit->
Is a forgetting unit, wherein C t-1 Is the information learned at time t-1.
5. The system for distributing a game micro-end map according to claim 4, wherein the output unit specifically comprises:
for conditionally performing corresponding output according to the input signal and hidden layer state, the calculation formula is
Specifically:
construction using Sigmoid layer
Information outputted in->
Will->
Mapping between (0, 1); />
Using a tanh layer handle
Regularized to between-1 and 1;
and multiplying the two types of information weights in the steps, and outputting a result.
6. The system for distributing a game micro-end map according to claim 4, wherein the input unit specifically comprises:
conditionally determining the value of the state of the inner hidden layer to be updated from the input signal, in particular by the tanh layer for generating new candidate values
(-1, 1) with a formula of +.>
；
Information learned at time t-1
And->
Multiplying and deleting redundant information;
adding
The new content to be added is updated as a unit, and its calculation formula is +>
。
7. The system for distributing a game micro-end map according to claim 4, wherein the forgetting unit specifically comprises:
for conditional screening of input and/or output information, in particular based on the output of the last moment
And the current input +.>
To generate a vector 0 to 1 +.>
And add->
Deciding whether to let the last learned information +.>
A pass or partial pass, wherein 0 indicates that no information is allowed to pass, 1 indicates that all information is allowed to pass, and the calculation formula is:
8. a system for distributing game micro-end maps according to claim 2 or 3, said time sequence training unit comprising a model training module, characterized in that:
the method is used for training a model by using a Back Propagation (BPTT) algorithm with time, wherein the BPTT algorithm is a training algorithm aiming at a circulating layer, and a mapping function can be obtained after training
Is a hidden layer state parameter +.>
。
9. The system for distributing a game micro-end map according to claim 1, wherein the prediction unit specifically comprises:
the model selection module is used for judging the basic attributes of the users accessing the micro-terminals, judging the superior attributes of the users according to the basic attributes of the users, and further selecting the corresponding prediction models;
model prediction module for inputting map sequence
The time sequence model is according to->
Calculating each map candidate +.>
And selecting the map id with the highest probability as a prediction result.
10. A method of distributing a game micro-end map, the method comprising the steps of:
acquiring user data uploaded to a server by a game client; the method comprises the steps of obtaining user data uploaded to a game server by a game client side, wherein the user data comprises user attribute data and user behavior data, and the user attribute data comprises: the user attribute data are superior attribute data to which the game roles belong, wherein each micro-terminal game comprises a plurality of superior attribute data, and each superior game attribute comprises a plurality of game roles; the user behavior data comprises a map id list accessed in the process of logging in and logging out of the game once, wherein the map id list is arranged in the access sequence;
constructing a training sample according to the user data, and further, creating a time sequence prediction model based on the training sample and the time recurrent neural network;
and acquiring a map sequence which is logged in the period of the user logging in and out of the game, calculating a sequence number of the map to be accessed by the user by using a timing prediction model, and executing early loading.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910032178.7A CN109925718B (en) | 2019-01-14 | 2019-01-14 | System and method for distributing game micro-terminal map |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910032178.7A CN109925718B (en) | 2019-01-14 | 2019-01-14 | System and method for distributing game micro-terminal map |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109925718A CN109925718A (en) | 2019-06-25 |
| CN109925718B true CN109925718B (en) | 2023-04-28 |
Family
ID=66984921
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910032178.7A Active CN109925718B (en) | 2019-01-14 | 2019-01-14 | System and method for distributing game micro-terminal map |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109925718B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110841280A (en) * | 2019-11-06 | 2020-02-28 | 福建天晴在线互动科技有限公司 | Method for online dynamic loading and checking of resources of micro-terminal |
| CN111729305B (en) * | 2020-06-23 | 2024-02-09 | 网易（杭州）网络有限公司 | Map scene preloading method, model training method, device and storage medium |
| CN112947969B (en) * | 2021-03-10 | 2022-12-16 | 支付宝(杭州)信息技术有限公司 | Page off-screen rendering method, device, equipment and readable medium |
| CN114791893B (en) * | 2021-12-15 | 2023-05-09 | 许磊 | Serialization system for random data access |
| CN114912082B (en) * | 2021-12-15 | 2023-05-09 | 许磊 | General computing task collaboration system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101358589A (en) * | 2003-11-24 | 2009-02-04 | 麦克罗斯塔克公司 | Microvalve device suitable for controlling a variable displacement compressor |
| CN104503784A (en) * | 2014-12-16 | 2015-04-08 | 珠海金山网络游戏科技有限公司 | Method and system for controlling micro-client downloading by using script |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103997544B (en) * | 2014-06-12 | 2017-08-25 | 网易（杭州）网络有限公司 | A kind of method and apparatus of resource downloading |
| CN107198884B (en) * | 2017-05-26 | 2021-05-11 | 徐州天草网络科技有限公司 | Method for efficiently loading game scenes |
| CN108764090B (en) * | 2018-05-18 | 2022-05-17 | 腾讯大地通途（北京）科技有限公司 | Regional transaction determination method, device, server and storage medium |
-
2019
- 2019-01-14 CN CN201910032178.7A patent/CN109925718B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101358589A (en) * | 2003-11-24 | 2009-02-04 | 麦克罗斯塔克公司 | Microvalve device suitable for controlling a variable displacement compressor |
| CN104503784A (en) * | 2014-12-16 | 2015-04-08 | 珠海金山网络游戏科技有限公司 | Method and system for controlling micro-client downloading by using script |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109925718A (en) | 2019-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109925718B (en) | System and method for distributing game micro-terminal map | |
| US20180349757A1 (en) | Learning service providing apparatus | |
| Fan et al. | Watching a small portion could be as good as watching all: Towards efficient video classification | |
| Espeholt et al. | Impala: Scalable distributed deep-rl with importance weighted actor-learner architectures | |
| US10817599B2 (en) | Application execution control utilizing ensemble machine learning for discernment | |
| US20220176248A1 (en) | Information processing method and apparatus, computer readable storage medium, and electronic device | |
| US20200242736A1 (en) | Method for few-shot unsupervised image-to-image translation | |
| CN112052948B (en) | Network model compression method and device, storage medium and electronic equipment | |
| CN111542841A (en) | System and method for content identification | |
| CN114616577A (en) | Identifying optimal weights to improve prediction accuracy in machine learning techniques | |
| CN112149709A (en) | Unsupervised classification of game play video using machine learning models | |
| KR102293791B1 (en) | Electronic device, method, and computer readable medium for simulation of semiconductor device | |
| US11902043B2 (en) | Self-learning home system and framework for autonomous home operation | |
| Solomes et al. | Efficient bird sound detection on the bela embedded system | |
| CN114925938A (en) | Electric energy meter running state prediction method and device based on self-adaptive SVM model | |
| Szot et al. | Large language models as generalizable policies for embodied tasks | |
| CN114556331A (en) | New frame for less-lens time action positioning | |
| CN109272165A (en) | Register probability predictor method, device, storage medium and electronic equipment | |
| CN112245934A (en) | Data analysis method, device and equipment for virtual resources in virtual scene application | |
| Wong et al. | State-space decomposition for reinforcement learning | |
| CN112148274A (en) | Method, system, article of manufacture, and apparatus for improving code characteristics | |
| KR102644593B1 (en) | An AI differentiation based HW-optimized Intelligent Software Development Tools for Developing Intelligent Devices | |
| JP7466702B2 (en) | Interpretable Imitation Learning by Discovering Prototype Options | |
| CN113191527A (en) | Prediction method and device for population prediction based on prediction model | |
| US11983240B2 (en) | Meta few-shot class incremental learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 519080 Room 102, 202, 302, 402 Room 327 Room 102, 202 Room 329 Room 302, 325 Qiandao Ring Road, Tangjiawan Town, High-tech Zone, Zhuhai City, Guangdong Province Applicant after: Zhuhai Jinshan Digital Network Technology Co.,Ltd. Address before: 519000 building 3, Jinshan Software Park, 325 Qiandao Ring Road, Xiangzhou District, Zhuhai City, Guangdong Province Applicant before: ZHUHAI KINGSOFT ONLINE GAME TECHNOLOGY Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |