[Biomod-commits] biomod2 / Maxent

[Damien Georges]

Dear Sami,

I just change biomod2 according to your post. Maxent temp directories 
are now created into the concerned species directory (resp. projection 
directory) to remove synchronization issues using MAXENT.

I guess you can now access saved objects within lapply-wrapper simply 
picking them into species directories.

All the best,

Damien

On 22/08/2012 23:42, Sami Domisch wrote:
> Dear Damien and Wilfried,
>
> first of all many thanks for providing the new biomod2 package! I 
> played a bit with the code in the multi species manual, and tried to 
> include MAXENT within the sfLapply-function in snowfall. However, 
> snowfall returns an error, stating that it "cannot open the 
> connection". After turning MAXENT off, the models run fine.
>
> > Error in checkForRemoteErrors(val) :
> >  9 nodes produced errors; first error: cannot open the connection
>
> I guess that this error derives from the temporary MAXENT-folder 
> created in setwd()? The single nodes try to access that folder which 
> is deleted after each run?
>
> Is it possible to place the MaxentTmpData-folder within each species' 
> folder during the model run? And related to that, is it possible to 
> save the modelling results (save.image) also in each species' folder? 
> In other words, how can I access the single species folder within the 
> lapply-wrapper?
>
> Again, thanks a lot for this nice tool!
> Best, Sami
>
>
> #----------------------------------
> library(biomod2)
> ### Load data
> mySpeciesOcc <- read.table("mySpeciesOcc.txt", h=T)
> coord = read.table("coordinates.txt", h=T)
> mySpeciesOcc = cbind(coord, mySpeciesOcc)
>
> ### Fill 'NAs'
> mySpeciesOcc[mySpeciesOcc==0] <- NA
>
> require(raster)
>
> myExpl = stack("present/bio1",
>                "present/bio12",
>                 "present/bio15")
>
> sp.names <- names(mySpeciesOcc[3:12])
>
> ### Create function to run models within 'lapply'
>
> MyBiomodSF <- function(sp.n){
>   cat('/n',sp.n,'modeling...')
>   ### definition of data for this run
>   ## i.e keep only the column of our species
>   myResp <- as.numeric(mySpeciesOcc[,sp.n])
>   #myResp <- as.numeric(mySpeciesOcc[,sp.n])
>   # get NAs id
>   na.id <- which(is.na(myResp))
>   # remove NAs to enforce PA sampling to be done on explanatory rasters
>   myResp <- myResp[-na.id]
>  # myRespCoord =mySpeciesOcc[1:2]
>   myRespCoord = mySpeciesOcc[-na.id,c('long','lat')]
>   myRespName = sp.n
>   ### Initialisation
>   myBiomodData <- BIOMOD_FormatingData(resp.var = myResp,
>                                        expl.var = myExpl,
>                                        resp.xy = myRespCoord,
>                                        resp.name = myRespName,
>                                        PA.nb.rep = 1,
>                                        PA.nb.absences = 10000,
>                                        PA.strategy = 'random')
>
>   ### Options definition
>   myBiomodOption <- BIOMOD_ModelingOptions()
>   ### Modelling
>   myBiomodModelOut <- BIOMOD_Modeling(
>     myBiomodData,
>     #models = 
> c('SRE','CTA','RF','MARS','FDA','MAXENT','GLM','GAM','GBM','ANN'),
>     models = c('SRE','RF','MAXENT'),
>     models.options = myBiomodOption,
>     NbRunEval=3,
>     DataSplit=70,
>     Yweights=NULL,
>     VarImport=3,
>     models.eval.meth = c('TSS','ROC'),
>     SaveObj = TRUE,
>     rescal.all.models = TRUE)
>
>   #   ### Building ensemble-models
>
>     myBiomodEM <- BIOMOD_EnsembleModeling(
>       modeling.output = myBiomodModelOut,
>       chosen.models = 'all',
>       eval.metric = c('TSS'),
>       eval.metric.quality.threshold = c(0.4), # e.g. 0.4
>       prob.mean = T,
>       prob.cv = T,
>       prob.ci = T,
>       prob.ci.alpha = 0.05,
>       prob.median = T,
>       committee.averaging = T,
>       prob.mean.weight = T,
>       prob.mean.weight.decay = 'proportional' )
>
>   ### Do projections on current variable
>   myBiomomodProj <- BIOMOD_Projection(
>     modeling.output = myBiomodModelOut,
>     new.env = myExpl,
>     proj.name = 'current',
>     selected.models = 'all',
>     binary.meth= 'ROC',
>     compress = 'xz',
>     clamping.mask = F)
>
>   ### Do ensemble-models projections on current varaiable
>   myBiomodEF <- BIOMOD_EnsembleForecasting(
>     projection.output = myBiomomodProj,
>     EM.output = myBiomodEM,
>     binary.meth = 'TSS',
>     total.consensus = TRUE)
>
> }
>
> #----------------------------------
>
>
> ### Init snowfall
> library(snowfall)
> sfInit(parallel=TRUE, cpus=8, type="SOCK")
>
> ### Export packages
> sfLibrary('biomod2', character.only=TRUE)
> sfLibrary('rgdal', character.only=TRUE)
>
> ### Export data to nodes
> sfExportAll()
>
> ### Check the items each node has loaded so far:
> sfClusterEval(ls())
>
> ### Do the run
> start <- Sys.time()
> result <- sfClusterApplyLB(sp.names, MyBiomodSF)
> Sys.time()-start
>
> ### Remove all variables from nodes
> # sfRemoveAll(except=c("sp.names", "MyBiomodSF"))
>
> ### Stop snowfall
> sfStop()
> #----------------------------------
>
>
> _______________________________________________
> Biomod-commits mailing list
> Biomod-commits at lists.r-forge.r-project.org
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/biomod-commits 
>