epclust/R/clustering.R

   1 #' @name clustering
   2 #' @rdname clustering
   3 #' @aliases clusteringTask1 computeClusters1 computeClusters2
   4 #'
   5 #' @title Two-stage clustering, withing one task (see \code{claws()})
   6 #'
   7 #' @description \code{clusteringTask1()} runs one full stage-1 task, which consists in
   8 #'   iterated stage 1 clustering (on nb_curves / ntasks energy contributions, computed
   9 #'   through discrete wavelets coefficients). \code{computeClusters1()} and
  10 #'   \code{computeClusters2()} correspond to the atomic clustering procedures respectively
  11 #'   for stage 1 and 2. The former applies the clustering algorithm (PAM) on a
  12 #'   contributions matrix, while the latter clusters a chunk of series inside one task
  13 #'   (~max nb_series_per_chunk)
  14 #'
  15 #' @param indices Range of series indices to cluster in parallel (initial data)
  16 #' @param getContribs Function to retrieve contributions from initial series indices:
  17 #'   \code{getContribs(indices)} outpus a contributions matrix
  18 #' @param contribs matrix of contributions (e.g. output of \code{curvesToContribs()})
  19 #' @inheritParams computeSynchrones
  20 #' @inheritParams claws
  21 #'
  22 #' @return For \code{clusteringTask1()} and \code{computeClusters1()}, the indices of the
  23 #'   computed (K1) medoids. Indices are irrelevant for stage 2 clustering, thus
  24 #'   \code{computeClusters2()} outputs a matrix of medoids
  25 #'   (of size limited by nb_series_per_chunk)
  26 NULL
  27
  28 #' @rdname clustering
  29 #' @export
  30 clusteringTask1 = function(
  31     indices, getContribs, K1, nb_series_per_chunk, ncores_clust=1, verbose=FALSE, parll=TRUE)
  32 {
  33     if (verbose)
  34         cat(paste("*** Clustering task on ",length(indices)," lines\n", sep=""))
  35
  36     wrapComputeClusters1 = function(inds) {
  37         if (parll)
  38             require("epclust", quietly=TRUE)
  39         if (verbose)
  40             cat(paste("   computeClusters1() on ",length(inds)," lines\n", sep=""))
  41         inds[ computeClusters1(getContribs(inds), K1) ]
  42     }
  43
  44     if (parll)
  45     {
  46         cl = parallel::makeCluster(ncores_clust)
  47         parallel::clusterExport(cl, varlist=c("getContribs","K1","verbose"), envir=environment())
  48     }
  49     while (length(indices) > K1)
  50     {
  51         indices_workers = .spreadIndices(indices, nb_series_per_chunk)
  52         if (parll)
  53             indices = unlist( parallel::parLapply(cl, indices_workers, wrapComputeClusters1) )
  54         else
  55             indices = unlist( lapply(indices_workers, wrapComputeClusters1) )
  56     }
  57     if (parll)
  58         parallel::stopCluster(cl)
  59
  60     indices #medoids
  61 }
  62
  63 #' @rdname clustering
  64 #' @export
  65 computeClusters1 = function(contribs, K1)
  66     cluster::pam(contribs, K1, diss=FALSE)$id.med
  67
  68 #' @rdname clustering
  69 #' @export
  70 computeClusters2 = function(medoids, K2,
  71     getRefSeries, nb_ref_curves, nb_series_per_chunk, ncores_clust=1,verbose=FALSE,parll=TRUE)
  72 {
  73     synchrones = computeSynchrones(medoids,
  74         getRefSeries, nb_ref_curves, nb_series_per_chunk, ncores_clust, verbose, parll)
  75     distances = computeWerDists(synchrones, ncores_clust, verbose, parll)
  76     medoids[ cluster::pam(distances, K2, diss=TRUE)$medoids , ]
  77 }
  78
  79 #' computeSynchrones
  80 #'
  81 #' Compute the synchrones curves (sum of clusters elements) from a matrix of medoids,
  82 #' using L2 distances.
  83 #'
  84 #' @param medoids Matrix of medoids (curves of same legnth as initial series)
  85 #' @param getRefSeries Function to retrieve initial series (e.g. in stage 2 after series
  86 #'   have been replaced by stage-1 medoids)
  87 #' @param nb_ref_curves How many reference series? (This number is known at this stage)
  88 #' @inheritParams claws
  89 #'
  90 #' @export
  91 computeSynchrones = function(medoids, getRefSeries,
  92     nb_ref_curves, nb_series_per_chunk, ncores_clust=1,verbose=FALSE,parll=TRUE)
  93 {
  94     computeSynchronesChunk = function(indices)
  95     {
  96         ref_series = getRefSeries(indices)
  97         #get medoids indices for this chunk of series
  98         for (i in seq_len(nrow(ref_series)))
  99         {
 100             j = which.min( rowSums( sweep(medoids, 2, ref_series[i,], '-')^2 ) )
 101             if (parll)
 102                 synchronicity::lock(m)
 103             synchrones[j,] = synchrones[j,] + ref_series[i,]
 104             counts[j,1] = counts[j,1] + 1
 105             if (parll)
 106                 synchronicity::unlock(m)
 107         }
 108     }
 109
 110     K = nrow(medoids)
 111     # Use bigmemory (shared==TRUE by default) + synchronicity to fill synchrones in //
 112     synchrones = bigmemory::big.matrix(nrow=K,ncol=ncol(medoids),type="double",init=0.)
 113     counts = bigmemory::big.matrix(nrow=K,ncol=1,type="double",init=0)
 114     # Fork (// run) only on Linux & MacOS; on Windows: run sequentially
 115     parll = (requireNamespace("synchronicity",quietly=TRUE)
 116         && parll && Sys.info()['sysname'] != "Windows")
 117     if (parll)
 118         m <- synchronicity::boost.mutex()
 119
 120     indices_workers = .spreadIndices(seq_len(nb_ref_curves), nb_series_per_chunk)
 121     for (inds in indices_workers)
 122     {
 123         if (verbose)
 124             cat(paste("--- Compute synchrones for ",length(inds)," lines\n", sep=""))
 125         if (parll)
 126             ignored <- parallel::mcparallel(computeSynchronesChunk(inds))
 127         else
 128             computeSynchronesChunk(inds)
 129     }
 130     if (parll)
 131         parallel::mccollect()
 132
 133     mat_syncs = matrix(nrow=K, ncol=ncol(medoids))
 134     vec_count = rep(NA, K)
 135     #TODO: can we avoid this loop?
 136     for (i in seq_len(K))
 137     {
 138         mat_syncs[i,] = synchrones[i,]
 139         vec_count[i] = counts[i,1]
 140     }
 141     #NOTE: odds for some clusters to be empty? (when series already come from stage 2)
 142     #      ...maybe; but let's hope resulting K1' be still quite bigger than K2
 143     mat_syncs = sweep(mat_syncs, 1, vec_count, '/')
 144     mat_syncs[ sapply(seq_len(K), function(i) all(!is.nan(mat_syncs[i,]))) , ]
 145 }
 146
 147 #' computeWerDists
 148 #'
 149 #' Compute the WER distances between the synchrones curves (in rows), which are
 150 #' returned (e.g.) by \code{computeSynchrones()}
 151 #'
 152 #' @param synchrones A matrix of synchrones, in rows. The series have same length as the
 153 #'   series in the initial dataset
 154 #' @inheritParams claws
 155 #'
 156 #' @export
 157 computeWerDists = function(synchrones, ncores_clust=1,verbose=FALSE,parll=TRUE)
 158 {
 159     n <- nrow(synchrones)
 160     delta <- ncol(synchrones)
 161     #TODO: automatic tune of all these parameters ? (for other users)
 162     nvoice   <- 4
 163     # noctave = 2^13 = 8192 half hours ~ 180 days ; ~log2(ncol(synchrones))
 164     noctave = 13
 165     # 4 here represent 2^5 = 32 half-hours ~ 1 day
 166     #NOTE: default scalevector == 2^(0:(noctave * nvoice) / nvoice) * s0 (?)
 167     scalevector  <- 2^(4:(noctave * nvoice) / nvoice) * 2
 168     #condition: ( log2(s0*w0/(2*pi)) - 1 ) * nvoice + 1.5 >= 1
 169     s0=2
 170     w0=2*pi
 171     scaled=FALSE
 172     s0log = as.integer( (log2( s0*w0/(2*pi) ) - 1) * nvoice + 1.5 )
 173     totnoct = noctave + as.integer(s0log/nvoice) + 1
 174
 175     computeCWT = function(i)
 176     {
 177         if (verbose)
 178             cat(paste("+++ Compute Rwave::cwt() on serie ",i,"\n", sep=""))
 179         ts <- scale(ts(synchrones[i,]), center=TRUE, scale=scaled)
 180         totts.cwt = Rwave::cwt(ts,totnoct,nvoice,w0,plot=0)
 181         ts.cwt = totts.cwt[,s0log:(s0log+noctave*nvoice)]
 182         #Normalization
 183         sqs <- sqrt(2^(0:(noctave*nvoice)/nvoice)*s0)
 184         sqres <- sweep(ts.cwt,2,sqs,'*')
 185         sqres / max(Mod(sqres))
 186     }
 187
 188     if (parll)
 189     {
 190         cl = parallel::makeCluster(ncores_clust)
 191         parallel::clusterExport(cl,
 192             varlist=c("synchrones","totnoct","nvoice","w0","s0log","noctave","s0","verbose"),
 193             envir=environment())
 194     }
 195
 196     # (normalized) observations node with CWT
 197     Xcwt4 <-
 198         if (parll)
 199             parallel::parLapply(cl, seq_len(n), computeCWT)
 200         else
 201             lapply(seq_len(n), computeCWT)
 202
 203     if (parll)
 204         parallel::stopCluster(cl)
 205
 206     Xwer_dist <- bigmemory::big.matrix(nrow=n, ncol=n, type="double")
 207     fcoefs = rep(1/3, 3) #moving average on 3 values (TODO: very slow! correct?!)
 208     if (verbose)
 209         cat("*** Compute WER distances from CWT\n")
 210
 211     computeDistancesLineI = function(i)
 212     {
 213         if (verbose)
 214             cat(paste("   Line ",i,"\n", sep=""))
 215         for (j in (i+1):n)
 216         {
 217             #TODO: 'circular=TRUE' is wrong, should just take values on the sides; to rewrite in C
 218             num <- filter(Mod(Xcwt4[[i]] * Conj(Xcwt4[[j]])), fcoefs, circular=TRUE)
 219             WX <- filter(Mod(Xcwt4[[i]] * Conj(Xcwt4[[i]])), fcoefs, circular=TRUE)
 220             WY <- filter(Mod(Xcwt4[[j]] * Conj(Xcwt4[[j]])), fcoefs, circular=TRUE)
 221             wer2    <- sum(colSums(num)^2) / sum( sum(colSums(WX) * colSums(WY)) )
 222             if (parll)
 223                 synchronicity::lock(m)
 224             Xwer_dist[i,j] <- sqrt(delta * ncol(Xcwt4[[1]]) * (1 - wer2))
 225             Xwer_dist[j,i] <- Xwer_dist[i,j]
 226             if (parll)
 227                 synchronicity::unlock(m)
 228         }
 229         Xwer_dist[i,i] = 0.
 230     }
 231
 232     parll = (requireNamespace("synchronicity",quietly=TRUE)
 233         && parll && Sys.info()['sysname'] != "Windows")
 234     if (parll)
 235         m <- synchronicity::boost.mutex()
 236
 237     for (i in 1:(n-1))
 238     {
 239         if (parll)
 240             ignored <- parallel::mcparallel(computeDistancesLineI(i))
 241         else
 242             computeDistancesLineI(i)
 243     }
 244     Xwer_dist[n,n] = 0.
 245
 246     if (parll)
 247         parallel::mccollect()
 248
 249     mat_dists = matrix(nrow=n, ncol=n)
 250     #TODO: avoid this loop?
 251     for (i in 1:n)
 252         mat_dists[i,] = Xwer_dist[i,]
 253     mat_dists
 254 }
 255
 256 # Helper function to divide indices into balanced sets
 257 .spreadIndices = function(indices, nb_per_chunk)
 258 {
 259     L = length(indices)
 260     nb_workers = floor( L / nb_per_chunk )
 261     if (nb_workers == 0)
 262     {
 263         # L < nb_series_per_chunk, simple case
 264         indices_workers = list(indices)
 265     }
 266     else
 267     {
 268         indices_workers = lapply( seq_len(nb_workers), function(i)
 269             indices[(nb_per_chunk*(i-1)+1):(nb_per_chunk*i)] )
 270         # Spread the remaining load among the workers
 271         rem = L %% nb_per_chunk
 272         while (rem > 0)
 273         {
 274             index = rem%%nb_workers + 1
 275             indices_workers[[index]] = c(indices_workers[[index]], indices[L-rem+1])
 276             rem = rem - 1
 277         }
 278     }
 279     indices_workers
 280 }