R-GP实现错误
所以我对R很陌生。我在用Mathematica导入数据时遇到了一些问题,所以我决定做一个切换,因为R更适合分析。我正在构建一些机器学习技术来对现在可以导入的数据进行分析。这是一个遗传编程实现,完成后应该对一些数据进行符号回归。除了错误之外,脚本应该差不多完成了(我需要对composition操作符进行编程,使division受保护,并完成基本函数的列表)。我之前在编写脚本时遇到了一个问题,该问题已解决()。我已经调试了大约一天的脚本,我已经没有主意了 我的错误消息是:R-GP实现错误,r,genetic-programming,R,Genetic Programming,所以我对R很陌生。我在用Mathematica导入数据时遇到了一些问题,所以我决定做一个切换,因为R更适合分析。我正在构建一些机器学习技术来对现在可以导入的数据进行分析。这是一个遗传编程实现,完成后应该对一些数据进行符号回归。除了错误之外,脚本应该差不多完成了(我需要对composition操作符进行编程,使division受保护,并完成基本函数的列表)。我之前在编写脚本时遇到了一个问题,该问题已解决()。我已经调试了大约一天的脚本,我已经没有主意了 我的错误消息是: Error in make
Error in makeStrName(nextGen) : object 'nextGen' not found
>
> #Print the string versions of the five functions with the lowest RMSE evolved.
> byRMSEList<-sortByRMSE(populationsBestTenStr)
Error: object 'totalTwo' not found
> for(i in 1:5)
+ {
+ byRMSEList[[i]]
+ }
Error: object 'byRMSEList' not found
makeStrName(nextGen)中出现错误:找不到对象“nextGen”
>
>#以最低的RMSE打印五个函数的字符串版本。
>(1:5中的i)的列表
+ {
+列表[[i]]
+ }
错误:找不到对象“byRMSEList”
这是我的剧本。我目前正在使用RStudio。感谢您抽出时间提供帮助:
library("datasets")
operators<-list("+","*","-","/","o")
funcs<-list("x","log(x)","sin(x)","cos(x)","tan(x)")
#Allows me to map a name to each element in a numerical list.
makeStrName<-function(listOfItems)
{
for(i in 1:length(listOfItems))
{
names(listOfItems)[i]=paste("x",i,sep="")
}
return(listOfItems)
}
#Allows me to replace each random number in a vector with the corresponding
#function in a list of functions.
mapFuncList<-function(funcList,rndNumVector)
{
for(i in 1:length(funcList))
{
rndNumVector[rndNumVector==i]<-funcList[i]
}
return(rndNumVector)
}
#Will generate a random function from the list of functions and a random sample.
generateOrganism<-function(inputLen,inputSeed, funcList)
{
set.seed(inputSeed)
rnd<-sample(1:length(funcList),inputLen,replace=T)
Org<-mapFuncList(funcList,rnd)
return(Org)
}
#Will generate a series of "Organisms"
genPopulation<-function(popSize,initialSeed,initialSize,functions)
{
population<-list()
for(i in 1:popSize)
{
population <- c(population,generateOrganism(initialSize,initialSeed+i,functions))
}
populationWithNames<-makeStrName(population)
return(populationWithNames)
}
#Turns the population of functions (which are actually strings in "") into
#actual functions. (i.e. changes the mode of the list from string to function).
funCreator<-function(snippet)
{
txt=snippet
function(x)
{
exprs <- parse(text = txt)
eval(exprs)
}
}
#Applies a fitness function to the population. Puts the best organism in
#the hallOfFame.
evalPopulation<-function(populationFuncList, inputData, outputData, populationStringList)
{
#rmse <- sqrt( mean( (sim - obs)^2))
for(i in 1:length(populationStringList))
{
stringFunc<-populationStringList[[i]]
total<-list(mode="numeric",length=length(inputData))
topTenPercentFunctionList<-list()
topTenPercentRMSEList<-list()
topTenPercentStringFunctionList<-list()
tempFunc<-function(x){x}
for(z in 1:length(inputData))
{
total<-c(total,(abs(populationFuncList[[i]](inputData[[z]])-outputData[[z]])))
tempFunc<-populationFuncList[[i]]
}
rmse<-sqrt(mean(total*total))
topTenPercentVal<-length(populationFuncList)*0.1
if(length(topTenPercentFunctionList)<topTenPercentVal||RMSE<min(topTenPercentRMSEList))
{
topTenPercentStringFunctionList<-c(topTenPercentStringFunctionList,stringFunc)
topTenPercentRMSEList<-c(topTenPercentRMSEList, rmse)
topTenPercentFunctionList<-c(topTenPercentFunctionList, tempFunc)
}
}
return(topTenPercentStringFunctionList)
}
#Get random operator
getRndOp<-function(seed)
{
set.seed(seed)
rndOpNum<-sample(1:length(operators),1,replace=T)
operation<-operators[[rndOpNum]]
return(operation)
}
#Mutation Operators
#This attaches a new appendage to an organism
endNodeMutation<-function(strFunc,seed)
{
op<-getRndOp(seed)
strFunc<-c(strFunc,op)
newAppendage<-generateOrganism(1,seed+2,funcs)
strFunc<-c(strFunc,newAppendage)
return(strFunc)
}
#This is a mutation that occurs at a random locaiton in an organism
rndNodeMutation<-function(strFunc,seed,secondSeed)
{
op<-getRndOp(seed)
halfStrFunc<-((length(strFunc))/2)
set.seed(seed)
randomStart<-sample(1:halfStrFunc,1,replace=T)
set.seed(secondSeed)
randomEnd<-2*(sample(1:length(halfStrFunc),1,replace=T))
strFuncUpdate<-substr(strFunc,randomStart,randomEnd)
strFuncUpdate<-c(strFuncUpdate,op)
newAppendage<-generateOrganism(1,seed+2,funcs)
strFuncUpdate<-c(strFuncUpdate,newAppendage)
return(strFuncUpdate)
}
#Crossover Operators
#Crossover operator that attaches otherStrFunc to strFunc at the endpoint of strFunc
crossoverConcatenationOperator<-function(strFunc,otherStrFunc)
{
newStrFunc<-c(strFunc,otherStrFunc)
return(newStrFunc)
}
#Crossover Operation that starts and ends at random points in the concatenation
randomCrossoverOperator<-function(strFunc,otherStrFunc,seed,secondSeed)
{
set.seed(seed)
wholeLength<-(length(strFunc)+length(otherStrFunc))
startRndNum<-sample(1:length(strFunc),1,replace=T)
set.seed(secondSeed)
endRndNum<-sample(length(strFunc):wholeLength,1,replace=T)
concatenatedFunc<-c(strFunc,otherStrFunc)
newFunc<-substr(concatenatedFunc,startRndNum,endRndNum)
return(newFunc)
}
evolve<-function(strFuncList,tenPercentStrFuncList)
{
#Detach the bottom ninety percent to the top ten percent
evolveList<-substr(strFuncList,length(tenPercentStrFuncList),length(strFuncList))
#Get sizes. Will use a random mutation, then random crossover, then
#random mutation, then random crossover at percentages with 0.05,0.45,0.05,0.45
#respectively
size<-length(evolveList)
mutateNum<-0.1*size
crossoverNum<-0.9*size
halfMutateNum<-0.05*size
halfCrossoverNum<-0.45*size
roundedMutateNum<-floor(mutateNum)
roundedCrossoverNum<-floor(crossoverNum)
roundedHalfMutateNum<-floor(halfMutateNum)
roundedHalfCrossoverNum<-floor(halfCrossoverNum)
#Calls the functions for those percentage of organisms in that order
for(i in 1:roundedHalfMutateNum)
{
set.seed(i)
rndOne<-sample(0:1000,1,replace=T)
set.seed(i+10000)
rndTwo<-sample(0:10000,1,replace=T)
newFunc<-rndNodeMutation(evolveList[[i]],rndOne,rndTWo)
evolveList[[i]]<-newFunc
}
for (i in roundedHalfMutateNum:(roundedHalfCrossoverNum+roundedHalfMutateNum))
{
set.seed(i)
rndOne<-sample(0:1000,1,replace=T)
set.seed(i+10000)
rndTwo<-sample(0:10000,1,replace=T)
newFunc<-rndCrossoverOperation(evolveList[[i]],evolveList[[i+1]],rndOne,rndTwo)
firstSubstr<-substr(evolveList,1,i-1)
secondSubstr<-substr(evolveLIst,i+2,length(evolveList))
halfSubstr<-c(firstSubstr,newFunc)
evolveList<-c(halfSubstr,secondSubstr)
}
for(i in (roundedHalfCrossoverNum+roundedHalfMutateNum):(roundedHalfCrossoverNum+roundedMutateNum))
{
set.seed(i)
rndOne<-sample(0:1000,1,replace=T)
set.seed(i+10000)
rndTwo<-sample(0:10000,1,replace=T)
newFunc<-rndNodeMutation(evolveList[[i]],rndOne,rndTWo)
evolveList[[i]]<-newFunc
}
for(i in (roundedHalfCrossoverNum+roundedMutateNum):(roundedCrossoverNum+roundedHalfMutateNum))
{
set.seed(i)
rndOne<-sample(0:1000,1,replace=T)
set.seed(i+10000)
rndTwo<-sample(0:10000,1,replace=T)
newFunc<-rndCrossoverOperation(evolveList[[i]],evolveList[[i+1]],rndOne,rndTwo)
firstSubstr<-substr(evolveList,1,i-1)
secondSubstr<-substr(evolveLIst,i+2,length(evolveList))
halfSubstr<-c(firstSubstr,newFunc)
evolveList<-c(halfSubstr,secondSubstr)
}
}
#Calculates the root mean squared of the functions in a string list.
#Then sorts the list by RMSE.
sortByRMSE<-function(strL)
{
for (z in 1:length(strL))
{
for(i in 1:length(strL))
{
nonStrFuncList<-lapply(strL,function(x){funCreator(x)})
totalTwo<-c(totalTwo,(abs(nonStrFuncList[[z]](inputData[[i]])-outputData[[i]])))
}
rmse<-sqrt(mean(totalTwo*totalTwo))
strFuncsLists<-strL[order(sapply(strL, '[[', rmse))]
}
return(strFuncsLists)
}
#Data, Output Goal
desiredFuncOutput<-list(1,4,9,16,25)
dataForInput<-list(1,2,3,4,5)
#Generate Initial Population
POpulation<-genPopulation(4,1,1,funcs)
POpulationFuncList <- lapply(setNames(POpulation,names(POpulation)),function(x){funCreator(x)})
#Get and save top ten percent in bestDudes
bestDudes<-evalPopulation(POpulationFuncList,dataForInput,desiredFuncOutput,POpulation)
#Evolve the rest
NewBottomNinetyPercent<-evolve(POpulation,bestDudes)
#Concatenate the two to make a new generation
nextGen<-c(bestDudes,NewBottomNinetyPercent)
#Declare lists,
populationsBestTenStr<-list()
populationsFuncList<-list()
#Run ten generations.
for(i in 1:10)
{
nextGen<-makeStrName(nextGen)
populationsFuncList<-lapply(setNames(nextGen,names(nextGen)),function(x){funCreator(x)})
populationsBestTenStr<-evalPopulation(populationsFuncList,dataForInput,desiredFuncOutput,nextGen)
nextGen<-evolve(populations,populationsBestTenStr)
}
#Print the string versions of the five functions with the lowest RMSE evolved.
byRMSEList<-sortByRMSE(populationsBestTenStr)
for(i in 1:5)
{
byRMSEList[[i]]
}
库(“数据集”)
操作员库(“数据集”)
操作员库(“数据集”)
操作员我的密码里有。看见“#Data,Output Goal desiredFuncOutputIts正好位于我发布的代码中对evalPopulation的调用之上。@初学者polymath好的,我修复了另一个bug,现在它可以通过evolve
工作,但我不知道你想用substr
做什么。它似乎任意截断了总体
值。我看到了那句话的评论,但我真的不明白。你能帮我吗?我必须去参加一个会议,但今天晚些时候我会再讨论这个问题。@KnowerPolymath它只是应该从人口中删除花花公子的吗?是的。基本上,这个想法是保持前10%的人(最好的人)不变,然后在后90%的人身上使用交叉和变异算子。在随机变异和随机交叉函数中,我使用它来进化种群,同时避免膨胀。因此,它(理想情况下)选择随机位置来更改内容,并丢弃起始和结束位置之外的内容。我能问一下为什么我的substr坏了,这样我就不会再犯同样的错误了?我查看了这些更改,但没有弄清楚。我的代码中确实有这些更改。看见“#Data,Output Goal desiredFuncOutputIts正好位于我发布的代码中对evalPopulation的调用之上。@初学者polymath好的,我修复了另一个bug,现在它可以通过evolve
工作,但我不知道你想用substr
做什么。它似乎任意截断了总体
值。我看到了那句话的评论,但我真的不明白。你能帮我吗?我必须去参加一个会议,但今天晚些时候我会再讨论这个问题。@KnowerPolymath它只是应该从人口中删除花花公子的吗?是的。基本上,这个想法是保持前10%的人(最好的人)不变,然后在后90%的人身上使用交叉和变异算子。在随机变异和随机交叉函数中,我使用它来进化种群,同时避免膨胀。因此,它(理想情况下)选择随机位置来更改内容,并丢弃起始和结束位置之外的内容。我能问一下为什么我的substr坏了,这样我就不会再犯同样的错误了?我看了一下这些变化,想不出来。
library("datasets")
operators<-list("+","*","-","/","o")
funcs<-list("x","log(x)","sin(x)","cos(x)","tan(x)")
# Fixed:
# evolveLIst inconsistently typed as evolveList
# rndCrossoverOperation inconsistently typed as randomCrossoverOperator
# rndTWo inconsistently typed as rndTwo
# broken substr
# broken condition leading to for(i in 1:0)
# misc. others
#Allows me to map a name to each element in a numerical list.
makeStrName<-function(listOfItems)
{
for(i in 1:length(listOfItems))
{
names(listOfItems)[i]=paste("x",i,sep="")
}
return(listOfItems)
}
#Allows me to replace each random number in a vector with the corresponding
#function in a list of functions.
mapFuncList<-function(funcList,rndNumVector)
{
for(i in 1:length(funcList))
{
rndNumVector[rndNumVector==i]<-funcList[i]
}
return(rndNumVector)
}
#Will generate a random function from the list of functions and a random sample.
generateOrganism<-function(inputLen,inputSeed, funcList)
{
set.seed(inputSeed)
rnd<-sample(1:length(funcList),inputLen,replace=T)
Org<-mapFuncList(funcList,rnd)
return(Org)
}
#Will generate a series of "Organisms"
genPopulation<-function(popSize,initialSeed,initialSize,functions)
{
population<-list()
for(i in 1:popSize)
{
population <- c(population,generateOrganism(initialSize,initialSeed+i,functions))
}
populationWithNames<-makeStrName(population)
return(populationWithNames)
}
#Turns the population of functions (which are actually strings in "") into
#actual functions. (i.e. changes the mode of the list from string to function).
funCreator<-function(snippet)
{
txt=snippet
function(x)
{
exprs <- parse(text = txt)
eval(exprs)
}
}
#Applies a fitness function to the population. Puts the best organism in
#the hallOfFame.
evalPopulation<-function(populationFuncList=POpulationFuncList, inputData=dataForInput, outputData=desiredFuncOutput,
populationStringList=POpulation)
{
#rmse <- sqrt( mean( (sim - obs)^2))
for(i in 1:length(populationStringList))
{
stringFunc<-populationStringList[[i]]
total<-as.numeric(length(inputData))
topTenPercentFunctionList<-list()
topTenPercentRMSEList<-list()
topTenPercentStringFunctionList<-list()
tempFunc<-function(x){x}
for(z in 1:length(inputData))
{
total<-c(total,(abs(populationFuncList[[i]](inputData[[z]])-outputData[[z]])))
tempFunc<-populationFuncList[[i]]
}
rmse<-sqrt(mean(total^2))
topTenPercentVal<-length(populationFuncList)*0.1
if(length(topTenPercentFunctionList)<topTenPercentVal||RMSE<min(topTenPercentRMSEList))
{
topTenPercentStringFunctionList<-c(topTenPercentStringFunctionList,stringFunc)
topTenPercentRMSEList<-c(topTenPercentRMSEList, rmse)
topTenPercentFunctionList<-c(topTenPercentFunctionList, tempFunc)
}
}
return(topTenPercentStringFunctionList)
}
#Get random operator
getRndOp<-function(seed)
{
set.seed(seed)
rndOpNum<-sample(1:length(operators),1,replace=T)
operation<-operators[[rndOpNum]]
return(operation)
}
#Mutation Operators
#This attaches a new appendage to an organism
endNodeMutation<-function(strFunc,seed)
{
op<-getRndOp(seed)
strFunc<-c(strFunc,op)
newAppendage<-generateOrganism(1,seed+2,funcs)
strFunc<-c(strFunc,newAppendage)
return(strFunc)
}
#This is a mutation that occurs at a random locaiton in an organism
rndNodeMutation<-function(strFunc,seed,secondSeed)
{
op<-getRndOp(seed)
halfStrFunc<-((length(strFunc))/2)
set.seed(seed)
randomStart<-sample(1:halfStrFunc,1,replace=T)
set.seed(secondSeed)
randomEnd<-2*(sample(1:length(halfStrFunc),1,replace=T))
strFuncUpdate<-substr(strFunc,randomStart,randomEnd)
strFuncUpdate<-c(strFuncUpdate,op)
newAppendage<-generateOrganism(1,seed+2,funcs)
strFuncUpdate<-c(strFuncUpdate,newAppendage)
return(strFuncUpdate)
}
#Crossover Operators
#Crossover operator that attaches otherStrFunc to strFunc at the endpoint of strFunc
crossoverConcatenationOperator<-function(strFunc,otherStrFunc)
{
newStrFunc<-c(strFunc,otherStrFunc)
return(newStrFunc)
}
#Crossover Operation that starts and ends at random points in the concatenation
rndCrossoverOperation<-function(strFunc,otherStrFunc,seed,secondSeed) # fixed function name
{
set.seed(seed)
wholeLength<-(length(strFunc)+length(otherStrFunc))
startRndNum<-sample(1:length(strFunc),1,replace=T)
set.seed(secondSeed)
endRndNum<-sample(length(strFunc):wholeLength,1,replace=T)
concatenatedFunc<-c(strFunc,otherStrFunc)
newFunc<-substr(concatenatedFunc,startRndNum,endRndNum)
return(newFunc)
}
evolve<-function(strFuncList=POpulation,tenPercentStrFuncList=bestDudes)
{
#Detach the bottom ninety percent to the top ten percent
evolveList<-strFuncList[!strFuncList %in% tenPercentStrFuncList] # fixed broken substring
#Get sizes. Will use a random mutation, then random crossover, then
#random mutation, then random crossover at percentages with 0.05,0.45,0.05,0.45
#respectively
size<-length(evolveList)
mutateNum<-0.1*size
crossoverNum<-0.9*size
halfMutateNum<-0.05*size
halfCrossoverNum<-0.45*size
roundedMutateNum<-floor(mutateNum)
roundedCrossoverNum<-floor(crossoverNum)
roundedHalfMutateNum<-floor(halfMutateNum)
roundedHalfCrossoverNum<-floor(halfCrossoverNum)
#Calls the functions for those percentage of organisms in that order
if(roundedHalfMutateNum < 1) roundedHalfMutateNum <- 1
for(i in 1:roundedHalfMutateNum)
{
set.seed(i)
rndOne<-sample(0:1000,1,replace=T)
set.seed(i+10000)
rndTwo<-sample(0:10000,1,replace=T)
newFunc<-rndNodeMutation(evolveList[[i]],rndOne,rndTwo) # fixed case
evolveList[[i]]<-newFunc
}
for (i in roundedHalfMutateNum:(roundedHalfCrossoverNum+roundedHalfMutateNum))
{
set.seed(i)
rndOne<-sample(0:1000,1,replace=T)
set.seed(i+10000)
rndTwo<-sample(0:10000,1,replace=T)
newFunc<-rndCrossoverOperation(evolveList[[i]],evolveList[[i+1]],rndOne,rndTwo)
firstSubstr<-substr(evolveList,1,i-1)
secondSubstr<-substr(evolveList,i+2,length(evolveList))
halfSubstr<-c(firstSubstr,newFunc)
evolveList<-c(halfSubstr,secondSubstr)
}
for(i in (roundedHalfCrossoverNum+roundedHalfMutateNum):(roundedHalfCrossoverNum+roundedMutateNum))
{
set.seed(i)
rndOne<-sample(0:1000,1,replace=T)
set.seed(i+10000)
rndTwo<-sample(0:10000,1,replace=T)
newFunc<-rndNodeMutation(evolveList[[i]],rndOne,rndTwo)
evolveList[[i]]<-newFunc
}
for(i in (roundedHalfCrossoverNum+roundedMutateNum):(roundedCrossoverNum+roundedHalfMutateNum))
{
set.seed(i)
rndOne<-sample(0:1000,1,replace=T)
set.seed(i+10000)
rndTwo<-sample(0:10000,1,replace=T)
newFunc<-rndCrossoverOperation(evolveList[[i]],evolveList[[i+1]],rndOne,rndTwo)
firstSubstr<-substr(evolveList,1,i-1)
secondSubstr<-substr(evolveList,i+2,length(evolveList))
halfSubstr<-c(firstSubstr,newFunc)
evolveList<-c(halfSubstr,secondSubstr)
}
}
#Calculates the root mean squared of the functions in a string list.
#Then sorts the list by RMSE.
sortByRMSE<-function(strL)
{
for (z in 1:length(strL))
{
for(i in 1:length(strL))
{
nonStrFuncList<-lapply(strL,function(x){funCreator(x)})
totalTwo<-c(totalTwo,(abs(nonStrFuncList[[z]](inputData[[i]])-outputData[[i]])))
}
rmse<-sqrt(mean(totalTwo*totalTwo))
strFuncsLists<-strL[order(sapply(strL, '[[', rmse))]
}
return(strFuncsLists)
}
#Data, Output Goal
desiredFuncOutput<-list(1,4,9,16,25)
dataForInput<-list(1,2,3,4,5)
#Generate Initial Population
POpulation<-genPopulation(4,1,1,funcs)
POpulationFuncList <- lapply(setNames(POpulation,names(POpulation)),function(x){funCreator(x)})
#Get and save top ten percent in bestDudes
bestDudes<-evalPopulation(POpulationFuncList,dataForInput,desiredFuncOutput,POpulation)
#Evolve the rest
NewBottomNinetyPercent<-evolve(POpulation,bestDudes)
#Concatenate the two to make a new generation
nextGen<-c(bestDudes,NewBottomNinetyPercent)
#Declare lists,
populationsBestTenStr<-list()
populationsFuncList<-list()
#Run ten generations.
for(i in 1:10)
{
nextGen<-makeStrName(nextGen)
populationsFuncList<-lapply(setNames(nextGen,names(nextGen)),function(x){funCreator(x)})
populationsBestTenStr<-evalPopulation(populationsFuncList,dataForInput,desiredFuncOutput,nextGen)
nextGen<-evolve(populations,populationsBestTenStr)
}
#Print the string versions of the five functions with the lowest RMSE evolved.
byRMSEList<-sortByRMSE(populationsBestTenStr)
for(i in 1:5)
{
byRMSEList[[i]]
}