R/pfilter.R

In pomp: Statistical Inference for Partially Observed Markov Processes

## particle filtering codes

setClass(
         "pfilterd.pomp",
         contains="pomp",
         slots=c(
           pred.mean="array",
           pred.var="array",
           filter.mean="array",
           paramMatrix="array",
           eff.sample.size="numeric",
           cond.loglik="numeric",
           saved.states="list",
           saved.params="list",
           seed="integer",
           Np="integer",
           tol="numeric",
           nfail="integer",
           loglik="numeric"
           ),
         prototype=prototype(
           pred.mean=array(data=numeric(0),dim=c(0,0)),
           pred.var=array(data=numeric(0),dim=c(0,0)),
           filter.mean=array(data=numeric(0),dim=c(0,0)),
           paramMatrix=array(data=numeric(0),dim=c(0,0)),
           eff.sample.size=numeric(0),
           cond.loglik=numeric(0),
           saved.states=list(),
           saved.params=list(),
           seed=as.integer(NA),
           Np=as.integer(NA),
           tol=as.double(NA),
           nfail=as.integer(NA),
           loglik=as.double(NA)
           )
         )

pfilter.internal <- function (object, params, Np,
                              tol, max.fail,
                              pred.mean, pred.var, filter.mean,
                              cooling, cooling.m, .mif2 = FALSE,
                              .rw.sd, seed, verbose,
                              save.states, save.params,
                              .transform,
                              .getnativesymbolinfo = TRUE) {

  pompLoad(object)

  ptsi.for <- gnsi.rproc <- gnsi.dmeas <- as.logical(.getnativesymbolinfo)
  mif2 <- as.logical(.mif2)
  transform <- as.logical(.transform)
  
  if (missing(seed)) seed <- NULL
  if (!is.null(seed)) {
    if (!exists(".Random.seed",where=.GlobalEnv)) { # need to initialize the RNG
      runif(1)
    }
    save.seed <- get(".Random.seed",pos=.GlobalEnv)
    set.seed(seed)
  }
  
  if (length(params)==0)
    stop(sQuote("pfilter")," error: ",sQuote("params")," must be specified",call.=FALSE)
  
  if (missing(tol))
    stop(sQuote("pfilter")," error: ",sQuote("tol")," must be specified",call.=FALSE)
  
  one.par <- FALSE
  times <- time(object,t0=TRUE)
  ntimes <- length(times)-1
  
  if (missing(Np))
    Np <- NCOL(params)
  if (is.function(Np)) {
    Np <- try(
              vapply(seq.int(from=0,to=ntimes,by=1),Np,numeric(1)),
              silent=FALSE
              )
    if (inherits(Np,"try-error"))
      stop("if ",sQuote("Np")," is a function, it must return a single positive integer",call.=FALSE)
  }
  if (length(Np)==1)
    Np <- rep(Np,times=ntimes+1)
  else if (length(Np)!=(ntimes+1))
    stop(sQuote("Np")," must have length 1 or length ",ntimes+1,call.=FALSE)
  if (any(Np<=0))
    stop("number of particles, ",sQuote("Np"),", must always be positive",call.=FALSE)
  if (!is.numeric(Np))
    stop(sQuote("Np")," must be a number, a vector of numbers, or a function",call.=FALSE)
  Np <- as.integer(Np)
  
  if (is.null(dim(params))) {
    one.par <- TRUE               # there is only one parameter vector
    coef(object) <- params        # set params slot to the parameters
    params <- matrix(
                     params,
                     nrow=length(params),
                     ncol=Np[1L],
                     dimnames=list(
                       names(params),
                       NULL
                       )
                     )
  }
  paramnames <- rownames(params)
  if (is.null(paramnames))
    stop(sQuote("pfilter")," error: ",sQuote("params")," must have rownames",call.=FALSE)
  
  x <- init.state(
                  object,
                  params=if (transform) {
                    partrans(object,params,dir="fromEstimationScale",
                             .getnativesymbolinfo=ptsi.for)
                  } else {
                    params
                  }
                  )
  statenames <- rownames(x)
  nvars <- nrow(x)
  ptsi.for <- FALSE
  
  ## set up storage for saving samples from filtering distributions
  if (save.states) {
    xparticles <- setNames(vector(mode="list",length=ntimes),time(object))
  } else {
    xparticles <- list()
  }
  if (save.params) {
    pparticles <- setNames(vector(mode="list",length=ntimes),time(object))
  } else {
    pparticles <- list()
  }
  
  random.walk <- !missing(.rw.sd)
  if (random.walk) {
    rw.names <- names(.rw.sd)
    if (is.null(rw.names)||!is.numeric(.rw.sd))
      stop(sQuote("pfilter")," error: ",sQuote(".rw.sd")," must be a named vector",call.=FALSE)
    if (!all(rw.names%in%paramnames))
      stop(
           sQuote("pfilter")," error: the rownames of ",
           sQuote("params")," must include all of the names of ",
           sQuote(".rw.sd"),"",call.=FALSE
           )
    sigma <- .rw.sd
  } else {
    rw.names <- character(0)
    sigma <- NULL
  }
  
  loglik <- rep(NA,ntimes)
  eff.sample.size <- numeric(ntimes)
  nfail <- 0
  npars <- length(rw.names)
  
  ## set up storage for prediction means, variances, etc.
  if (pred.mean)
    pred.m <- matrix(
                     data=0,
                     nrow=nvars+npars,
                     ncol=ntimes,
                     dimnames=list(
                       variable=c(statenames,rw.names),
                       time=time(object))
                     )
  else
    pred.m <- array(data=numeric(0),dim=c(0,0))
  
  if (pred.var)
    pred.v <- matrix(
                     data=0,
                     nrow=nvars+npars,
                     ncol=ntimes,
                     dimnames=list(
                       variable=c(statenames,rw.names),
                       time=time(object))
                     )
  else
    pred.v <- array(data=numeric(0),dim=c(0,0))
  
  if (filter.mean)
    if (random.walk)
      filt.m <- matrix(
                       data=0,
                       nrow=nvars+length(paramnames),
                       ncol=ntimes,
                       dimnames=list(
                         variable=c(statenames,paramnames),
                         time=time(object))
                       )
    else
      filt.m <- matrix(
                       data=0,
                       nrow=nvars,
                       ncol=ntimes,
                       dimnames=list(
                         variable=statenames,
                         time=time(object))
                       )
  else
    filt.m <- array(data=numeric(0),dim=c(0,0))

  for (nt in seq_len(ntimes)) {
    
    if (mif2) {	  
      cool.sched <- cooling(nt=nt,m=cooling.m)
      sigma1 <- sigma*cool.sched$alpha
    } else {
      sigma1 <- sigma
    }
    
    ## transform the parameters if necessary
    if (transform) tparams <- partrans(object,params,dir="fromEstimationScale",
                                       .getnativesymbolinfo=ptsi.for)
    ptsi.for <- FALSE
    
    ## advance the state variables according to the process model
    X <- try(
             rprocess(
                      object,
                      xstart=x,
                      times=times[c(nt,nt+1)],
                      params=if (transform) tparams else params,
                      offset=1,
                      .getnativesymbolinfo=gnsi.rproc
                      ),
             silent=FALSE
             )
    if (inherits(X,'try-error'))
      stop(sQuote("pfilter")," error: process simulation error",call.=FALSE)
    gnsi.rproc <- FALSE
    
    if (pred.var) { ## check for nonfinite state variables and parameters
      problem.indices <- unique(which(!is.finite(X),arr.ind=TRUE)[,1L])
      if (length(problem.indices)>0) {  # state variables
        stop(
             sQuote("pfilter")," error: non-finite state variable(s): ",
             paste(rownames(X)[problem.indices],collapse=', '),
             call.=FALSE
             )
      }
      if (random.walk) { # parameters (need to be checked only if 'random.walk=TRUE')
        problem.indices <- unique(which(!is.finite(params[rw.names,,drop=FALSE]),arr.ind=TRUE)[,1L])
        if (length(problem.indices)>0) {
          stop(
               sQuote("pfilter")," error: non-finite parameter(s): ",
               paste(rw.names[problem.indices],collapse=', '),
               call.=FALSE
               )
        }
      }
    }
    
    ## determine the weights
    weights <- try(
                   dmeasure(
                            object,
                            y=object@data[,nt,drop=FALSE],
                            x=X,
                            times=times[nt+1],
                            params=if (transform) tparams else params,
                            log=FALSE,
                            .getnativesymbolinfo=gnsi.dmeas
                            ),
                   silent=FALSE
                   )
    if (inherits(weights,'try-error'))
      stop(sQuote("pfilter")," error: error in calculation of weights",call.=FALSE)
    if (!all(is.finite(weights)))
      stop(sQuote("pfilter")," error: ",sQuote("dmeasure")," returns non-finite value",call.=FALSE)
    gnsi.dmeas <- FALSE
    
    ## compute prediction mean, prediction variance, filtering mean,
    ## effective sample size, log-likelihood
    ## also do resampling if filtering has not failed
    xx <- try(
              .Call(
                    pfilter_computations,
                    X,params,Np[nt+1],
                    random.walk,
                    sigma1,
                    pred.mean,pred.var,
                    filter.mean,one.par,
                    weights,tol
                    ),
              silent=FALSE
              )
    if (inherits(xx,'try-error')) {
      stop(sQuote("pfilter")," error",call.=FALSE)
    }
    all.fail <- xx$fail
    loglik[nt] <- xx$loglik
    eff.sample.size[nt] <- xx$ess
    
    x <- xx$states
    params <- xx$params
    
    if (pred.mean)
      pred.m[,nt] <- xx$pm
    if (pred.var)
      pred.v[,nt] <- xx$pv
    if (filter.mean)
      filt.m[,nt] <- xx$fm
    
    if (all.fail) { ## all particles are lost
      nfail <- nfail+1
      if (verbose)
        message("filtering failure at time t = ",times[nt+1])
      if (nfail>max.fail)
        stop(sQuote("pfilter")," error: too many filtering failures",call.=FALSE)
    }
    
    if (save.states) {
      xparticles[[nt]] <- x
      dimnames(xparticles[[nt]]) <- setNames(dimnames(xparticles[[nt]]),c("variable","rep"))
    }
    
    if (save.params) {
      pparticles[[nt]] <- params
      dimnames(pparticles[[nt]]) <- setNames(dimnames(pparticles[[nt]]),c("variable","rep"))
    }
    
    if (verbose && (nt%%5==0))
      cat("pfilter timestep",nt,"of",ntimes,"finished\n")
    
  }
  
  if (!is.null(seed)) {
    assign(".Random.seed",save.seed,pos=.GlobalEnv)
    seed <- save.seed
  }

  if (nfail>0)
    warning(sprintf(ngettext(nfail,msg1="%d filtering failure occurred in ",
                             msg2="%d filtering failures occurred in "),nfail),
            sQuote("pfilter"),call.=FALSE)

  pompUnload(object)

  new(
      "pfilterd.pomp",
      object,
      pred.mean=pred.m,
      pred.var=pred.v,
      filter.mean=filt.m,
      paramMatrix=if (mif2) params else array(data=numeric(0),dim=c(0,0)),
      eff.sample.size=eff.sample.size,
      cond.loglik=loglik,
      saved.states=xparticles,
      saved.params=pparticles,
      seed=as.integer(seed),
      Np=as.integer(Np),
      tol=tol,
      nfail=as.integer(nfail),
      loglik=sum(loglik)
      )
}

setMethod(
          "pfilter",
          signature=signature(object="pomp"),
          function (object, params, Np,
                    tol = 1e-17,
                    max.fail = Inf,
                    pred.mean = FALSE,
                    pred.var = FALSE,
                    filter.mean = FALSE,
                    save.states = FALSE,
                    save.params = FALSE,
                    seed = NULL,
                    verbose = getOption("verbose"),
                    ...) {
            if (missing(params)) params <- coef(object)
            pfilter.internal(
                             object=object,
                             params=params,
                             Np=Np,
                             tol=tol,
                             max.fail=max.fail,
                             pred.mean=pred.mean,
                             pred.var=pred.var,
                             filter.mean=filter.mean,
                             save.states=save.states,
                             save.params=save.params,
                             seed=seed,
                             verbose=verbose,
                             .transform=FALSE,
                             ...
                             )
          }
          )

setMethod(
          "pfilter",
          signature=signature(object="pfilterd.pomp"),
          function (object, params, Np, tol, ...) {
            if (missing(params)) params <- coef(object)
            if (missing(Np)) Np <- object@Np
            if (missing(tol)) tol <- object@tol
            pfilter(
                    object=as(object,"pomp"),
                    params=params,
                    Np=Np,
                    tol=tol,
                    ...
                    )
          }
          )