CN107682117A - A kind of design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm - Google Patents
A kind of design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm Download PDFInfo
- Publication number
- CN107682117A CN107682117A CN201710815343.7A CN201710815343A CN107682117A CN 107682117 A CN107682117 A CN 107682117A CN 201710815343 A CN201710815343 A CN 201710815343A CN 107682117 A CN107682117 A CN 107682117A
- Authority
- CN
- China
- Prior art keywords
- chicken
- individual
- degree
- optimization
- chicken group
- 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.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/0001—Systems modifying transmission characteristics according to link quality, e.g. power backoff
- H04L1/0009—Systems modifying transmission characteristics according to link quality, e.g. power backoff by adapting the channel coding
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
- H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
- H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0056—Systems characterized by the type of code used
- H04L1/0057—Block codes
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0056—Systems characterized by the type of code used
- H04L1/0061—Error detection codes
- H04L1/0063—Single parity check
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0056—Systems characterized by the type of code used
- H04L1/0064—Concatenated codes
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
A kind of design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm, specifically includes following steps:The Probability Forms of conjugation distribution and the feature of swarm optimization algorithm, degree of foundation distribution optimization problem model;Establish optimization object function;Design improves chicken colony optimization algorithm;Computing is iterated, exports globally optimal solution, LT codes is obtained more excellent decoding performance so as to obtain preferably degree distribution.Required average number of coded symbols is minimised as optimization aim when the present invention is with successfully decoded, and optimization problem is solved using improved chicken group algorithm, and more excellent degree distribution can be obtained while algorithm the convergence speed is ensured makes LT code decoding performances more excellent.
Description
Technical field
The invention belongs to mobile radio telecommunications technical field, is related to a kind of based on the long LT of short code for improving chicken colony optimization algorithm
The design method of code degree distribution.
Background technology
As a kind of typical forward error correction coding scheme, no code check code is to overcome time varying channel and the unknown letter of status information
The defects of road is present provides a good technological means.With the digital fountain code of representative no code check code
The it is proposed of (Digital Fountain Code, DFC), the development and application of DFC technologies progressively receive the extensive of various circles of society
Concern, and have become study hotspot.DFC basic thought is:Assuming that several raw data symbols are encoded, number
According to transmitting terminal coded identification is then endlessly sent as fountain, export the coded identification data stream of random length, connect
As long as receiving end can receive sufficient amount of coded identification just can successfully recover all raw data symbols with maximum probability.
2002, Luby proposed the specific implementation LT (Luby Transform) of the first practicable digital fountain code
Code.Because LT codes have the characteristics that good autgmentability, coding and decoding process are simple, calculating degree is low, lead in broadcast communication, deep space
The fields such as letter, wireless sensor network, high in the clouds storage have been obtained for extensive use.
In LT codes, transmitting terminal is distributed according to degree produces coded identification, and receiving terminal recovers also according to degree distributed intelligence
Raw data symbols.Therefore during LT coding and decoding, its performance of Degree distributions plays vital effect.Identical degree
It is distributed in different application occasion and different degrees of influence is produced to the decoding performance of LT codes, wherein compares typically mutually unison
Decoding performance under the application scenarios of different code length data is distributed in differ greatly.With reference to information source data code length characteristic, reasonable selection
There is important theoretical research value in the decoding performance for improving LT codes with optimization Degree distributions, related research has become nothing
The hot issue of the line communications field.
At present, researcher is applied to intelligent optimization algorithm to solve LT code degree distribution optimization design problems, example
As Deng Zaihui et al. use particle swarm optimization algorithm (the Particle Swarm Optimization with gradient
Gradient, PSO-G) design is optimized to the distribution of the degree of LT codes, using the algorithm to optimize to a certain extent can be with
Obtain preferably degree distribution, but due to particle cluster algorithm in optimization process population diversity easy to be lost, therefore the optimization method
The defects of in the presence of locally optimal solution and slow convergence rate is easily trapped into.2014, MengXianbing et al. proposed a kind of new
Bionic Algorithm:Chicken colony optimization algorithm (Chicken Swarm Optimization, CSO), the optimized algorithm is to simulate chicken
A kind of random algorithm of group's foraging behavior and grade classification as optimization basis, compared with other intelligent optimization algorithms, its advantage
Mainly have:(1) location Update Strategy expands effective search space, is advantageous to try to achieve global optimum;(2) by establishing grade
Order keeps population diversity, can cooperate with and look for food between different chicken groups, make algorithm be not easy to be absorbed in local extremum.Simulation result table
Bright, CSO can obtain more preferable low optimization accuracy and speed of searching optimization in complicated optimum problem solution.Change at present on CSO correlations
The work entered has become the focus of researcher.Based on background above, this patent plays the advantage of chicken group's algorithm, there is provided
A kind of design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm (Enhance CSO, ECSO).
The content of the invention
It is an object of the present invention to provide a kind of design based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm
Method.The present invention can find more excellent degree distribution for the long LT codes of short code and provide an effective solution, with other algorithms
Compare, more excellent LT code performances and faster convergence rate can be obtained.The particular content of the present invention comprises the following steps:
Step 1:According to the form of probability of degree distribution and the position feature of improvement chicken colony optimization algorithm, the distribution of degree of foundation
Optimization problem model, optimized using a degree distribution as population at individual position, it is assumed that raw data symbols quantity is k, degree
The quantity comprising probable value is D in distribution, and angle value is that the raw data symbols quantity needed for a coded identification is d, maximum
Angle value is expressed as dmax, produce the degree distribution probability value that angle value is d and be represented by Ω (d), then a complete degree distributed architecture can
It is expressed as Ω={ Ω (d1), Ω (d2), K, Ω (dmax), and meetWherein 0 < Ω (d) < 1.
Step 2:The degree distribution optimization problem model carried with reference to step 1, the related notion of chicken colony optimization algorithm is introduced, it is fixed
The population scale of adopted chicken group is N, position vector x, and the size of search space is expressed as D, and using chicken group body position as excellent
Change variable, because individual body position is corresponding with degree distributed architecture, so the position vector of i-th of individual is represented by xi=
[pI, 1, pI, 2, K, pI, j, K, pI, D], pI, jParameter probable value in ∈ (0,1) degree of a representation distributed architecture,Its
In have xi=Ωi, pI, D=Ωi(dmax)。
Step 3:With reference to chicken group's related notion is carried in the degree distribution optimization model and step 2 that step 1 is carried, according to code length
The concrete condition of size, select suitable degree distribution form to optimize, if code length is slightly shorter, orphan's distribution form can be directly selected
Optimize, now individual body position is the probability of the angle value from 1 to k, therefore has dmax=k, if code length is slightly long, due to chicken, group is excellent
Change algorithm search space can become big with the increase of code length, and search space scope become conference complicate the issue so as to
Effect of optimization is influenceed, therefore degree of rarefication distributed architecture form can be selected to optimize with the size of less search space, such as
Angle value size only considers the index that power is 2, and the angle value of other positions is defaulted as 0, i.e. dj=2j, j=1,2, K, D, meet simultaneously
dmax=2D, dmax< k.
Step 4:Required average number of coded symbols is minimized as optimization aim during using successfully decoded, i.e., successfully recovers
The number of coded symbols received during all raw data symbols is more few better, and it is to translate the coding needed for a code word to define T
The quantity of symbol, then minimizing average coding symbol number can be embodied by decoding overheads, and decoding overheads are designated asWherein E represents it is expected.In the Optimized model of proposition, calculated to optimize suitable for colony intelligence
The characteristic of method, the average value of coded identification is selected to replace the decoding overheads target good and bad as the distribution of measurement degree in optimization process
Function, it is represented byWherein f represents fitness value, and m is sample size, i.e., for identical degree point
Cloth carries out the number of encoding and decoding, and T (r) represents average number of coded symbols required during successfully decoded in r-th of sample, therefore
Optimal solution is found in distribution optimization problem is spent can be defined as OS(Ω)=arg (min (f)).
Step 5:Improved chicken colony optimization algorithm is designed, then the degree distribution optimization problem in step 4 is solved, it is raw
Into certain scale N chicken group, a kind of degree distributed architecture of the corresponding optimization problem of each individual in chicken group, to all individual in chicken group
Body carries out position initialization, and using initial position as initial degree distribution ΩI, define cock, hen, chicken and hen in population
The quantitative proportion of mother is respectively RN, HN, CN, MN, and all individual fitness of chicken group are calculated according to the object function in step 4
Value, chicken group's hierarchy is established according to the quality for calculating gained fitness value, that is, determines character species of the individual in chicken group, knot
Variety classes individual proportion is closed, selects several to possess the individual of adaptive optimal control angle value as cock, selects several tools
Each body for having worst fitness value is used as hen as chicken, remaining other individuals.According to cock quantity to all individuals of chicken group
It is grouped, each packet includes a cock, some hens and a small amount of chicken, and hen selection belongs to point nearest from it
The mother-child relationship (MCR) of group, hen and chicken is established using random fashion.
Step 6:Location status is carried out more to whole cocks in chicken group using amplitude constraint strategy, limit dimensions updating mechanism
Newly.Wherein, amplitude constraint strategy can be described as:It is individual in chicken group it can be seen from the Optimized model carried in step 1 and step 2
Each dimension of body position is of equal value actually with the probable value in degree distributed architecture, then cock location updating scope should be controlled
System is in the range of 0 and 1, and otherwise renewal is invalid, in view of directly using cock more new formula in typical chicken group algorithm to chicken group
In whole cocks carry out location updating easily produce above-mentioned renewal inefficiency, amplitude constraint plan is used to cock location updating
Slightly to control position changes in amplitude, then cock position updating process can be described as xI, j(t+1)=xI, j(t)(1+c1Randn
(0, σ2)), wherein t represents iterations, xI, jAnd x (t+1)I, j(t) i-th of individual is illustrated respectively in the t times and t+1
The positional value of jth dimension during secondary iteration, c1To limit the controlling elements of rangeability, Randn (0, σ2) represent to be desired for 0
Variance is σ2Gaussian Profile.Limit dimensions updating mechanism can be described as:In chicken group's algorithm, cock, which has, preferentially searches out food
Right, then it has bigger food search space, therefore the validity of cock location updating largely affects calculation
Method optimizes performance, and all dimensions of cock position update simultaneously in typical chicken group algorithm, it is contemplated that Degree distributions probability
The sensitivity of value, especially when dimension is very big, above-mentioned renewal principle most likely results in some dimensions updatings in cock position
Validity is offset by the renewal of other dimensions, so can reduce the generation of the above situation using limit dimensions updating mechanism, that is, is existed
Only randomly select several dimensions in cock position when dimension is larger to be updated, the location status of other dimensions keeps constant.
Step 7;Location status is carried out to whole hens in chicken group using the hen more new formula in typical chicken group algorithm
Renewal.
Step 8:Location status renewal is carried out to whole chickens in chicken group using Optimal Learning strategy.Optimal Learning strategy
It can be described as:Chicken mother's search of food can only be surrounded in view of the chicken in typical chicken group algorithm, i.e., chicken can only be from small
Chicken mother obtains information and carries out location updating, can weaken the global search of algorithm if now chicken mother is in same state
Ability, therefore Optimal Learning strategy is used to chicken location updating, chicken position updating process can be described asWherein FL represents the study of chicken
Ability level and FL ∈ [0,2], xM, j(t) i-th of chicken mother position that jth is tieed up in the t times iteration is represented,
Represent the cock of current contribution globally optimal solution in the position that the t times iteration jth is tieed up, c2It is that chicken follows institute for Studying factors
Select the size of the learning ability of cock.
Step 9:It is right using differential evolution (Differential Evolution, DE) strategy under current iteration number
All individual is updated optimization again in chicken group, and definition zoom factor is M, crossover probability CR, and variation mode uses DE/
Rand/1/bin, wherein DE represent differential evolution, and the vector that rand represents currently to be made a variation randomly generates, and 1 represents to utilize difference
The number of vector, bin represent that cross-mode is binomial pattern (binomial, bin).Differential evolution strategy includes three behaviour
Make step, respectively make a variation, intersect and select.Mutation operation can be described as:The each individual generation one being followed successively by chicken group
Individual corresponding variation vector, one variation of often generation select other three different individuals from chicken group at random when vectorial,
Mutation operation process can be expressed asWherein vi(t+1) represent
I-th of individual corresponding variation vector in the t+1 times iteration in chicken group,WithTable respectively
Show three individuals randomly selected from the chicken group updated for the first time, and i ≠ y1≠y2≠y3, y1, y2, y3∈ [1, K, N], M ∈
[0,2],To be vectorial by variation,For difference vector.Crossover operation can be described as:In order to
Increase the diversity of population, cross selection is carried out to each individual in chicken group and its corresponding variation vector, produces experiment individual
Crossover operation process can be expressed as
Wherein uI, j(t+1) i-th of individual trial vector that jth is tieed up in the t+1 times iteration, Rand (j) caused by chicken group are represented
Represent caused between [0,1] and obey equally distributed j-th of random number, CR ∈ [0,1], jrandRepresent between [0, n]
Caused random integers.Selection operation can be described as:Under determining by the individual updated for the first time to turn into chicken group
Generation member, fitness value is calculated to testing individual corresponding to more new individual for the first time and its according to object function, then using greedy
Greedy strategy is selected, and selection operation process can be expressed as xi(t+1)=ui(t+1)for f(ui(t+1)) < f (xi(t+
1))。
Step 10:Calculating is iterated, when iterations reaches maximum iteration, terminates the searching process of the algorithm
And export a body position corresponding to final optimal solution and spend distributed architecture, completing LT codes degree long to short code according to which is distributed
Optimization design.
The present invention realizes a kind of design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm, the party
Required average number of coded symbols minimizes the average value generation for as optimization aim, selecting coded identification when method is using successfully decoded
For the decoding overheads object function good and bad as the distribution of measurement degree, optimization problem is solved using improved chicken group algorithm,
Location status renewal is successively carried out to whole cocks in chicken group using amplitude constraint strategy, limit dimensions updating mechanism, using most
Excellent learning strategy carries out location status renewal to whole chickens in chicken group, after all individual renewals terminate in chicken group, quotes
Differential evolution strategy is updated optimization again to all individuals, and the design method can while algorithm the convergence speed is ensured
Obtaining more excellent degree distribution makes LT code performances more excellent.
Brief description of the drawings:
Fig. 1 is FB(flow block) of the present invention;
Fig. 2 is the simulated effect figure that the present invention is applied to sparse distribution optimization form;
Fig. 3 is the simulated effect figure that the present invention is applied to orphan's distribution optimization form.
Embodiment:
The purport of the present invention is to propose a kind of design side based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm
Method, the degree distribution that this method obtains can effectively lift the performance of the long LT codes of short code.
Embodiment of the present invention is described further in detail for 1, accompanying drawing 2 and accompanying drawing 3 below in conjunction with the accompanying drawings.
LT codes degree distribution optimization design method problem long to short code carries out mathematical description, according to the probability distribution shape of degree distribution
Formula and the position feature for improving chicken colony optimization algorithm, degree of foundation distribution optimization problem model are required during by successfully decoded to be averaged
Number of coded symbols minimizes and is used as optimization aim, i.e., the coding symbol number received when successfully recovering all raw data symbols
Amount is more few better, and it is the quantity for translating the coded identification needed for a code word to define T, then minimizing average coding symbol number can be with
Embodied by decoding overheads, decoding overheads are
Wherein E represents expectation, and k represents raw data symbols quantity.In the Optimized model of proposition, in order to suitable for gunz
The characteristic of energy optimized algorithm, the average value of coded identification is selected to replace decoding overheads excellent as the distribution of measurement degree in optimization process
Bad object function, object function are
Wherein f represents fitness value, and Ω represents some complete degree distribution, and m is sample size, i.e., for identical degree point
Cloth carries out the number of encoding and decoding, and T (r) represents average number of coded symbols required during successfully decoded in r-th of sample.Therefore exist
Finding optimal solution in degree distribution optimization problem can be defined as
OS(Ω)=arg (min (f)) (3)
Wherein arg represents the Ω values for making f reach minimum value, OSOptimal solution is represented, then above formula smaller, the LT that can be understood as f
The performance of code is better, and the degree distribution that Ω is represented is more excellent.
Assuming that the quantity comprising probable value is D in degree distribution, angle value is the raw data symbols needed for a coded identification
Quantity is d, and maximum angle value is expressed as dmax, produce the degree distribution probability value that angle value is d and be represented by Ω (d), then one it is complete
Degree distributed architecture be represented by
Ω={ Ω (d1), Ω (d2), K, Ω (dmax)} (4)
Wherein 0 < Ω (d) < 1,The related notion of chicken colony optimization algorithm is introduced, defines chicken group's
Population scale is N, position vector x, and the size of search space is expressed as D, and using chicken group body position as optimized variable, by
It is corresponding with degree distributed architecture in individual body position, so the position vector of i-th of individual is represented by
xi=[pI, 1, pI, 2, K, pI, j, K, pI, D] (5)
Whereinxi=Ωi, pI, D=Ωi(dmax), pI, jParameter in ∈ (0,1) degree of a representation distributed architecture
Probable value.According to the concrete condition of code length size, suitable degree distribution form is selected to optimize, can be direct if code length is slightly shorter
Selection orphan's distribution form optimizes, and now individual body position is the probability of the angle value from 1 to k, therefore has dmax=k, if code length
Slightly long, because the meeting of chicken colony optimization algorithm becomes big with the increase of code length, and search space scope becomes conference and answers problem
Hydridization can select degree of rarefication distributed architecture form to optimize with the big of less search space so as to influence effect of optimization
It is small, such as angle value size only considers the index that power is 2, the angle value of other positions is defaulted as 0, i.e. dj=2j, j=1,2, K, D, together
When meet dmax=2D, dmax< k.
Design improved chicken colony optimization algorithm to solve degree distribution optimization problem, generation certain scale N chicken group, chicken
A kind of degree distributed architecture of the corresponding optimization problem of each individual in group, to all individual carries out position initialization in chicken group, and
Using initial position as initial degree distribution ΩI, define cock in population, hen, the quantitative proportion difference of chicken and hen mother
For RN, HN, CN, MN, all individual fitness values of chicken group are calculated according to object function, the fitness value according to obtained by calculating
Quality establishes chicken group's hierarchy, with reference to variety classes individual proportion, for selecting several to possess adaptive optimal control angle value
Body selects several that there is each body of worst fitness value to be used as hen as chicken, remaining other individuals as cock.According to
Cock quantity is grouped to all individuals of chicken group, and each packet includes a cock, some hens and a small amount of chicken, female
Chicken selection belongs to the packet nearest from it, and the mother-child relationship (MCR) of hen and chicken is using random fashion foundation, using amplitude constraint plan
Slightly, limit dimensions updating mechanism carries out location status renewal to whole cocks in chicken group, using Optimal Learning strategy in chicken group
Whole chickens carry out location status renewal, when in chicken group it is all individual renewal terminate after, quote differential evolution strategy to whole
Individual is updated optimization again.
As shown in figure 1, calculating is iterated using chicken group's algorithm is improved, when iterations reaches the greatest iteration time of setting
During number, terminate searching process and export final most goodness distributed architecture, completing LT codes degree long to short code according to which is distributed
Optimization design.
The optimization method of the present invention is briefly described below by 2 instantiations.Select k=64's in example 1
The optimization of code length degree of progress distribution, it is sparse using the optimization form of degree of rarefication distribution in order to reduce the complexity of optimization problem
Degree distribution form is represented by Ω={ Ω (d1), Ω (d2), Ω (d4), Ω (d8), Ω (d16), Ω (d32), therefore dimension D=
6, initial degree distribution ΩIRandomly generate.The optimization of k=50 code length degree of progress distribution is selected in example 2, is distributed using orphan's degree
Optimization form, orphan's distribution form is represented by Ω={ Ω (d1), Ω (d2), K, Ω (d49), Ω (d50), dimension D=50,
Initial degree distribution ΩI=0.2,0.2, K, 1/i (i-1), and K, 0.2 }, wherein i=3,4, K, k-1.Assuming that passed in ideal communication channel
Defeated information, the condition of setting include population scale N=100, maximum iteration 80, cock, hen, chicken and chicken mother
Ratio be respectively RN=0.15, HN=0.7, CN=0.15, MN=0.5, calculate the sample size m=8 of adaptive value, scale
Factor M=0.5, crossover probability CR=0.8, c1=0.6, c2=0.4.In order to evaluate the validity of the design method, to above-mentioned
Example carries out the Monte Carlo simulation experiment of 40 times.
Design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm is applied to sparse distribution optimization shape
The simulation result of formula as shown in Fig. 2 wherein with the design side based on the particle cluster algorithm with gradient and typical chicken group algorithm
Method compares, and solid line, dotted line and chain-dotted line are represented based on the particle cluster algorithm with gradient, typical chicken group algorithm and changed respectively
Enter the degree distribution optimization design method of chicken group's algorithm.Required average number of coded symbols is minimised as when now with successfully decoded
The optimal solution of optimization aim is OS={ 0.0835,0.6660,0.0822,0.0596,0.0801,0.0286 }, it is imitative based on 40 times
True result, average fitness value when calculating optimal solution is f=73.7750.Based on the long LT of short code for improving chicken colony optimization algorithm
Code degree distribution design method be applied to orphan's distribution optimization form simulation result as shown in figure 3, wherein with based on band gradient
Particle cluster algorithm and the design method of typical chicken group algorithm compare, solid line, dotted line and chain-dotted line represent to be based on respectively
Particle cluster algorithm with gradient, typical chicken group algorithm and the degree distribution optimization design method for improving chicken group's algorithm.Now with solution
The optimal solution that required average number of coded symbols is minimised as optimization aim during code success is OS=0.1644,0.4275,
0.0530, K, 0.0102 }, based on 40 simulation results, average fitness value when calculating optimal solution is f=58.9094.When repeatedly
The present invention is already available to still to keep most when preferably fitness value and iterations are 80 when generation number reaches 15
Excellent state, simulation result illustrate that the method based on the distribution of improved chicken colony optimization algorithm design degree can be with faster convergence rate
Obtain preferably degree distribution.
Claims (4)
1. a kind of design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm, it is comprised the following steps that:
Step 1:According to the form of probability of degree distribution and the position feature of improvement chicken colony optimization algorithm, degree of foundation distribution optimization
Problem model, optimized using a degree distribution as population at individual position, it is assumed that raw data symbols quantity is k, degree distribution
In the quantity comprising probable value be D, angle value is that raw data symbols quantity needed for a coded identification is d, maximum angle value
It is expressed as dmax, produce the degree distribution probability value that angle value is d and be represented by Ω (d), then a complete degree distributed architecture can represent
For Ω={ Ω (d1), Ω (d2), K, Ω (dmax), and meetWherein 0 < Ω (d) < 1;
Step 2:The degree distribution optimization problem model carried with reference to step 1, the related notion of chicken colony optimization algorithm is introduced, define chicken
The population scale of group is N, and position vector x, the size of search space is expressed as D, and becomes chicken group body position as optimization
Amount, because individual body position is corresponding with degree distributed architecture, so the position vector of i-th of individual is represented by xi=[pI, 1,
pI, 2, K, pI, j, K, pI, D], pI, jParameter probable value in ∈ (0,1) degree of a representation distributed architecture,Wherein there is xi
=Ωi, pI, D=Ωi(dmax);
Step 3:With reference to chicken group's related notion is carried in the degree distribution optimization model and step 2 that step 1 is carried, according to code length size
Concrete condition, select suitable degree distribution form to optimize, if code length is slightly shorter, the progress of orphan's distribution form can be directly selected
Optimization, now individual body position is the probability of the angle value from 1 to k, therefore has dmax=k, if code length is slightly long, calculated because chicken group optimizes
The search space of method can become big with the increase of code length, and search space scope becomes conference and complicated the issue so as to influence
Effect of optimization, therefore degree of rarefication distributed architecture form can be selected to optimize with the size of less search space, such as angle value
Size only considers the index that power is 2, and the angle value of other positions is defaulted as 0, i.e. dj=2j, j=1,2, K, D, while meet dmax=
2D, dmax< k;
Step 4:Required average number of coded symbols is minimized as optimization aim during using successfully decoded, i.e., successfully recovers all
The number of coded symbols received during raw data symbols is more few better, and it is to translate the coded identification needed for a code word to define T
Quantity, then minimizing average coding symbol number can be embodied by decoding overheads, and decoding overheads are designated as
Wherein E represents it is expected, in the Optimized model of proposition, for the characteristic suitable for colony intelligence optimized algorithm, in optimization process
Select the average value of coded identification to replace the decoding overheads object function good and bad as the distribution of measurement degree, be represented byWherein f represents fitness value, and m is sample size, i.e., is distributed for identical degree and carries out encoding and decoding
Number, T (r) represents required average number of coded symbols during successfully decoded in r-th of sample, therefore is asked in degree distribution optimization
Optimal solution is found in topic can be defined as OS(Ω)=arg (min (f));
Step 5:Improved chicken colony optimization algorithm is designed, then the degree distribution optimization problem in step 4 is solved, generation one
Establish rules mould N chicken group, a kind of degree distributed architecture of the corresponding optimization problem of each individual in chicken group, whole individuals in chicken group is entered
Row position initialization, and using initial position as initial degree distribution ΩI, define cock, hen, chicken and hen mother in population
Quantitative proportion be respectively RN, HN, CN, MN, all individual fitness values of chicken group are calculated according to the object function in step 4,
Quality according to gained fitness value is calculated establishes chicken group's hierarchy, that is, determines character species of the individual in chicken group, with reference to
Variety classes individual proportion, select several to possess the individual of adaptive optimal control angle value as cock, select several to have
Each body of worst fitness value is used as hen, all individuals of chicken group is entered according to cock quantity as chicken, remaining other individuals
Row packet, each packet include a cock, some hens and a small amount of chicken, and hen selection belongs to point nearest from it
The mother-child relationship (MCR) of group, hen and chicken is established using random fashion;
Step 6:Location status renewal is carried out to whole cocks in chicken group using amplitude constraint strategy, limit dimensions updating mechanism;
Step 7;Location status renewal is carried out to whole hens in chicken group using the hen more new formula in typical chicken group algorithm;
Step 8:Location status renewal is carried out to whole chickens in chicken group using Optimal Learning strategy;Step 9:In current iteration
Under number, using differential evolution (Differential Evolution, DE) strategy to all individual is carried out more again in chicken group
New optimization, definition zoom factor are M, crossover probability CR, and variation mode uses DE/rand/1/bin, wherein DE expression difference
Evolve, the vector that rand represents currently to be made a variation randomly generates, and 1 represents the number using difference vector, and bin represents cross-mode
For binomial pattern (binomial, bin), differential evolution strategy includes three operating procedures, respectively makes a variation, intersects and selects
Select;
Step 10:Calculating is iterated, when iterations reaches maximum iteration, terminates the searching process of the algorithm and defeated
Go out a body position corresponding to final optimal solution and spend distributed architecture, the excellent of LT codes degree distribution long to short code is completed according to which
Change design.
2. a kind of design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm, it is characterised in that in step 6
Amplitude constraint strategy can be described as:It can be seen from the Optimized model carried in step 1 and step 2, a body position in chicken group
Each dimension actually with degree distributed architecture in probable value be of equal value, then cock location updating scope should be controlled 0
In the range of 1, otherwise renewal is invalid, in view of directly using cock more new formula in typical chicken group algorithm to complete in chicken group
Portion cock carries out location updating and easily produces above-mentioned renewal inefficiency, uses amplitude constraint strategy to control cock location updating
Position changes in amplitude processed, then cock position updating process can be described as xI, j(t+1)=xI, j(t)(1+c1Randn (0, σ2)),
Wherein t represents iterations, xI, jAnd x (t+1)I, j(t) i-th of individual is illustrated respectively in the t times and the t+1 times iteration
The positional value of jth dimension, c1To limit the controlling elements of rangeability, Randn (0, σ2) represent that it is σ to be desired for 0 variance2's
Gaussian Profile;Limit dimensions updating mechanism in step 6 can be described as:In chicken group's algorithm, cock, which has, preferentially searches out food
Right, then it has bigger food search space, therefore the validity of cock location updating largely affects calculation
Method optimizes performance, and all dimensions of cock position update simultaneously in typical chicken group algorithm, it is contemplated that Degree distributions probability
The sensitivity of value, especially when dimension is very big, above-mentioned renewal principle most likely results in some dimensions updatings in cock position
Validity is offset by the renewal of other dimensions, so can reduce the generation of the above situation using limit dimensions updating mechanism, that is, is existed
Only randomly select several dimensions in cock position when dimension is larger to be updated, the location status of other dimensions keeps constant.
3. a kind of design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm, it is characterised in that in step 8
Optimal Learning strategy can be described as:Chicken mother's search of food can only be surrounded in view of the chicken in typical chicken group algorithm,
I.e. chicken can only obtain information from chicken mother and carry out location updating, can weaken if now chicken mother is in same state
The ability of searching optimum of algorithm, therefore Optimal Learning strategy is used to chicken location updating, chicken position updating process can be retouched
State forWherein FL represents chicken
Learning ability level and FL ∈ [0,2], xM, j(t) i-th of chicken mother position that jth is tieed up in the t times iteration is represented,Represent the cock of current contribution globally optimal solution in the position that the t times iteration jth is tieed up, c2It is chicken for Studying factors
Follow the size of the learning ability of selected cock.
A kind of 4. design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm, it is characterised in that step 9
In, mutation operation can be described as:The each one corresponding variation of individual generation being followed successively by chicken group is vectorial, often generates one
It is individual variation it is vectorial when other three different individuals are selected from chicken group at random, mutation operation process can be expressed asWherein vi(t+1) represent that i-th of individual is in t+ in chicken group
Corresponding variation vector during 1 iteration,WithRepresent respectively from the chicken group updated for the first time
Three individuals randomly selected, and i ≠ y1≠y2≠y3, y1, y2, y3∈ [1, K, N], M ∈ [0,2],To be become incorgruous
Amount,For difference vector;Crossover operation can be described as:In order to increase the diversity of population, to chicken group
In each individual and its corresponding variation vector carry out cross selection, producing the crossover operation process of experiment individual can be expressed asWherein uI, j(t+1) represent in chicken group
Caused i-th of individual trial vector that jth is tieed up in the t+1 times iteration, Rand (j) represent the caused clothes between [0,1]
From equally distributed j-th of random number, CR ∈ [0,1], jrandRepresent the caused random integers between [0, n];Selection operation
It can be described as:In order to determine that member of future generation can be turned into by the individual updated for the first time in chicken group, according to object function pair
First more new individual and its corresponding experiment individual calculate fitness value, are then selected using greedy strategy, selection operation
Process can be expressed as xi(t+1)=ui(t+1)for f(ui(t+1)) < f (xi(t+1))。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710815343.7A CN107682117B (en) | 2017-09-11 | 2017-09-11 | Short code length LT code degree distribution design method based on improved chicken flock optimization algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710815343.7A CN107682117B (en) | 2017-09-11 | 2017-09-11 | Short code length LT code degree distribution design method based on improved chicken flock optimization algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107682117A true CN107682117A (en) | 2018-02-09 |
CN107682117B CN107682117B (en) | 2020-06-16 |
Family
ID=61135696
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710815343.7A Active CN107682117B (en) | 2017-09-11 | 2017-09-11 | Short code length LT code degree distribution design method based on improved chicken flock optimization algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107682117B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109450596A (en) * | 2018-11-12 | 2019-03-08 | 湖南瑞利德信息科技有限公司 | Coding method, coding/decoding method, encoding device, decoding device, storage medium and terminal |
CN110086568A (en) * | 2019-04-02 | 2019-08-02 | 中山大学 | The LDPC code interpretation method of dynamic select strategy based on information updating variable range |
CN110119778A (en) * | 2019-05-10 | 2019-08-13 | 辽宁大学 | A kind of equipment method for detecting health status improving chicken group's optimization RBF neural |
CN110555283A (en) * | 2019-09-11 | 2019-12-10 | 湖北工业大学 | Efficient LT code degree distribution design method and system |
CN111273606A (en) * | 2020-03-09 | 2020-06-12 | 西南交通大学 | Tool posture optimization method for geometric error compensation of five-axis machine tool |
CN112365196A (en) * | 2020-12-03 | 2021-02-12 | 国网河北省电力有限公司信息通信分公司 | Spark-based power supply planning method for distributed improved differential evolution algorithm |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104880949A (en) * | 2015-05-27 | 2015-09-02 | 江南大学 | Method of acquiring workpiece-processing optimal scheduling based on improved chicken flock algorithm |
CN106534009A (en) * | 2016-11-29 | 2017-03-22 | 安徽理工大学 | Improved variable step size equivariant adaptive blind source separation method |
CN106778809A (en) * | 2016-11-23 | 2017-05-31 | 安徽理工大学 | A kind of blind source separation method based on improvement chicken group's algorithm |
-
2017
- 2017-09-11 CN CN201710815343.7A patent/CN107682117B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104880949A (en) * | 2015-05-27 | 2015-09-02 | 江南大学 | Method of acquiring workpiece-processing optimal scheduling based on improved chicken flock algorithm |
CN106778809A (en) * | 2016-11-23 | 2017-05-31 | 安徽理工大学 | A kind of blind source separation method based on improvement chicken group's algorithm |
CN106534009A (en) * | 2016-11-29 | 2017-03-22 | 安徽理工大学 | Improved variable step size equivariant adaptive blind source separation method |
Non-Patent Citations (2)
Title |
---|
CHIH-MING CHEN等: ""On the Optimization of Degree Distributions in LT Code with Covariance Matrix Adaptation Evolution Strategy"", 《IEEE CONGRESS ON EVOLUTIONARY COMPUTATION》 * |
邓在辉等: ""采用带梯度的粒子群优化算法的LT码度分布设计"", 《北京邮电大学学报》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109450596A (en) * | 2018-11-12 | 2019-03-08 | 湖南瑞利德信息科技有限公司 | Coding method, coding/decoding method, encoding device, decoding device, storage medium and terminal |
CN109450596B (en) * | 2018-11-12 | 2022-02-01 | 湖南瑞利德信息科技有限公司 | Encoding method, decoding method, encoding device, decoding device, storage medium, and terminal |
CN110086568A (en) * | 2019-04-02 | 2019-08-02 | 中山大学 | The LDPC code interpretation method of dynamic select strategy based on information updating variable range |
CN110086568B (en) * | 2019-04-02 | 2020-09-22 | 中山大学 | LDPC code decoding method based on dynamic selection strategy with variable message updating range |
CN110119778A (en) * | 2019-05-10 | 2019-08-13 | 辽宁大学 | A kind of equipment method for detecting health status improving chicken group's optimization RBF neural |
CN110119778B (en) * | 2019-05-10 | 2024-01-05 | 辽宁大学 | Equipment health state detection method for improving chicken flock optimization RBF neural network |
CN110555283A (en) * | 2019-09-11 | 2019-12-10 | 湖北工业大学 | Efficient LT code degree distribution design method and system |
CN110555283B (en) * | 2019-09-11 | 2022-11-29 | 湖北工业大学 | Efficient LT code degree distribution design method and system |
CN111273606A (en) * | 2020-03-09 | 2020-06-12 | 西南交通大学 | Tool posture optimization method for geometric error compensation of five-axis machine tool |
CN112365196A (en) * | 2020-12-03 | 2021-02-12 | 国网河北省电力有限公司信息通信分公司 | Spark-based power supply planning method for distributed improved differential evolution algorithm |
Also Published As
Publication number | Publication date |
---|---|
CN107682117B (en) | 2020-06-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107682117A (en) | A kind of design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm | |
Lee et al. | Graph embedding-based wireless link scheduling with few training samples | |
Chen et al. | A GNN-based supervised learning framework for resource allocation in wireless IoT networks | |
Nguyen et al. | Non-cooperative energy efficient power allocation game in D2D communication: A multi-agent deep reinforcement learning approach | |
EP3422261A1 (en) | Global optimization, searching and machine learning method based on lamarck acquired genetic principle | |
Wang et al. | Learn to compress CSI and allocate resources in vehicular networks | |
Supraja et al. | Optimized neural network for spectrum prediction using genetic algorithm in cognitive radio networks | |
CN108924935A (en) | A kind of power distribution method in NOMA based on nitrification enhancement power domain | |
CN111490853A (en) | Channel coding parameter identification method based on deep convolutional neural network | |
CN112153616B (en) | Power control method in millimeter wave communication system based on deep learning | |
AU2018267755A1 (en) | Array element arrangement method for L-type array antenna based on inheritance of acquired characteristics | |
CN112836822B (en) | Federal learning strategy optimization method and device based on width learning | |
Lan et al. | Progressive feature transmission for split classification at the wireless edge | |
CN112307927A (en) | BP network-based identification research for MPSK signals in non-cooperative communication | |
Shodamola et al. | A machine learning based framework for KPI maximization in emerging networks using mobility parameters | |
CN115829027A (en) | Comparative learning-based federated learning sparse training method and system | |
CN113038612A (en) | Cognitive radio power control method based on deep learning | |
CN106257849B (en) | Frequency spectrum sensing method based on multi-target quantum firefly search mechanisms | |
Zec et al. | Specialized federated learning using a mixture of experts | |
CN114204971A (en) | Iterative aggregation beamforming design and user equipment selection method | |
CN116112934A (en) | End-to-end network slice resource allocation method based on machine learning | |
Zhang et al. | Federated multi-task learning with non-stationary heterogeneous data | |
CN114337883B (en) | CNN collaborative spectrum sensing method and system for covariance matrix Cholesky decomposition | |
Singhal et al. | Greedy Shapley Client Selection for Communication-Efficient Federated Learning | |
CN105376185A (en) | Constant modulus blind equalization processing method based on optimization of DNA shuffled frog leaping algorithm in communication system |
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 |