CN109739959B - Method and device used in topic association calculation - Google Patents
Method and device used in topic association calculation Download PDFInfo
- Publication number
- CN109739959B CN109739959B CN201811457347.3A CN201811457347A CN109739959B CN 109739959 B CN109739959 B CN 109739959B CN 201811457347 A CN201811457347 A CN 201811457347A CN 109739959 B CN109739959 B CN 109739959B
- Authority
- CN
- China
- Prior art keywords
- topic
- iteration
- population
- individuals
- topics
- 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
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present disclosure relates to a method and apparatus for use in topic association computation. The method comprises the following steps: obtaining a population for iteration in an SCA algorithm by distributing random weights to a plurality of topics to be selected, wherein each topic to be selected has a corresponding weight component in each individual of the population; iteratively updating the individuals in the population by using an SCA algorithm, wherein the fitness function used by the SCA algorithm takes the individuals as variables, and the weight component of the topic to be selected is corrected according to the statistical quantity of the relevance between the topic to be selected and the concerned topic in the analytical formula of the fitness function; and outputting the population optimal solution individuals after the iteration of the SCA algorithm is finished as the relevance degrees of the concerned topics and the plurality of topics to be selected. The optimal weight of the optimal individual can reach the optimal weight which is adaptive to the statistical number of the characterization relevance, and the iteration process of the SCA algorithm has the characteristic of high efficiency, so that the topic relevance calculation is efficiently carried out.
Description
Technical Field
The present disclosure relates to the field of data mining, and in particular, to a method and an apparatus for use in topic association calculation.
Background
At present, the knowledge reserve can be improved only by continuous learning in the knowledge explosion age. To facilitate learning, many enterprises have built knowledge portals. The user can log in a knowledge portal website of a company, enter a knowledge base to learn and pay attention to the interested topics. Meanwhile, the system can also recommend related topics including articles, books, industry experts and the like according to topics concerned by the user.
The core of topic association calculation is topic association degree. Currently, a cosine similarity algorithm is generally adopted to calculate topic relevance. However, the cosine similarity algorithm needs to calculate complex semantic information, such as probability distribution of feature words, to obtain word vectors, and then calculates included angles between the word vectors to obtain the association between topics, which is complex in calculation and low in efficiency.
Disclosure of Invention
In view of the above, an object of the present disclosure is to provide a method and an apparatus for topic association calculation, so as to achieve the purpose of efficiently performing the associated topic calculation.
In a first aspect of embodiments of the present disclosure, there is provided a method for use in topic association computation, the method comprising: obtaining a population for iteration in an SCA algorithm (sine cosine algorithm) by distributing random weights to a plurality of topics to be selected, wherein each topic to be selected in each individual of the population has a corresponding weight component; iteratively updating the individuals in the population by using an SCA algorithm, wherein a fitness function used by the SCA algorithm takes the individuals as variables, and in an analytical formula of the fitness function, the weight component of the topic to be selected is corrected by the relevance statistical quantity between the topic to be selected and the concerned topic; and outputting the optimal solution individuals in the population as the association degrees of the concerned topics and the plurality of topics to be selected after the iteration of the SCA algorithm is finished.
Optionally, the statistical number of the relevance between the topic to be selected and the topic concerned is the number of times that the topic to be selected and the topic concerned occur in the same document; the fitness function fit (W)k) The analytical formula (D) is as follows:
wherein, WkRepresents the kth individual in the population; | Topic | represents the number of topics that the user has paid attention to, | T | represents the number of the topics to be selected, and Topic |iRepresenting the ith topic of interest; t is tjRepresents the jth topic to be selected, wkjRepresents WkThe weight component of the jth topic to be selected, count (topic)i,tj) The value of the function return of (c) indicates the ith attentionThe number of times that the topic and the jth topic to be selected appear in the same document materials.
Optionally, the method further comprises: in each iteration of the SCA algorithm iteration, obtaining an updated dynamic elimination value, wherein the dynamic elimination value is gradually reduced along with the increase of the iteration times; in each iteration, in the case that the number of the individuals allowed to be eliminated is greater than or equal to the dynamic elimination value, eliminating the individuals allowed to be eliminated from the population in the number corresponding to the dynamic elimination value, and in the case that the number of the individuals allowed to be eliminated is smaller than the dynamic elimination value, eliminating the individuals allowed to be eliminated from the population.
Optionally, the method further comprises: in each iteration of the SCA algorithm iteration, the fitness function is stored corresponding to the function value of the population individual obtained by the current iteration; in each iteration of the SCA algorithm iteration, calculating a population average fitness value according to the function value saved in the last iteration and the number of individuals in the population; and in each iteration of the SCA algorithm iteration, comparing a function value of the fitness function corresponding to the population individuals obtained in the current iteration with the population average fitness value calculated in the current iteration, and screening out the individuals allowed to be eliminated from the population.
Optionally, the method further comprises: when one or more individuals in the population are eliminated, acquiring new individuals corresponding to the quantity of the eliminated individuals according to a new individual generation algorithm; the new individual generation algorithm comprises: obtaining random step length by random amplitude of the distance between the eliminated individual and the optimal solution individual in the current iteration according to a sine or cosine function, and obtaining the weight of a new individual by accumulating the random step length on the basis of the weight of the eliminated individual.
Optionally, the method further comprises: and recommending topics according to the relevance of the concerned topics and a plurality of topics to be selected.
Optionally, the recommending topics according to the relevance of the concerned topics and a plurality of topics to be selected comprises: and selecting one or more topics to be selected for recommendation according to the relevance of the concerned topics and the topics to be selected from high to low.
In a second aspect of embodiments of the present disclosure, there is provided an apparatus for use in topic association computation, the apparatus comprising: the weight distribution module is configured to distribute random weights to a plurality of topics to be selected to obtain a population for iteration in an SCA algorithm, wherein each topic to be selected in each individual of the population has a corresponding weight component. And the iteration module is configured to perform iterative update on the individuals in the population by using an SCA algorithm, wherein a fitness function used by the SCA algorithm takes the individuals as variables, and in an analytic expression of the fitness function, the weight component of the topic to be selected is corrected by the relevance statistic quantity between the topic to be selected and the concerned topic. And the output module is configured to output the optimal solution individuals in the population after the iteration of the SCA algorithm is finished as the association degrees of the concerned topics and the plurality of topics to be selected.
Optionally, the apparatus used in topic association calculation further comprises: a dynamic culling value acquisition module configured to acquire an updated dynamic culling value in each iteration of the SCA algorithm iteration, the dynamic culling value gradually decreasing as the number of iterations increases. And the elimination module is configured to eliminate the eliminated individuals from the population by the quantity corresponding to the dynamic elimination value in the case that the quantity of the eliminated individuals is greater than or equal to the dynamic elimination value in each iteration, and eliminate the eliminated individuals from the population in the case that the quantity of the eliminated individuals is smaller than the dynamic elimination value.
Optionally, the apparatus used in topic association calculation further comprises: and the storage module is configured to store the function value of the fitness function corresponding to the population individual obtained by the current iteration in each iteration of the SCA algorithm iteration. And the average fitness calculation module is configured to calculate a population average fitness value according to the function value saved in the last iteration and the number of individuals in the population in each iteration of the SCA algorithm iteration. And the elimination screening module is configured to screen out the individuals allowed to be eliminated from the population by comparing the function value of the fitness function corresponding to the population individuals obtained in the current iteration with the population average fitness value calculated in the current iteration in each iteration of the SCA algorithm iteration.
Optionally, the apparatus used in topic association calculation further comprises: and the new individual generation module is configured to obtain new individuals corresponding to the number of the eliminated individuals according to a new individual generation algorithm when one or more individuals in the population are eliminated. The new individual generation algorithm comprises: obtaining random step length by random amplitude of the distance between the eliminated individual and the optimal solution individual in the current iteration according to a sine or cosine function, and obtaining the weight of a new individual by accumulating the random step length on the basis of the weight of the eliminated individual.
Optionally, the apparatus used in topic association calculation further comprises: and the recommending module is configured to recommend topics according to the relevance between the concerned topics and a plurality of topics to be selected.
Optionally, the recommendation module is configured to select one or more topics to be selected for recommendation according to the relevance of the concerned topic to several topics to be selected from high to low.
In a third aspect of embodiments of the present disclosure, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, carries out the steps of a method for use in topic association calculations as described in any one of the embodiments of the first aspect of the present disclosure.
In a fourth aspect of embodiments of the present disclosure, there is provided an electronic device, comprising: a memory having a computer program stored thereon; a processor for executing the computer program in the memory to implement the steps of the method for use in topic association calculation as described in any one of the embodiments in the first aspect of the disclosure.
The technical proposal of the present disclosure obtains a population for iteration in the SCA algorithm by distributing random weights to a plurality of topics to be selected, performs iterative update by using individuals in the SCA algorithm population, in the analytic expression of the fitness function used by the algorithm, the weight component of the topic to be selected is corrected according to the statistical quantity of the relevance between the topic to be selected and the concerned topic, the process of the SCA algorithm for iteratively updating the individuals in the population is an optimizing process, so that a weight vector of a globally optimal individual can be obtained when the iteration of the algorithm is finished, the weight component of each topic to be selected of the optimal individual can reach the optimal weight component which is adaptive to the statistical quantity representing the relevance, the relevance can be accurately represented, therefore, the optimal solution individuals can be used as the relevance of the concerned topic and the plurality of topics to be selected for output. And the SCA algorithm iteration process has the characteristics of simple realization, stronger global optimization capability, small calculated amount and less time consumption, so the method and the system realize the purpose of efficiently performing topic correlation calculation.
Additional features and advantages of the disclosure will be set forth in the detailed description which follows.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure and are incorporated in and constitute a part of this specification, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure without limiting the disclosure. In the drawings:
FIG. 1 is a schematic diagram of an implementation environment shown in accordance with an exemplary embodiment.
Fig. 2 is a flow chart illustrating a method used in topic association computation according to an exemplary embodiment in the first aspect of the present disclosure.
Fig. 3 is a flow diagram illustrating a method used in topic association computation according to another exemplary embodiment in the first aspect of the present disclosure.
Fig. 4 is a graph of an exponential function shown in accordance with an exemplary embodiment in a first aspect of the present disclosure.
Fig. 5 is a flow diagram illustrating a method used in topic association computation according to yet another exemplary embodiment in the first aspect of the present disclosure.
Fig. 6 is a flow chart illustrating a method used in topic association computation according to yet another exemplary embodiment in the first aspect of the present disclosure.
Fig. 7 is a flow chart illustrating a method used in topic association computation according to yet another exemplary embodiment in the first aspect of the present disclosure.
Fig. 8 is a flow chart illustrating a method used in topic association computation according to yet another exemplary embodiment in the first aspect of the present disclosure.
Fig. 9 is a block diagram illustrating an apparatus for use in topic association computation according to an exemplary embodiment in a second aspect of the present disclosure.
Fig. 10 is a block diagram illustrating an apparatus for use in topic association computation according to another exemplary embodiment in the second aspect of the present disclosure.
FIG. 11 is a block diagram illustrating an electronic device in accordance with an example embodiment.
Detailed Description
The following detailed description of specific embodiments of the present disclosure is provided in connection with the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating the present disclosure, are given by way of illustration and explanation only, not limitation.
FIG. 1 is a schematic diagram of an implementation environment shown in accordance with an exemplary embodiment. The implementation environment includes: server 101, user terminal 102. The server 101 applies the method provided by the embodiment of the present disclosure to topic association calculation. The server 101 may receive a topic attention instruction sent by the user terminal 102 through the network, record a topic that the user has paid attention to, and calculate the association degree between the topic that the user has paid attention to and a topic to be selected by using the method used in the topic association calculation provided by the embodiment of the present disclosure. And selecting topics from the topics to be selected according to the calculated association degree to recommend the user.
The network between the server 101 and the user terminal 102 may include, but is not limited to: WiFi (Wireless Fidelity), 2G, 3G, 4G, etc. networks.
Fig. 2 is a flow chart illustrating a method used in topic association computation according to an exemplary embodiment in the first aspect of the present disclosure. The method may be applied to a server. For example, it can be applied to the server 101 shown in fig. 1. As shown in fig. 2, the method may include:
in step 210, a population for iteration in the SCA algorithm is obtained by assigning random weights to a number of topics to be selected, where each topic to be selected has a corresponding weight component in each individual of the population.
The topics of interest may include one or more topics.
For the convenience of the present disclosure, the topics of interest and the band selection topics are expressed in the form of vectors. For example, there are two topics of interest, which can be represented as a vector: topic ═ or (Topic)1,topic2). The topic to be selected can be n different topic vocabularies extracted from each blog article recently browsed by the user in the knowledge system, and the topic to be selected can be expressed by vectors as follows: t ═ T (T)1,t2,t3,...,tn)。
SCA algorithm, sine cosine algorithm. SCA is a meta-heuristic algorithm, which is a numerical optimization calculation method based on self-organization and group intelligence on sine and cosine functions. Starting with a set of random solutions, the initial population generated by assigning random weights in step 220 of the present disclosure. This stochastic solution is iteratively evaluated and optimized by an objective function, namely the fitness function described in step 220 of the present disclosure.
For example, assuming that a random weight W is assigned to each topic to be selected of the n topics to be selected, the topic to be selected weight vector may be represented as W ═ W (W ═ W)1,w2,w3,...,wn). Assuming that m random weights are distributed to each topic to be selected in n topics, generating m topic weight vectors to be selected to form an initial population with iterative algorithm, wherein each individual in the initial population is a topic weight vector to be selected:
W1＝(w11,w12,w13,...,w1n),
W2＝(w21,w22,w23,...,w2n),
Wm＝(wm1,wm2,wm3,...,wmn)
in step 220, an SCA algorithm is used to iteratively update the individuals in the population, wherein a fitness function used by the SCA algorithm takes the individuals as variables, and in an analytic expression of the fitness function, a weight component of the topic to be selected is corrected according to a statistical number of associations between the topic to be selected and the topic concerned.
In the SCA algorithm, the calculation formula for updating the individuals in the population at each iteration is as follows:
wherein the content of the first and second substances,
indicating the position of the ith individual in the search space during the t-th iteration,
representing the position of the ith individual in the search space during the t +1 th iteration, wherein alpha is a linearly decreasing function and is expressed as:
where c is a constant of 2, T is the current iteration number, T is the total iteration number, β is [0,2 π]Random number in the range, η is [0,2 ]]Random number in the range, Pi tIs the global optimum individual in the t-th iteration process, r is [0,1 ]]Random numbers within a range.
The fitness function of the SCA algorithm is defined differently when solving different problems. In the present disclosure, the definition of the fitness function is not limited as long as the fitness function takes the individuals of the population as variables, and in the analytic expression of the fitness function, the weight component of the topic to be selected is corrected by the statistical number of the relevance between the topic to be selected and the topic concerned. Thus, in the iterative process of the SCA algorithm of the present disclosure, in the analytic expression of the fitness function, the weight component of the topic to be selected is corrected by the statistical number of the relevance between the topic to be selected and the topic concerned, so that the weight vector of each individual is continuously corrected and changed, and finally a weight vector called as an optimal solution individual can be output. The weight component of each topic to be selected of the optimal individual can reach the optimal weight component which is adaptive to the relevance statistic number, and the relevance can be accurately represented. And the magnitude of each dimension weight component in the optimal solution individual weight vector is the relevance of each topic to be selected. Therefore, in the iteration process, the fitness function of the disclosure can be used for evaluating the quality of each individual weight, and the individual weight is updated according to the fitness function fit ().
For example, in an embodiment of the present disclosure, the statistical number of the association between the topic to be selected and the topic of interest is the number of times that the topic to be selected and the topic of interest occur in the same document. In the implementation environment shown in fig. 1, the server 101 may store several pieces of document data in advance.
In the analytic expression of the fitness function, the implementation manner of correcting the weight component of the topic to be selected by the statistical number of the relevance between the topic to be selected and the concerned topic is not limited. For example, in an embodiment where the statistical number of relevance is the same number of occurrences, the weight component of the topic to be selected may be corrected by multiplying the number of occurrences of the topic to be selected and the topic of interest in the same piece of document data by the weight component of the topic to be selected.
For example, the fitness function fit (W) used by the present disclosurek) The analytical formula (2) can be as follows:
wherein, WkRepresenting the populationThe kth individual; | Topic | represents the number of topics that the user has paid attention to, | T | represents the number of the topics to be selected, and Topic |iRepresenting the ith topic of interest; t is tjRepresents the jth topic to be selected, wkjRepresents WkThe weight component of the jth topic to be selected, count (topic)i,tj) The function return value of (a) indicates the number of times that the ith topic of interest and the jth topic to be selected appear in the same document materials.
Under the definition of the fitness function in the present disclosure, the smaller the value of fit () is, the higher the relevance between the weight vector of the Topic to be selected, that is, the individual and the Topic that the user has paid attention to is. When the algorithm finishes iteration, a weight vector of a global optimal individual can be obtained, the weight component of each dimension of the weight vector of the global optimal individual is the association degree of each topic to be selected, and the higher the weight component is, the higher the association degree between the topic to be selected and the topic concerned by the user is. The fitness function is simple in definition and small in calculation amount, and therefore calculation efficiency of the relevance is further improved.
In step 230, the optimal solution individuals in the population after the SCA algorithm iteration is finished are output as the association degrees of the topic of interest and the several topics to be selected.
It can be seen that, since the present disclosure obtains a population for iteration in the SCA algorithm by assigning random weights to a plurality of topics to be selected, performs iterative update using individuals in the SCA algorithm population, in the analytic expression of the fitness function used by the algorithm, the weight component of the topic to be selected is corrected according to the statistical quantity of the relevance between the topic to be selected and the concerned topic, the process of the SCA algorithm for iteratively updating the individuals in the population is an optimizing process, so that a weight vector of a globally optimal individual can be obtained when the iteration of the algorithm is finished, the weight component of each topic to be selected of the optimal individual can reach the optimal weight component which is adaptive to the statistical quantity representing the relevance, the relevance can be accurately represented, therefore, the optimal solution individuals can be used as the relevance of the concerned topic and the plurality of topics to be selected for output. The SCA algorithm iteration process has the characteristics of simple realization, stronger global optimization capability, small calculated amount and less time consumption, so that the purpose of efficiently calculating the topic association degree is realized.
Fig. 3 is a flow diagram illustrating a method used in topic association computation according to another exemplary embodiment in the first aspect of the present disclosure. The method may be applied to a server. For example, it can be applied to the server 101 shown in fig. 1. In this embodiment, further in each iteration of the SCA algorithm iteration, an updated dynamic elimination value is obtained, which gradually decreases as the number of iterations increases. In each iteration, in the case that the number of the individuals allowed to be eliminated is greater than or equal to the dynamic elimination value, eliminating the individuals allowed to be eliminated from the population in the number corresponding to the dynamic elimination value, and in the case that the number of the individuals allowed to be eliminated is smaller than the dynamic elimination value, eliminating the individuals allowed to be eliminated from the population. As shown in fig. 3, the method may include:
in step 310, a population for iteration in the SCA algorithm is obtained by assigning random weights to a plurality of topics to be selected, each topic to be selected having a corresponding weight component in each individual of the population.
In step 320, it is determined whether the current iteration count is less than a preset maximum iteration count.
In step 321, if the current iteration number is less than the preset maximum iteration number, an updated dynamic elimination value is obtained, the number of eliminated individuals is cleared, and the individuals of the population are traversed.
For example, the updated dynamic culling value Num can be calculated according to the following formuladynamic：
Wherein, NumdynamicIs a dynamic elimination value]Represents a rounding function which can be rounding up or rounding down, n is the number of population individuals, tcurAnd tmaxRespectively the current iteration number and the preset maximum iteration number. The exponential function y ═ a as shown in fig. 4x(a > 1) graph, the exponential function with e as the base is a>1, so that the formula corresponds to<The function value is monotonously increased when 0 is adopted. In the above formula of the present disclosure the index of e is
Is a negative number gradually changing from-1 to 0, and
will continuously change from small to big and to 1, so
Is a value less than 1 and decreasing from large to small and gradually changing to 0. Thus, Num can be guaranteeddynamicThe value of (c) gradually decreases as the number of iterations increases.
In step 322, the currently traversed individuals are updated. If the fitness function is better than the fitness function before updating corresponding to the updated function value of the currently traversed individual, the updated individual is reserved, otherwise, the individual before updating is reserved.
In step 323, it is determined whether the currently traversed individual is allowed to be eliminated. If it is determined that the currently traversed individual is not allowed to be eliminated, return is made to step 325.
In step 324, if the currently traversed individual is allowed to be eliminated and the number of eliminated individuals is less than the dynamic elimination value, the individual is eliminated from the population.
When the individuals are eliminated, the number of the eliminated individuals in the current iteration is correspondingly increased by one.
In step 325, it is determined whether there are more individuals that have not been traversed.
In step 326, if there are more individuals that have not been traversed, the traversal continues and returns to step 322.
In step 327, if the traversal has been completed, one is added to the current number of iterations, returning to step 320.
In step 330, if the current iteration number is greater than or equal to the preset maximum iteration number, the optimal solution individuals in the population in the last iteration are used as the association degrees of the concerned topic and the multiple topics to be selected for output.
In this embodiment, a dynamic elimination strategy based on an optimization process is adopted. The dynamic elimination strategy is to eliminate a part of individuals allowed to be eliminated from the population according to the dynamic elimination value updated in each iteration in the iteration process. The dynamic elimination value is adopted in the method, so that the elimination number is larger in the initial stage of population iteration, global exploration is carried out in the whole search space as far as possible in the initial stage of search so as to find a more accurate feasible solution, the value is smaller in the later stage of iteration, local exploitation is facilitated in the later stage of exploration, the accuracy of the feasible solution is improved, and the problems that a plurality of individuals are eliminated in the later stage to increase the calculated amount and slow algorithm convergence are avoided. Therefore, the dynamic elimination strategy of the embodiment conforms to the population optimization process, so that the elimination number is gradually reduced along with the iteration from the beginning, the convergence of the algorithm is ensured, and the feasible solution precision and the calculation efficiency are improved.
Fig. 5 is a flow diagram illustrating a method used in topic association computation according to yet another exemplary embodiment in the first aspect of the present disclosure. The method may be applied to a server. For example, it can be applied to the server 101 shown in fig. 1. In this embodiment, further in each iteration of the SCA algorithm iteration, a function value of the fitness function corresponding to the population individual obtained by the current iteration is saved; in each iteration of the SCA algorithm iteration, calculating a population average fitness value according to the function value saved in the last iteration and the number of individuals in the population; and in each iteration of the SCA algorithm iteration, comparing a function value of the fitness function corresponding to the population individuals obtained in the current iteration with the population average fitness value calculated in the current iteration, and screening out the individuals allowed to be eliminated from the population. As shown in fig. 5, the method may include:
in step 510, a population for iteration in the SCA algorithm is obtained by assigning random weights to a number of topics to be selected, each topic to be selected having a corresponding weight component in each individual of the population.
In step 520, it is determined whether the current iteration count is less than a preset maximum iteration count.
In step 521, if the current iteration number is less than the preset maximum iteration number, a population average fitness value is calculated according to the function value saved in the last iteration and the number of individuals in the population, and the individuals in the population are traversed.
For example, the population average fitness value avrrageFit may be calculated according to the following formula:
in step 522, the currently traversed individuals are updated. If the fitness function is better than the fitness function before updating corresponding to the updated function value of the currently traversed individual, the updated individual is reserved, otherwise, the individual before updating is reserved.
In step 523, it is determined whether the currently traversed individuals are allowed to be eliminated by comparing the function values of the fitness function corresponding to the currently traversed individuals with the population average fitness value. If the currently traversed individuals are judged not to be eliminated according to the comparison result, the step 525 is entered.
In step 524a, if it is determined that the currently traversed individuals are allowed to be eliminated according to the comparison result, it is determined whether the number of eliminated individuals in the current iteration reaches the upper limit. If the number of eliminated individuals in the current iteration has reached the upper limit, go to step 525.
In step 524b, if the number of eliminated individuals in the current iteration has not reached the upper limit, the individuals are eliminated from the population and step 525 is entered.
When the individuals are eliminated, the number of the eliminated individuals in the current iteration is correspondingly increased by one.
In step 525, the function value of the fitness function corresponding to the currently traversed individual is saved, and whether there are any traversed individuals is determined.
In step 526, if there are more individuals that have not been traversed, the traversal continues and returns to step 522.
In step 527, if the traversal has been completed, the fitness function is saved as the function value corresponding to the population individual in the current iteration, and the current iteration number is incremented by one, and the process returns to step 520.
For example, a vector FitStore may be defined to hold the function values of fitness functions corresponding to population individuals in this iteration, as follows:
in step 530, if the current iteration number is greater than or equal to the preset maximum iteration number, the optimal solution individuals in the population in the last iteration are used as the association degrees of the concerned topic and the multiple topics to be selected for output.
It will be appreciated that in the iterative process using the SCA algorithm, the individuals of the population are continually updated and the fitness function value is getting better. For individuals in the population
In other words, the position of its position in the search space after being locally updated becomes
Comparing the fitness value before and after updating, and reserving the individual with better fitness value as the fitness value
Participate in the next iteration. If the fitness value of a certain individual is better than the fitness value of the globally optimal individual, the globally optimal individual is updated. A new generation of population is generated if each individual in the population completes a local update and the next iteration begins. However, there is also a relative fitness value in the new population generatedGood and relatively poor. In this embodiment, the FitStore is updated every time the whole population iterates once, and is used as a basis for calculating the averageFit in the next iteration process. Therefore, the algorithm can use averageFit as a judgment basis from the beginning of the first iteration to decide which individuals can be eliminated. For example, in the case of optimization by maximizing the fitness value in the embodiments of the present disclosure, individuals with fitness values less than averageFit may be eliminated, and conversely, individuals with fitness values greater than averageFit may be eliminated.
Therefore, according to the embodiment, individuals with poor fitness values are screened out according to the average fitness value, the quality of the whole population is improved, and the calculation efficiency and accuracy of the relevance are improved.
Fig. 6 is a flow chart illustrating a method used in topic association computation according to yet another exemplary embodiment in the first aspect of the present disclosure. The method may be applied to a server. For example, it can be applied to the server 101 shown in fig. 1. In this embodiment, further when one or more individuals in the population are eliminated, new individuals corresponding to the number of eliminated individuals are obtained according to a new individual generation algorithm. The new individual generation algorithm comprises: obtaining random step length by random amplitude of the distance between the eliminated individual and the optimal solution individual in the current iteration according to a sine or cosine function, and obtaining the weight of a new individual by accumulating the random step length on the basis of the weight of the eliminated individual. As shown in fig. 6, the method may include:
in step 610, a population for iteration in the SCA algorithm is obtained by assigning random weights to a number of topics to be selected, each topic to be selected having a corresponding weight component in each individual of the population.
In step 620, it is determined whether the current iteration count is less than a preset maximum iteration count.
In step 621, if the current iteration number is less than the preset maximum iteration number, the individuals of the population are traversed.
In step 622, the currently traversed individuals are updated. If the fitness function is better than the fitness function before updating corresponding to the updated function value of the currently traversed individual, the updated individual is reserved, otherwise, the individual before updating is reserved.
In step 623, it is determined whether the currently traversed individual is allowed to be eliminated. If the currently traversed individual is not allowed to be eliminated, return is made to step 625.
In step 624, if the currently traversed individuals are allowed to be eliminated and the number of eliminated individuals in the current iteration does not reach the upper limit, the individuals are eliminated from the population, and new individuals corresponding to the currently eliminated individuals are obtained according to a new individual generation algorithm.
For example, a new individual may be generated based on a sine-cosine function, the formula as follows:
wherein D represents the currently traversed individual
And the weight vector of (2) and the optimal solution individual x in the current iteration*Distance between weight vectors of (a):
is a new individual to be generated, and,
is the currently rejected individual, is ∈ [ -1,1 [ ]]And p is a random number between 0 and 1.
In step 625, it is determined whether there are more individuals that have not been traversed.
In step 626, if there are more individuals not traversed, the traversal continues and returns to step 622.
In step 627, if the traversal has been completed, one is added to the current iteration number, returning to step 620.
In step 630, if the current iteration number is greater than or equal to the preset maximum iteration number, the optimal solution individuals in the population in the last iteration are used as the association degrees of the concerned topic and the multiple topics to be selected for output.
Therefore, in the embodiment, under the condition of eliminating one or more individuals, corresponding new individuals are generated, and the generation of the new individuals refers to globally optimal individuals, so that the evolution direction is ensured, the quality of the whole population is improved, and the accuracy of the association degree is further improved.
Fig. 7 is a flow chart illustrating a method used in topic association computation according to yet another exemplary embodiment in the first aspect of the present disclosure. The method may be applied to a server. For example, it can be applied to the server 101 shown in fig. 1. In this embodiment, the embodiments shown in the above figures are combined to calculate topic relevance. As shown in fig. 7, the method may include:
in step 710, a population for iteration in an SCA algorithm is obtained by assigning random weights to a number of topics to be selected, each topic to be selected having a corresponding weight component in each individual of the population.
In step 720, it is determined whether the current iteration count is less than a preset maximum iteration count.
In step 721, if the current iteration time is less than the preset maximum iteration time, a population average fitness value is calculated according to the function value saved in the last iteration and the number of individuals in the population, an updated dynamic elimination value is obtained, the number of eliminated individuals is cleared, and the individuals in the population are traversed.
In step 722, the currently traversed individuals are updated. If the fitness function is better than the fitness function before updating corresponding to the updated function value of the currently traversed individual, the updated individual is reserved, otherwise, the individual before updating is reserved.
In step 723, it is determined whether the currently traversed individuals are allowed to be eliminated by comparing the function values of the fitness function corresponding to the currently traversed individuals with the population average fitness value. If the currently traversed individuals are determined not to be eliminated according to the comparison result, the step 725 is entered.
In step 724a, if it is determined that the currently traversed individuals are allowed to be eliminated according to the comparison result, it is determined whether the number of eliminated individuals in the current iteration is smaller than the dynamic elimination value. If not, step 725 is entered.
In step 724b, if the number of eliminated individuals in the current iteration is smaller than the dynamic elimination value, the individuals are eliminated from the population, new individuals corresponding to the currently eliminated individuals are obtained according to a new individual generation algorithm, and the process goes to step 725.
When the individuals are eliminated, the number of the eliminated individuals in the current iteration is correspondingly increased by one.
In step 725, the function value of the fitness function corresponding to the currently traversed individual is saved, and whether there are any non-traversed individuals is determined.
In step 726, if there are more individuals not traversed, the traversal continues and returns to step 722.
In step 727, if the traversal is completed, the fitness function is saved corresponding to the function value of the population individual in the current iteration, one is added to the current iteration time, and the process returns to step 720.
In step 730, if the current iteration number is greater than or equal to the preset maximum iteration number, the optimal solution individuals in the population in the last iteration are used as the association degrees of the concerned topic and the multiple topics to be selected for output.
It can be seen that, in this embodiment, a certain improvement is performed on the SCA algorithm, a dynamic elimination strategy is introduced into the algorithm, the elimination number is dynamically determined based on the optimization process, the global optimization and the local optimization processes of the algorithm can be balanced by the dynamic elimination strategy, both the accuracy of feasible solution and the convergence of the algorithm can be ensured, and the inferior individuals can be eliminated by using the average fitness value as the basis for whether to eliminate an individual, so that the overall population quality is improved, new individuals are generated when the inferior individuals are eliminated, and the generation of the new individuals refers to the positions of the eliminated individuals and the positions of the optimal values of the search space, so that the capacity of local search is improved and the evolution direction is ensured.
Fig. 8 is a flow chart illustrating a method used in topic association computation according to yet another exemplary embodiment in the first aspect of the present disclosure. The method may be applied to a server. For example, it can be applied to the server 101 shown in fig. 1. In the embodiment, topic recommendation is further carried out according to the relevance of the concerned topic and a plurality of topics to be selected. As shown in fig. 8, the method may include:
in step 810, a population for iteration in an SCA algorithm is obtained by assigning random weights to a number of topics to be selected, wherein each topic to be selected has a corresponding weight component in each individual of the population.
In step 820, iteratively updating the individuals in the population by using an SCA algorithm, wherein a fitness function used by the SCA algorithm takes the individuals as variables, and in an analytic expression of the fitness function, a weight component of the topic to be selected is corrected by a statistical number of associations between the topic to be selected and the topic concerned.
In step 830, the optimal solution individuals in the population after the SCA algorithm iteration is finished are output as the association degrees of the topic of interest and the multiple topics to be selected.
In step 840, topic recommendation is performed according to the relevance of the concerned topic and a plurality of topics to be selected.
For example, one or more topics to be selected can be selected for recommendation according to the relevance of the concerned topics and a plurality of topics to be selected from high to low.
In the following, the focused topics and the topics to be selected in a certain scene are taken as examples for explanation. Let the topics that the user has focused on are "Java" and "MySql", respectively, denoted as vector Topic ═ (Java, MySql). The topic to be selected is obtained from different topic words extracted from a blog recently browsed by a user, and is expressed in a vector form:
t ═ T (java web, Hadoop, SpringBoot, JavaScript, MySql database cluster,
oracle, MySql Performance optimization, Vue, MVC, design Pattern)
Randomly generating M weight vectors for the vector T of the topic to be selected, via step 810:
W1＝(0.15，0.32，0.51，0.17，0.46，0.80，0.34，0.73，0.92，0.85)；
W2＝(0.22，0.58，0.19，0.10，0.84，0.79，0.92，0.13，0.46，0.75)；
W3＝(0.46，0.63，0.50，0.04，0.77，0.61，0.34，0.73，0.92，0.85)；……
WM＝(0.93，0.24，0.52，0.54，0.44，0.09，0.71，0.55，0.39，0.78)。
the M weight vectors are iteratively changed through step 820, and finally, the weight vector W of the globally optimal individual is output in step 830 as (0.43, 0.59, 0.64, 0.21, 0.84, 0.39, 0.93, 0.07, 0.19, 0.24), and is sorted through step 840 into (0.93, 0.84, 0.64, 0.59, 0.43, 0.39, 0.24, 0.21, 0.19, 0.07). The topic sequence set to be selected corresponding to the weight vector after the sorting is as follows: (MySql performance optimization, MySql database cluster, SpringBoot, Hadoop, JavaWeb, Oracle, design Pattern, JavaScript, MVC, VUE). If the system is set to recommend a related topic with the highest similarity to the user, the topic to be selected, namely MySql performance optimization, can be recommended to the user; if two associated topics are recommended, the topics to be selected, namely 'MySql performance optimization' and 'MySql database cluster', can be recommended to the user; by analogy, as long as there are multiple alternative topics, multiple associated topics can be recommended.
Therefore, the relevance between the concerned topic and the plurality of topics to be selected is obtained through calculation by utilizing an SCA algorithm which is simple in realization, strong in global optimization capability, small in calculation amount and short in time consumption, and the topic to be selected with the relevance meeting the requirement is selected from the plurality of topics to be selected as the relevant topic for recommendation according to the relevance between the concerned topic and the plurality of topics to be selected, so that the purpose of efficiently recommending the relevant topic can be achieved.
Fig. 9 is a block diagram illustrating an apparatus 900 for use in topic association computation according to an exemplary embodiment in the second aspect of the disclosure. The apparatus may be configured with a server. For example, the server 101 shown in fig. 1 may be configured. As shown in fig. 9, the apparatus may include: a weight assignment module 910, an iteration module 920, and an output module 930.
The weight assignment module 910 may be configured to obtain a population for iteration in the SCA algorithm by assigning random weights to a number of topics to be selected, where each topic to be selected has a corresponding weight component in each individual of the population.
The iteration module 920 may be configured to iteratively update the individuals in the population by using an SCA algorithm, where a fitness function used by the SCA algorithm uses the individuals as variables, and in an analytic expression of the fitness function, a weight component of the topic to be selected is modified by a statistical number of associations between the topic to be selected and the topic concerned.
The output module 930 may be configured to output, as the association degrees between the topic of interest and the several topics to be selected, the optimal solution individuals in the population after the SCA algorithm iteration is finished.
In this embodiment, since the weight assignment module 910 assigns random weights to a plurality of topics to be selected to obtain a population for iteration in the SCA algorithm, the iteration module 920 performs iterative update using individuals in the SCA algorithm population, in the analytic expression of the fitness function used by the algorithm, the weight component of the topic to be selected is corrected according to the statistical quantity of the relevance between the topic to be selected and the concerned topic, the process of the SCA algorithm for iteratively updating the individuals in the population is an optimizing process, so that a weight vector of a globally optimal individual can be obtained when the iteration of the algorithm is finished, the weight component of each topic to be selected of the optimal individual can reach the optimal weight component which is adaptive to the statistical quantity representing the relevance, the relevance can be accurately represented, the output module 930 may thus output the optimal solution individuals as the association degrees of the topic of interest and the several topics to be selected. And the SCA algorithm iteration process has the characteristics of simple realization, stronger global optimization capability, small calculated amount and less time consumption, so the method and the device realize the purpose of efficiently calculating the topic association degree.
Fig. 10 is a block diagram of an apparatus 1000 for use in topic association computation shown in accordance with another exemplary embodiment in the second aspect of the present disclosure. The apparatus may be configured with a server. For example, the server 101 shown in fig. 1 may be configured. As shown in fig. 10, the apparatus may optionally further include: a dynamic cull value acquisition module 940 configured to acquire an updated dynamic cull value in each iteration of the SCA algorithm iteration, the dynamic cull value gradually decreasing as the number of iterations increases. A culling module 941 configured to cull, in each iteration, a corresponding number of individuals allowed to be culled from the population in accordance with the dynamic culling value if the number of individuals allowed to be culled is equal to or greater than the dynamic culling value, and cull, in each iteration, individuals allowed to be culled from the population if the number of individuals allowed to be culled is less than the dynamic culling value. In the embodiment, due to the adoption of the dynamic elimination value, the elimination number has a larger initial value in population iteration, global exploration is carried out in the whole search space as far as possible in the initial stage of search so as to find a more accurate feasible solution, the value is smaller in the later stage of iteration, local exploitation is facilitated in the later stage of exploration, the accuracy of the feasible solution is improved, and the phenomena that a plurality of individuals are eliminated in the later stage to increase the calculated amount and slow the convergence of the algorithm are avoided. Therefore, the dynamic elimination strategy of the embodiment conforms to the population optimization process, so that the elimination number is gradually reduced along with the iteration from the beginning, the convergence of the algorithm is ensured, and the feasible solution precision and the calculation efficiency are improved.
Optionally, as shown in fig. 10, the apparatus 1000 used in topic association calculation may further include: a saving module 950 configured to save, in each iteration of the SCA algorithm iteration, a function value of the fitness function corresponding to the population individual obtained by the current iteration. And an average fitness calculation module 951 configured to calculate a population average fitness value according to the function value saved in the last iteration and the number of individuals in the population in each iteration of the SCA algorithm iteration. A culling and screening module 952 configured to, in each iteration of the SCA algorithm iteration, screen out individuals that are allowed to be culled from the population by comparing a function value of the fitness function corresponding to a population individual obtained in a current iteration with a population average fitness value calculated in the current iteration. In the embodiment, individuals with relatively poor fitness values are screened out according to the average fitness value, the quality of the whole population is improved, and the calculation efficiency and accuracy of the relevance are improved.
Optionally, as shown in fig. 10, the apparatus 1000 used in topic association calculation may further include: a new individual generation module 960 configured to, when one or more individuals in the population are eliminated, obtain new individuals corresponding to the number of eliminated individuals according to a new individual generation algorithm. The new individual generation algorithm comprises: obtaining random step length by random amplitude according to sine or cosine function on the basis of the distance between the eliminated individual and the optimal solution individual in the current iteration, and obtaining the weight of a new individual by accumulating the random step length on the basis of the weight of the eliminated individual. In the embodiment, under the condition of eliminating one or more individuals, corresponding new individuals are generated, and the generation of the new individuals refers to globally optimal individuals, so that the evolution direction is ensured, the quality of the whole population is improved, and the accuracy of the association degree is further improved.
Optionally, as shown in fig. 10, the apparatus 1000 used in topic association calculation may further include: a recommendation module 970. The recommending module 970 can be configured to recommend topics according to the relevance of the topics to be selected.
Therefore, the recommendation module 970 can recommend topics according to the relevance between the concerned topics and a plurality of topics to be selected, and the purpose of efficiently recommending relevant topics can be achieved.
With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.
Fig. 11 is a block diagram illustrating an electronic device 1100 in accordance with an example embodiment. For example, the electronic device 1100 may be provided as a server. Referring to fig. 11, electronic device 1100 includes a processor 1122, which can be one or more in number, and a memory 1132 for storing computer programs executable by processor 1122. The computer programs stored in memory 1132 may include one or more modules that each correspond to a set of instructions. Further, the processor 1122 may be configured to execute the computer program to perform the above-described method used in topic association calculations.
Additionally, the electronic device 1100 may also include a power component 1126 and a communication component 1150, the power component 1126 may be configured to perform power management of the electronic device 1100, and the communication component 1150 may be configured to enable communication, e.g., wired or wireless communication, of the electronic device 1100. In addition, the electronic device 1100 may also include an input/output (I/O) interface 1158. The electronic device 1100 may operate based on an operating system stored in memory 1132, such as Windows Server, Mac OS XTM, UnixTM, Linux, and the like.
In another exemplary embodiment, a computer readable storage medium comprising program instructions which, when executed by a processor, implement the steps of the above method for use in topic association calculation is also provided. For example, the computer-readable storage medium may be the memory 1132 described above including program instructions executable by the processor 1122 of the electronic device 1100 to perform the methods described above for use in topic association calculations.
The preferred embodiments of the present disclosure are described in detail with reference to the accompanying drawings, however, the present disclosure is not limited to the specific details of the above embodiments, and various simple modifications may be made to the technical solution of the present disclosure within the technical idea of the present disclosure, and these simple modifications all belong to the protection scope of the present disclosure.
It should be noted that the various features described in the above embodiments may be combined in any suitable manner without departing from the scope of the invention. In order to avoid unnecessary repetition, various possible combinations will not be separately described in this disclosure. In addition, any combination of various embodiments of the present disclosure may be made, and the same should be considered as the disclosure of the present disclosure, as long as it does not depart from the spirit of the present disclosure.
Claims (10)
1. A method for use in topic association computation, comprising:
the method comprises the steps that random weights are distributed to a plurality of topics to be selected to obtain a population for iteration in a Sine and Cosine Algorithm (SCA), wherein each topic to be selected in each individual of the population has a corresponding weight component;
iteratively updating the individuals in the population by using the SCA algorithm, wherein a fitness function used by the SCA algorithm takes the individuals as variables, and in an analytical formula of the fitness function, the weight component of the topic to be selected is corrected by the relevance statistical quantity between the topic to be selected and the concerned topic;
and outputting the optimal solution individuals in the population as the association degrees of the concerned topics and the plurality of topics to be selected after the iteration of the SCA algorithm is finished.
2. The method according to claim 1, wherein the statistical number of the relevance between the topic to be selected and the topic of interest is the number of times that the topic to be selected and the topic of interest occur in the same document material;
the fitness function fit (W)k) The analytical formula (D) is as follows:
wherein, WkRepresents the kth individual in the population; | Topic | represents the number of topics that the user has paid attention to, | T | represents the number of the topics to be selected, and Topic |iRepresenting the ith topic of interest; t is tjRepresents the jth topic to be selected, wkjRepresents WkThe weight component of the jth topic to be selected, count (topic)i,tj) The function return value of (a) indicates the number of times that the ith topic of interest and the jth topic to be selected appear in the same document materials.
3. The method of claim 1, further comprising:
in each iteration of the SCA algorithm iteration, obtaining an updated dynamic elimination value, wherein the dynamic elimination value is gradually reduced along with the increase of the iteration times;
in each iteration, in the case that the number of the individuals allowed to be eliminated is greater than or equal to the dynamic elimination value, eliminating the individuals allowed to be eliminated from the population in the number corresponding to the dynamic elimination value, and in the case that the number of the individuals allowed to be eliminated is smaller than the dynamic elimination value, eliminating the individuals allowed to be eliminated from the population.
4. The method according to any one of claims 1-3, further comprising:
in each iteration of the SCA algorithm iteration, the fitness function is stored corresponding to the function value of the population individual obtained by the current iteration;
in each iteration of the SCA algorithm iteration, calculating a population average fitness value according to the function value saved in the last iteration and the number of individuals in the population;
and in each iteration of the SCA algorithm iteration, comparing a function value of the fitness function corresponding to the population individuals obtained in the current iteration with the population average fitness value calculated in the current iteration, and screening out the individuals allowed to be eliminated from the population.
5. The method according to any one of claims 1-3, further comprising:
when one or more individuals in the population are eliminated, acquiring new individuals corresponding to the quantity of the eliminated individuals according to a new individual generation algorithm;
the new individual generation algorithm comprises: obtaining random step length by random amplitude of the distance between the eliminated individual and the optimal solution individual in the current iteration according to a sine or cosine function, and obtaining the weight of a new individual by accumulating the random step length on the basis of the weight of the eliminated individual.
6. The method of claim 1, further comprising:
and recommending topics according to the relevance of the concerned topics and a plurality of topics to be selected.
7. The method of claim 6, wherein the topic recommendation according to the relevance of the topic of interest to a number of topics to be selected comprises: and selecting one or more topics to be selected for recommendation according to the relevance of the concerned topics and the topics to be selected from high to low.
8. An apparatus for use in topic association computation, comprising:
the weight distribution module is configured to distribute random weights to a plurality of topics to be selected to obtain a population for iteration in a Sine and Cosine Algorithm (SCA), wherein each topic to be selected in each individual of the population has a corresponding weight component;
the iteration module is configured to perform iterative updating on the individuals in the population by using the SCA algorithm, wherein a fitness function used by the SCA algorithm takes the individuals as variables, and in an analytic expression of the fitness function, the weight component of the topic to be selected is corrected by the relevance statistical quantity between the topic to be selected and the concerned topic;
and the output module is configured to output the optimal solution individuals in the population after the iteration of the SCA algorithm is finished as the association degrees of the concerned topics and the plurality of topics to be selected.
9. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 7.
10. An electronic device, comprising:
a memory having a computer program stored thereon;
a processor for executing the computer program in the memory to carry out the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811457347.3A CN109739959B (en) | 2018-11-30 | 2018-11-30 | Method and device used in topic association calculation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811457347.3A CN109739959B (en) | 2018-11-30 | 2018-11-30 | Method and device used in topic association calculation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109739959A CN109739959A (en) | 2019-05-10 |
| CN109739959B true CN109739959B (en) | 2021-02-26 |
Family
ID=66358390
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811457347.3A Active CN109739959B (en) | 2018-11-30 | 2018-11-30 | Method and device used in topic association calculation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109739959B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6581056B1 (en) * | 1996-06-27 | 2003-06-17 | Xerox Corporation | Information retrieval system providing secondary content analysis on collections of information objects |
| CN102663100A (en) * | 2012-04-13 | 2012-09-12 | 西安电子科技大学 | Two-stage hybrid particle swarm optimization clustering method |
| CN105005820A (en) * | 2015-04-03 | 2015-10-28 | 北京理工大学 | Target assignment optimizing method based on particle swarm algorithm of population explosion |
| CN106682682A (en) * | 2016-10-20 | 2017-05-17 | 北京工业大学 | Method for optimizing support vector machine based on Particle Swarm Optimization |
| WO2018109145A1 (en) * | 2016-12-16 | 2018-06-21 | Koninklijke Philips N.V. | Context sensitive help function for medical device monitor/defibrillators |
-
2018
- 2018-11-30 CN CN201811457347.3A patent/CN109739959B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6581056B1 (en) * | 1996-06-27 | 2003-06-17 | Xerox Corporation | Information retrieval system providing secondary content analysis on collections of information objects |
| CN102663100A (en) * | 2012-04-13 | 2012-09-12 | 西安电子科技大学 | Two-stage hybrid particle swarm optimization clustering method |
| CN105005820A (en) * | 2015-04-03 | 2015-10-28 | 北京理工大学 | Target assignment optimizing method based on particle swarm algorithm of population explosion |
| CN106682682A (en) * | 2016-10-20 | 2017-05-17 | 北京工业大学 | Method for optimizing support vector machine based on Particle Swarm Optimization |
| WO2018109145A1 (en) * | 2016-12-16 | 2018-06-21 | Koninklijke Philips N.V. | Context sensitive help function for medical device monitor/defibrillators |
Non-Patent Citations (1)
| Title |
|---|
| "Hybridizing sine cosine algorithm with multi-orthogonal search strategy for engineering design problems";Rizk M等;《Journal of Computational Design and Engineering》;20170818;第5卷(第2期);249-273页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109739959A (en) | 2019-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109902708B (en) | Recommendation model training method and related device | |
| Ning et al. | Constrained multi-objective optimization using constrained non-dominated sorting combined with an improved hybrid multi-objective evolutionary algorithm | |
| EP2950226A1 (en) | New heuristic for optimizing non-convex function for learning to rank | |
| CN112511342B (en) | Network slicing method and device, electronic equipment and storage medium | |
| CN110378434A (en) | Training method, recommended method, device and the electronic equipment of clicking rate prediction model | |
| CN111310053A (en) | Information recommendation method, device, equipment and storage medium | |
| CN111506820B (en) | Recommendation model, recommendation method, recommendation device, recommendation equipment and recommendation storage medium | |
| JP2007317068A (en) | Recommending device and recommending system | |
| CN106062732A (en) | Data processing system, calculation node and data processing method | |
| CN108470075A (en) | A kind of socialization recommendation method of sequencing-oriented prediction | |
| CN110472016B (en) | Article recommendation method and device, electronic equipment and storage medium | |
| CN113536097A (en) | Recommendation method and device based on automatic feature grouping | |
| GB2599137A (en) | Method and apparatus for neural architecture search | |
| Liu et al. | Evolving graph construction for successive recommendation in event-based social networks | |
| CN114822533B (en) | Voice interaction method, model training method, electronic device and storage medium | |
| CN116050540A (en) | Self-adaptive federal edge learning method based on joint bi-dimensional user scheduling | |
| CN109739959B (en) | Method and device used in topic association calculation | |
| US9111213B2 (en) | Method for constructing a tree of linear classifiers to predict a quantitative variable | |
| CN108550019A (en) | A kind of resume selection method and device | |
| CN107870862A (en) | Construction method, traversal method of testing and the computing device of new control forecast model | |
| CN110705889A (en) | Enterprise screening method, device, equipment and storage medium | |
| CN108053260A (en) | A kind of method and system that extending user is determined according to statistics interest-degree | |
| CN108182600A (en) | A kind of method and system that extending user is determined according to weighted calculation | |
| CN108038739A (en) | A kind of method and system that extending user is determined according to the statistics degree of association | |
| CN113541993B (en) | Network evaluation method and device, network index processing method, device and medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |