CN102163249B - Interactive evolution optimization method for curtain design - Google Patents
Interactive evolution optimization method for curtain design Download PDFInfo
- Publication number
- CN102163249B CN102163249B CN 201110094666 CN201110094666A CN102163249B CN 102163249 B CN102163249 B CN 102163249B CN 201110094666 CN201110094666 CN 201110094666 CN 201110094666 A CN201110094666 A CN 201110094666A CN 102163249 B CN102163249 B CN 102163249B
- Authority
- CN
- China
- Prior art keywords
- user
- curtain
- evolution
- evaluation
- individuality
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000013461 design Methods 0.000 title claims abstract description 35
- 230000002452 interceptive effect Effects 0.000 title claims abstract description 17
- 238000005457 optimization Methods 0.000 title claims abstract description 8
- 238000011156 evaluation Methods 0.000 claims abstract description 51
- 230000008569 process Effects 0.000 claims abstract description 22
- 230000003993 interaction Effects 0.000 claims abstract description 6
- 108090000623 proteins and genes Proteins 0.000 claims description 18
- 230000019771 cognition Effects 0.000 claims description 6
- 230000001149 cognitive effect Effects 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 5
- 230000004069 differentiation Effects 0.000 claims description 4
- 101150008942 J gene Proteins 0.000 claims description 2
- 238000005286 illumination Methods 0.000 claims 1
- 230000002708 enhancing effect Effects 0.000 abstract 1
- 238000010187 selection method Methods 0.000 abstract 1
- 238000012549 training Methods 0.000 description 14
- 238000013528 artificial neural network Methods 0.000 description 13
- 239000000047 product Substances 0.000 description 9
- 239000003795 chemical substances by application Substances 0.000 description 8
- 230000000694 effects Effects 0.000 description 8
- 230000003044 adaptive effect Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 230000010429 evolutionary process Effects 0.000 description 5
- 206010016256 fatigue Diseases 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 230000002068 genetic effect Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 230000036651 mood Effects 0.000 description 3
- 101150095029 W gene Proteins 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 239000012141 concentrate Substances 0.000 description 2
- 238000012938 design process Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 208000019901 Anxiety disease Diseases 0.000 description 1
- 230000036506 anxiety Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000004737 colorimetric analysis Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000003292 diminished effect Effects 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 239000012467 final product Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000002921 genetic algorithm search Methods 0.000 description 1
- 238000010353 genetic engineering Methods 0.000 description 1
- 230000008303 genetic mechanism Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 230000008521 reorganization Effects 0.000 description 1
- 229920006395 saturated elastomer Polymers 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
Images
Landscapes
- User Interface Of Digital Computer (AREA)
Abstract
The invention discloses an interactive evolution optimization method for curtain design, which guides user to look for the preferred design scheme and specifically comprises the steps of: (1) before evolution begins, allowing a user to set fuzzy preference information, including the degree of influence of using occasion, style and preference information of curtain on individuals, and generating a particular initial population therefrom by a computer; (2) providing a plurality of individual evaluation ways for the user in the process of human-computer interaction, including multi-set type fuzzy set evaluation, exact value evaluation and automatic evaluation, and designing selection methods of different evaluation ways; and (3) in the process of evolution, supporting the use of different coding manners and corresponding evolution algorithms between different evolution generations in order to facilitate the search of the user for curtain schemes in different areas. While the above method is provided, the corresponding system is hereby developed. The method of the invention aims at inspiring the user creation and enhancing the efficiency of designing curtain product in order to raise the market competitiveness of curtain.
Description
Technical field
This patent belongs to the evolutionary computation field, is specifically related to a kind of interactive evolutionary optimization method for the curtain design, can be used for guiding the user to carry out the curtain design for scheme.
Background technology
Along with expanding economy and growth in the living standard; the consumer is also more and more higher to the requirement of this family product of curtain; usually the factor that can take all factors into consideration each side is selected; and outward appearance wherein is undoubtedly the factor that the consumer values the most at this, and outstanding appearance design often can significantly improve the curtain competitiveness of product in market.Traditional simple dependence designer's intention and the mode that inspiration designs can't be satisfied the demand, and design brand-new works because the designer is difficult to cast aside the constraint of original scheme.This present situation for a change, the present invention is incorporated into evolutionary computation in the curtain design, can excite designer's inspiration greatly, instructs design process.
Evolutionary computation be the biological evolutionary process of simulation with mechanism find the solution the highly-parallel of problem, at random, adaptive artificial intelligence technology, come down to a kind of iterative adaptive probability searching method that is based upon on natural selection principle and the genetic mechanism, it represents the structure of various complexity by different coding techniquess, and by intersecting, the direction that variation waits the natural law of genetic manipulation and the survival of the fittest to come guidance learning and determine to search for, thereby produce population of new generation, population is evolved to comprise the state of approximate optimal solution.
In recent years, the example in existing many innovative designs that evolutionary computation successfully is incorporated into product.For example assist at the 15 national computer and design in " the paper-cut pattern innovative design method of support evolution is studied and realized " literary composition of delivering with the graphics academic conference, with the design of evolutionary computation for paper-cut pattern, generate new patterns; Stepping on word at soft work is in No. 0186137 the software " based on the dress designing software of the individual uncertain adaptive value interactive genetic algorithm of evolving ", and evolutionary computation is used for the design of clothes fashion, seeks the colour match and the style that meet user preference most.Though on going result has obtained certain success, because the evaluation method that adopts, evolution strategy etc. are comparatively single, very flexible, user's personalization is supported not enough, so ubiquity efficiency of evolution not high, problems such as ease for use difference.People such as Jimmy Secretan are in " Picbreeder:Evolving Pictures Collaboratively Online " literary composition that CHI meeting in 2008 is delivered, introduced the online picture evolutionary system by its team develops, this system has adopted the mode of evolution based on the weights network, can evolve out from the simple picture of a width of cloth and satisfy the complicated picture of customer requirements, improve the effect of evolving greatly.But in this cover system, owing to can't generate the defective that exists on initial population with particular community and the evolution strategy, cause and between the diversity of population and convergence, to reach balance preferably, increase the multifarious speed of convergence that can slow down population simultaneously greatly of population, vice versa.Although existing these achievements in research exist various problems, still to the invention provides many-sided useful guidance.
Evolutionary computation is constantly promoting its application in productive life in the significant achievements that obtain aspect many theoretical researches, embodies huge realistic meaning.Through consulting pertinent literature, present this technology of no-trump still is applied to the precedent in the curtain design, if can be at the appearance design problem of curtain, and efficient, the practical evolvement method of design one cover will alleviate designer's workload greatly, promotes the development of curtain production.Simultaneously also can consider it is generalized in the design field of other products, bring into play bigger benefit.
Summary of the invention
The present invention proposes the complete individuality of a cover and generate, reach alternately evolvement method, can style, colour match and the pattern etc. of curtain be designed by evolution algorithm, and developed corresponding system accordingly, its basic goal is to realize the intellectuality of curtain design, excites designer's creation inspiration.Characteristics of the present invention are: at first allow the user to set the generation that ambiguous preference information is used for initial population, can reduce the region of search greatly like this; Secondly among the present invention for the user provides multiple individual evaluation mode, to be user-friendly to; The 3rd, in whole evolutionary process, with changing individual coded system per family, decide scope and the form of search with this.Each characteristic realizes by corresponding functional module, and then of the present invention three big modules reach that wherein employed method is as follows:
1, initial population generation module
At the beginning of the curtain Evolutionary Design, for the user provides preference information option is set by computing machine, the user arranges corresponding preference information according to the target customers of designed curtain product, generates specific initial population according to these information by computing machine.Particular content comprises:
1.1, any a product has target customers at the beginning of design, have more targetedly design and can significantly improve the competitiveness of product in market.Based on above consideration, the present invention supports user's setting section preference information before the beginning of evolving, and comprising: the use occasion of curtain, curtain style and preference information are to the influence degree of individual generative process etc.
1.2, after the setting that obtains the user, by the HSI model in the colorimetry preference information is applied in the individual generative process, thereby produces specific initial population.
More than these are set, can make the initial population that generates concentrate on certain specific region, reduced the search volume, make things convenient for the user to find fast to meet the design proposal of self preference.
2, human-computer interaction module
In whole curtain intelligent design process, man-machine interaction directly affects the quality of final design proposal as a vital ring.This module is evaluated as the master for the user provides multiple close friend, complete man-machine interaction mode with many collection gesture fuzzy set, as replenishing of this mode, provides exact value evaluation and automatic Evaluation, and particular content comprises:
2.1, collect the evaluation of gesture fuzzy set more: comprise the fuzzy set of multiple collection gesture among the present invention, the evaluation language that comprises of fuzzy set of every kind of collection gesture is different, and for example the fuzzy sets of 3 collection gesture are:
And the fuzzy set of 5 collection gesture is:
Wherein
Expression collects j evaluation language in the fuzzy set that gesture is i, and adjective is for estimating the content of language.In the different phase of evolving, provide a kind of fuzzy set that collects gesture for certain generation evolution is individual by computing machine automatically according to the strategy of formulating, therefrom select one to estimate language as the adaptive value of individuality by the user.This kind evaluation method is simple and easy to usefulness, can effectively alleviate the evaluation of user burden.
2.2, the exact value evaluation: be that individuality is given a perfect number 0 to 100 as the adaptive value of individuality by the user.This kind mode when the user thinks that fuzzy set can't well have been estimated individuality, can select to use this mode as replenishing that many collection gesture fuzzy sets are estimated.This mode is the discriminate individuals quality obviously, but to the having relatively high expectations of user, easily makes the user occur estimating tired.
2.3, automatic Evaluation: this mode utilizes the neural metwork training agent model to replace the user that individuality is estimated, and reaches requirement and user in training precision and thinks and be incorporated in the evolutionary process in case of necessity.Three layers of feedforward neural network are adopted in the training of agent model, can rely on its powerful local approximation capability to satisfy the user to the requirement of individual evaluation, can select to use this evaluation method when the user estimates appearance fatigue.
In the present invention, the user can be as required switched above-mentioned evaluation method when estimating arbitrarily in Different Evolutionary, alleviating to greatest extent when the user estimates burden, also can ensure individuality is made effective evaluation, has improved practicality.
3, evolution module
This module for the user provides a whole set of flexibly, satisfies the evolution operation of multiple evolution requirement by the individuality in the population being done multi-form coding and comprehensively being adopted multiple evolution algorithm.Particular content comprises:
3.1, institute's extracting method support comprises binary-coded character string encoding and integer coding to the individual different coding form that adopts of evolving among the present invention, and can switch in the generation of evolving arbitrarily, uses different evolution algorithms under the different coding forms.
3.2, computing machine is according to the coding mode selection evolution algorithm, under the string of binary characters coding form, adopt interactive genetic algorithm search optimal case, adopt interactive evolution strategy search optimal case under the integer coding form, two kinds of methods directly influence the zone of search.When the user can evolve, the regional area individuality that meets self preference is done meticulousr evolution in global scope.
The evolvement method that adopts among the present invention fully combines the advantage of Different Evolutionary algorithm, makes more efficiently, meticulous to the evolution of curtain individuality, more has practical value.
In a word, intelligent curtain method for designing proposed by the invention, can be under the prerequisite that guarantees efficiency of evolution, evolution effect, reduce user's workload to greatest extent, alleviate burden for users, it is fast to satisfy in the curtain design field pace of product renewal, the characteristics that appearance requirement is high, this invention also may be used in the design of other various product, thereby has high practical value and wide application prospect.
Description of drawings
Fig. 1 is the interactive evolutionary optimization method flow diagram that is used for the curtain design that proposes among the present invention;
Fig. 2 is the curtain design system surface chart according to institute of the present invention extracting method exploitation;
Fig. 3 is the distribution plan of different use occasions on the colourity dish of curtain;
Fig. 4 is the interactive genetic algorithm process flow diagram that the present invention adopts under the string of binary characters coding mode;
Fig. 5 is the interactive evolution strategy process flow diagram that the present invention adopts under the integer coding pattern;
Embodiment
This part elaborates to embodiments of the present invention in conjunction with concrete accompanying drawing.Among the present invention the process flow diagram of the method that proposes as shown in Figure 1, concrete implementation step is also drafted according to this figure.
For demonstrating fully the value of interactive evolutionary optimization method proposed by the invention, corresponding software systems in proposition method, have also been developed.In the performance history of system, reach best visual effect for making designed curtain, all curtain all adopt the 3D model, adopt the texture mapping technology to represent the design effect of curtain.Related system adopts Visual C++6.0 as developing instrument, builds operation platform based on MFC, utilizes the OpenGL realization to the operation of 3D model, and system interface as shown in Figure 2.
Step 1: finish the initialization of evolution population.This step comprises two aspect contents: set preference information and generate initial population.Specifically finished by following two sub-steps:
Step 101: set preference information.The preference information that can set comprises that use occasion, curtain style and the preference information of curtain are to the influence degree of individual generative process herein.For use occasion, have the typical case below the present invention has chosen and represent the occasion of meaning: bedroom, parlor, office building, hotel and other; The curtain style comprises: simple and elegant, pure and fresh, enthusiasm, strong and other; Influence degree is a percentage from 0 to 100% that is set by the user.
Consider different colors to the influence of people's mood and with the harmony of surrounding environment, different use occasions influence the distribution of color, different styles influence the saturation degree of color, the percentage of setting has then been pointed out the exact level of this influence.According to the achievement of correlative study, the bedroom should make the people feel warm, comfortable with look, should select dim redness for use, tones such as pink colour; The parlor should make people's mood happy, also will consider the utilization to light, should select pale green for use, orange and white etc.; Office building belongs to the workplace, should be able to overcome anxiety, mood tired out with look, should select blueness, green etc. for use; The public place is counted in the hotel, should select more popular color for use, avoids using the color with distinct characteristic; If select other then do not limit.Four kinds of styles influence the saturation degree of color, and from simple and elegant to strong, it is saturated that color is tending towards gradually, if select other not limit equally.The setting of preceding two preferences is to the influence degree of individuality, and the influence degree that is set by the user determines.
Step 102: generate initial population.The preference information that back is set is to set by the form of language description, can't directly apply it in the evolutionary process.For realizing above function, the present invention has adopted the HSI color model.The setting of use occasion is influenced H (tone) component of model, and Fig. 3 is the distribution situations of different use occasions on the colourity dish, and three coordinate axis of RGB are divided 360 ° equally among the figure, and 0 ° is red, and 120 ° be green, and 240 ° is blueness.The tone H that P is ordered is that the vector of P and the angle of red axle are arrived in the center of circle.Among the present invention, the use occasion according to the user sets generates the vector that becomes a certain angle with red axle at respective regions by evenly distributing automatically by computing machine, if the user selects " other ", then generates a certain vector automatically in global scope.
The setting of style is influenced S (saturation degree) component of model, the saturation degree interval of " simple and elegant ", " pure and fresh ", " enthusiasm " and " strong " four correspondences is respectively in this option: [0-24%], [25%-49%], [50%-74%], [75%-100%], " other " corresponding interval then is [0-100%].Computing machine can generate an intensity value by evenly distributing automatically according to user's selection equally in respective bins.The setting of preference information effect degree is influenced I (brightness) component of model, and this moment can be directly with the influence degree of user's setting, is converted into real number form and is applied in the model and goes.The user is converted into user's preference information according to following formula (1) value of each component in the RGB model by computing machine after having set initial preference information.
According to above result, finish painted to floral designs, from material database, read in the Window curtain structure model of 3ds form then, utilize the OpenGL technology to finish pinup picture and demonstration to model, can obtain a complete curtain individuality, repeat all initialization of individuality in finishing population of this process.
Step 2: the evolution parameter is set, finishes interactive the evolution and prepare.At first, be evolution individual choice coded system by the user, the present invention supports binary-coded character string encoding and integer coding, and can switch individual coded system in the generation of evolving arbitrarily, different coded system correspondences different evolution algorithms.Secondly, for the different coding form arranges relevant parameter, if the user selects the binary-coded character string encoding, then need set crossover probability, variation probability and maximum evolutionary generation; If the selection integer coding then need be set maximum variation step-length and maximum evolutionary generation.
The user should consider the coded system that should select from the hunting zone angle of expection.Because used evolution algorithm difference, the binary-coded character string encoding stresses extensive search, and integer coding stresses Local Search, and particular content will describe in detail in subsequent step.
Step 3: the user finishes evaluation to the curtain individuality by human-computer interaction module.To the evaluation of curtain individuality, the present invention comprises for the user provides multiple individual evaluation form: collect the evaluation of gesture fuzzy set, exact value evaluation and automatic Evaluation more for the convenience of the user.The substep of every kind of evaluation method correspondence is as described below:
Step 301: use many collection gesture fuzzy sets that individuality is estimated.Under this kind mode, computing machine provides 3,5 for the user, the fuzzy set of 7 three kinds of collection gesture, and the contained evaluation number of languages of fuzzy set of different collection gesture is different, and the fuzzy set preamble of 3 collection gesture and 5 collection gesture was introduced, and the fuzzy set of 7 collection gesture is:
Because the evolution initial stage, individual difference was bigger, the user uses the fuzzy set of little collection gesture can finish evaluation, uses this moment the evaluation collection of big collection gesture can increase the evaluation of user burden on the contrary.After difference between individuality diminished gradually, it was necessary to use the fuzzy set of big collection gesture to be only because this moment little collection gesture fuzzy set can't embody individual quality.
For the objective selection evaluation of situation collection that can be individual according to evolution, the present invention at first quantizes evolving individual and estimating to collect, and has formulated concrete system of selection then, and whole process is finished automatically by computing machine.Concrete implementation step is as follows:
Step 3011: the evaluating ability that quantizes the fuzzy set of different collection gesture.Different collection gesture fuzzy sets are because contained evaluation number of languages difference, evaluation effect to individuality also is inequality, for portraying this difference, the present invention introduces concept of information entropy, each fuzzy set is considered as an information source, then estimates language and namely can be considered the signal that the equiprobability in the information source occurs.With S
iExpression collection gesture is the fuzzy set of i, and the information entropy that can try to achieve three information sources according to the computing formula of information entropy is respectively:
Because the fuzzy set of high granularity has best evaluation effect, from the angle of probability scale, the evaluation differentiation rate of ambiguity in definition collection is shown below:
The evaluation differentiation rate that can try to achieve three fuzzy sets according to the result in the formula (2) is: R (S
3)=56.4%, R (S
5)=82.7%, R (S
7)=100%.
Step 3012: the cognitive process that quantizes the user.The user is the preference to curtain at the evolution initial stage, and the relative later stage is very indeterminate, so the general performance type difference of initial stage curtain is also bigger, and along with the carrying out of evolving, individual phenotype converges to user's preference point gradually, and difference also reduces gradually.The present invention quantizes cognitive this subjective process of user accordingly.
Still adopt the method for computing information entropy.Be located in the curtain design problem, per generation population K curtain individuality arranged, each individuality comprises several separate parts, and each part is called a gene meaning unit, with each gene meaning unit as an information source independently.With
J gene meaning unit representing t i individuality in generation, with
Represent in t generation, the identical phenotypic occurrence number of i j individual gene meaning unit, then t in the information entropy of j gene meaning unit be:
Wherein
According to following formula, when j gene meaning unit phenotype of all individualities in t generation was all identical, information entropy was got minimum value, got maximal value when all inequality, so the span of the information entropy of gene meaning unit is: H (x
j(t)) ∈ [0, log
2K].For ease of follow-up comparison, asking information entropy is done normalized.The present invention has defined the cognition degree of user to a certain gene meaning unit from the angle of probability scale, the cognition degree d of t j gene meaning unit in generation
j(t) be defined as:
In the formula: H (x)
Max=log
2K
The user t for the time cognitive sharpness D (t) by the user maximal value of the cognition degree of gene meaning unit and interval that minimum value constitutes are determined that formula table is shown:
Step 3013: the selection that collects the gesture fuzzy set more.This step is the fuzzy set of a certain collection gesture of certain generation evolution individual choice automatically, so that the user finishes individual evaluation.
The selection of fuzzy set should be used the fuzzy set of little collection gesture as far as possible under the prerequisite that does not influence evaluation effect.According to R (S in the preamble
i) with the definition of D (t), the two between same zone in value and have similar Changing Pattern, the choice criteria of the many collection gesture fuzzy set that adopts in the present invention is accordingly: select to estimate differentiation rate minimum, and greater than the fuzzy set in (D (t)) upper bound between cognitive articulation area.As long as the user still adopts the evaluation of many collection gesture fuzzy sets, then before per generation evaluation beginning, all need carry out and once select operation.
Above strategy can guarantee that not only the fuzzy set that collects gesture greatly in time participates in the evaluation procedure, and can eliminate the accuracy of fuzzy set to guarantee to estimate of little collection gesture as early as possible.The cognitive fluctuation of user no matter occurs and still be the sharpness interval width and be 0 situation, this strategy all can be selected, and has stronger robustness.If the user thinks that the fuzzy set of maximum collection gesture still can't satisfy it and estimate needs, switch to the exact value evaluation model to guarantee evaluation effect.
Step 302: use the exact value between 0 to 100 that individuality is estimated.Under this kind mode, the user only needs according to self preference, gives a certain exact value for each individuality and gets final product.The evaluation of user result is directly as individual fitness under the computer recording, and is used for individual choice in follow-up phase.This mode easy operating can reflect the preference difference that the user is trickle, but also very easily increases the weight of user's burden simultaneously, causes the user to occur estimating fatigue too early, the accuracy that influence is estimated, and at this moment, the user can select the automatic Evaluation mode to lessen fatigue.
Step 303: use agent model to carry out automatic Evaluation.Because the agent model among the present invention is to adopt three layers of feedforward neural network to make up, and neural network itself needs a relatively long training process, therefore when evolving beginning, automatically begin the training of network, after training precision reaches requirement, can determine whether in certain generation, to use agent model to replace the user to estimate by the user.Concrete steps are:
Step 3031: the obtaining of neural network training desired data.After the user estimated individuality, obtain evolution individuality and evaluation result data thereof automatically by computing machine, remain to and preserve to concentrate, these data are respectively as learning sample and the test sample book of neural network.The learning sample collection is designated as: P
l(t)={ (x
k, f (x
k)), k=1,2 ..., N
l, in the formula: f (x
k) the individual x of the evolution that provides of expression user
kAdaptive value, N
lBe the learning sample number.The test sample book collection obtains the individual evaluation of evolving based on nearest several generations user, can be expressed as: P
u(t)={ (x
m, f (x
m)), m=1,2 ..., N
u, in the formula: f (x
m) the individual x of the evolution that provides of expression user
mAdaptive value, N
uBe the learning sample number.
Step 3032: the training of neural network.The learning algorithm of neural network adopts the BP algorithm among the present invention.
Train according to the learning sample set pair neural network that obtained in the last step.Training error in the training process can be expressed as:
In order to test the generalization ability of the neural network after the training, P
uIn the individual x of evolution
l, l=1,2 ..., N
uAs the input of neural network, can obtain the output of neural network
Consider these outputs and the difference that the user estimates, be defined as follows test error:
e
tMore little, show that the generalization ability of neural network is more strong.Setting the test error threshold value is
When
The time, think that the generalization ability of neural network has reached requirement, can be applicable to the estimation of follow-up evolution individual fitness; Otherwise, think still can not meet the demands, need again to its training, till meeting the demands.
Step 3033: the evolution individual fitness based on agent model is estimated.The user estimates the fatigue that will certainly cause the user to the curtain individuality for a long time, the accuracy that influence is estimated, and at this moment, under the prerequisite of the structure of finishing agent model, the user can select to use agent model evaluation.
Step 4: to the individual operation of evolving of carrying out of evolving.The mode of evolution that has adopted multiple evolution algorithm to combine among the present invention, according to the definite evolution algorithm that should take of the coding form of individuality, corresponding substep is under the different coding form:
Step 401: select the string of binary characters coding form.Under this kind coded system, adopt interactive genetic algorithm to carry out follow-up evolution operation, the evolution process flow diagram is as shown in Figure 4.
Be located in this curtain design problem, each individuality comprises W gene meaning unit, and to each gene meaning unit, its allelic span is [0, w
i-1] (i={1,2 ..., W}).When using interactive genetic algorithm to evolve, adaptive value according to the curtain individuality of being given by the user before, adopt league matches to select to keep more excellent individuality, intersect respectively and mutation operation in W gene meaning unit according to the crossover probability of setting and variation probability, the result of final comprehensive each unit generates new individuality.
Step 402: select the integer coding form.Under this kind coded system, adopt interactive evolution strategy to carry out follow-up evolution operation, the evolution process flow diagram is as shown in Figure 5.
Under the integer coding form, directly with the value of each gene meaning unit in the individuality coding as this unit, the coded combination of W unit is got up to constitute the coding of individuality.According to the related content of evolution strategy, this moment, t was for the i in the population individual x
i(t) can be expressed as:
In the formula
σ (t) desirable 3.0.The user is at first from selecting preference individuality, optional one or more individualities when former generation.Based on selected individuality, target variable x is adopted discrete recombination, tactful factor sigma may is adopted the intermediate value reorganization.In the sudden change link, the sudden change formula that the maximum variation step-length λ that utilizes the user to set formulates is:
In the formula
The random number that N (0,1)-obedience just too distributes;
N
iThe random number of (0,1)-just too distribute at the new obedience of component i
Newly-generated number of individuals should be population scale K, the execution that the process of individual choice, evolution of repeating can promote to evolve.
Step 5: preserve optimal case, finish curtain and evolve.By above process, the user can finally finish the evolution of curtain.This moment, the user can preserve the optimal case that finds.In whole evolutionary process, if the user can select to reinitialize the curtain population satisfied inadequately to current curtain scheme all the time, begins new evolution.
Claims (1)
1. be used for the interactive evolutionary optimization method of curtain design, it is characterized by this method whole evolutionary optimization process is divided into three big steps, specifically comprise following content:
(1) the initialized process of evolution population, the ambiguous preference information to curtain for user's setting, be applied in the generative process of population according to the HSI model by computing machine, comprise: the user can arrange the preference information to the curtain use occasion, this arranges the distribution that influences colourity H, and the user can arrange the preference information to the curtain style, and this arranges the distribution that influences saturation degree S, the user can arrange preference information to the individual influence degree that generates of curtain, and this arranges influences intensity of illumination I;
(2) process of man-machine interaction, quantized the evaluating ability of different collection gesture fuzzy sets, introduce concept of information entropy, regard the fuzzy set of every kind of collection gesture as an independently information source, the signal in the information source regarded as in evaluation language in the fuzzy set, calculate the information entropy of each information source, with the fuzzy set of information entropy maximum standard as a comparison, the evaluation differentiation rate of ambiguity in definition collection is shown below:
In the following formula, S
iExpression collection gesture is the fuzzy set of i;
Quantized user's cognition, be located in the curtain design problem, per generation population K curtain individuality arranged, each individuality comprises several separate parts, each part is called a gene meaning unit, with each gene meaning unit as an information source independently; With
J gene meaning unit representing t i individuality in generation, with
Represent in t generation, the identical phenotypic occurrence number of i j individual gene meaning unit, then t in the information entropy of j gene meaning unit be:
Wherein
Situation with information entropy maximum in the gene meaning unit is standard of comparison, defines user's cognition sharpness d of t j gene meaning unit in generation
j(t) be:
In the formula: H (x)
Max=log
2K,
The user t for the time cognitive sharpness D (t) by the user maximal value of the cognitive sharpness of gene meaning unit and interval that minimum value constitutes are determined that formula table is shown:
(3) interactive process of evolving is by supporting different individual coding forms, utilize suitable evolution algorithm that the curtain population is evolved under the corresponding encoded form, and supports the user to switch coding form to satisfy the requirement to evolving.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110094666 CN102163249B (en) | 2011-04-12 | 2011-04-12 | Interactive evolution optimization method for curtain design |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110094666 CN102163249B (en) | 2011-04-12 | 2011-04-12 | Interactive evolution optimization method for curtain design |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102163249A CN102163249A (en) | 2011-08-24 |
CN102163249B true CN102163249B (en) | 2013-07-10 |
Family
ID=44464474
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201110094666 Expired - Fee Related CN102163249B (en) | 2011-04-12 | 2011-04-12 | Interactive evolution optimization method for curtain design |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102163249B (en) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106097418A (en) * | 2016-06-14 | 2016-11-09 | 江苏师范大学 | Cartoon character face verification method for designing based on Interactive evolutionary algorithm |
CN110495254B (en) * | 2017-01-13 | 2022-04-01 | 路创技术有限责任公司 | Method and apparatus for configuring a load control system |
CN107292230B (en) * | 2017-05-09 | 2020-07-28 | 华南理工大学 | Embedded finger vein identification method based on convolutional neural network and having counterfeit detection capability |
CN107194073B (en) * | 2017-05-24 | 2020-08-28 | 郑州航空工业管理学院 | Fuzzy adaptive value interactive evolution optimization method for indoor wall clock design |
CN107463309B (en) * | 2017-08-21 | 2020-12-25 | 郑州航空工业管理学院 | Entropy maximization criterion interactive evolution optimization method for wallpaper design |
CN109544300A (en) * | 2018-11-27 | 2019-03-29 | 景德镇陶瓷大学 | A kind of design of ceramic products system and method based on interactive genetic algorithm |
CN110197527A (en) * | 2019-05-17 | 2019-09-03 | 广州慧阳信息科技有限公司 | Curtain model display system and method |
CN117649466B (en) * | 2024-01-30 | 2024-05-14 | 深圳市崇宁实业有限公司 | System and method for generating decorative drawing engraving concave patterns based on AI algorithm |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7979365B2 (en) * | 2006-01-31 | 2011-07-12 | The Board Of Trustees Of The University Of Illinois | Methods and systems for interactive computing |
CN101105817A (en) * | 2006-07-12 | 2008-01-16 | 猫王国际开发有限公司 | Garment design aid system and its method |
CN100409244C (en) * | 2006-09-05 | 2008-08-06 | 中山大学 | Intelligent dress style design system based on digital family |
CN101819408B (en) * | 2010-04-16 | 2011-08-31 | 东华大学 | Method for interactively designing differential fiber spinning process based on data drive |
-
2011
- 2011-04-12 CN CN 201110094666 patent/CN102163249B/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
CN102163249A (en) | 2011-08-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102163249B (en) | Interactive evolution optimization method for curtain design | |
CN103984721B (en) | Books individuation search method based on interactive evolutionary optimization | |
CN100421585C (en) | Method for establishing mixed expert system of maintaining cigarette leaf group formulation | |
CN110097609A (en) | A kind of fining embroidery texture moving method based on sample territory | |
CN110222908A (en) | A kind of short-term electricity demand forecasting method based on data-driven | |
CN104408281B (en) | It is a kind of to mix type interactive evolution optimization method for what portable wine pot was designed | |
CN108664730A (en) | A kind of dynamic color design method towards multi-modal industrial products | |
Chao | The fractal artistic design based on interactive genetic algorithm | |
CN113705098A (en) | Air duct heater modeling method based on PCA and GA-BP network | |
Cai | User preference adaptive fitness of interactive genetic algorithm based ceramic disk pattern generation method | |
McCormack et al. | Growing and Evolving 3-D Prints | |
Yang et al. | An Intelligent Designing Approach for Personalized Products Using Product Gene | |
Maeda et al. | Rhythm generation method for automatic musical composition using genetic algorithm | |
Sasaki et al. | Portfolio optimization by fuzzy interactive genetic algorithm | |
Önduygu | Graphagos: evolutionary algorithm as a model for the creative process and as a tool to create graphic design products | |
CN106097418A (en) | Cartoon character face verification method for designing based on Interactive evolutionary algorithm | |
Schein et al. | Structural analysis as driver in surface-based design approaches | |
Ota et al. | Kansei clothing retrieval system using features extracted by autoencoder | |
KR101807175B1 (en) | Light theraphy providing method and server | |
Maeda et al. | Automatic generation method of twelve tone row for musical composition used genetic algorithm | |
Wen et al. | Study on the aesthetic evaluation method of applying symmetry analysis in industrial product styling design | |
Zeng et al. | A memetic algorithm for Chinese traditional music composition | |
CN117689450B (en) | Digital marketing system based on big data | |
Nuraliev et al. | The use of methods for optimizing the genetic algorithm for the colors of a fractal structure pattern in carpet design | |
Gan et al. | Application of Isomorphic Block Decoration Based on VR and Internet of Things in Fashion Design |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20130710 Termination date: 20160412 |