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).
  10 #'   \code{clusteringTask2()} runs a full stage-2 task, which consists in synchrones
  11 #'   and then WER distances computations, before applying the clustering algorithm.
  12 #'   \code{computeClusters1()} and \code{computeClusters2()} correspond to the atomic
  13 #'   clustering procedures respectively for stage 1 and 2. The former applies the
  14 #'   clustering algorithm (PAM) on a contributions matrix, while the latter clusters
  15 #'   a chunk of series inside one task (~max nb_series_per_chunk)
  16 #'
  17 #' @param indices Range of series indices to cluster in parallel (initial data)
  18 #' @param getContribs Function to retrieve contributions from initial series indices:
  19 #'   \code{getContribs(indices)} outpus a contributions matrix
  20 #' @param contribs matrix of contributions (e.g. output of \code{curvesToContribs()})
  21 #' @param distances matrix of K1 x K1 (WER) distances between synchrones
  22 #' @inheritParams computeSynchrones
  23 #' @inheritParams claws
  24 #'
  25 #' @return For \code{clusteringTask1()} and \code{computeClusters1()}, the indices of the
  26 #'   computed (K1) medoids. Indices are irrelevant for stage 2 clustering, thus
  27 #'   \code{computeClusters2()} outputs a big.matrix of medoids
  28 #'   (of size limited by nb_series_per_chunk)
  29 NULL
  30
  31 #' @rdname clustering
  32 #' @export
  33 clusteringTask1 = function(
  34     indices, getContribs, K1, nb_series_per_chunk, ncores_clust=1, verbose=FALSE, parll=TRUE)
  35 {
  36     if (verbose)
  37         cat(paste("*** Clustering task 1 on ",length(indices)," lines\n", sep=""))
  38
  39     if (parll)
  40     {
  41         cl = parallel::makeCluster(ncores_clust)
  42         parallel::clusterExport(cl, varlist=c("getContribs","K1","verbose"), envir=environment())
  43     }
  44     while (length(indices) > K1)
  45     {
  46         indices_workers = .spreadIndices(indices, nb_series_per_chunk)
  47         indices <-
  48             if (parll)
  49             {
  50                 unlist( parallel::parLapply(cl, indices_workers, function(inds) {
  51                     require("epclust", quietly=TRUE)
  52                     inds[ computeClusters1(getContribs(inds), K1, verbose) ]
  53                 }) )
  54             }
  55             else
  56             {
  57                 unlist( lapply(indices_workers, function(inds)
  58                     inds[ computeClusters1(getContribs(inds), K1, verbose) ]
  59                 ) )
  60             }
  61     }
  62     if (parll)
  63         parallel::stopCluster(cl)
  64
  65     indices #medoids
  66 }
  67
  68 #' @rdname clustering
  69 #' @export
  70 clusteringTask2 = function(medoids, K2,
  71     getRefSeries, nb_ref_curves, nb_series_per_chunk, ncores_clust=1,verbose=FALSE,parll=TRUE)
  72 {
  73     if (verbose)
  74         cat(paste("*** Clustering task 2 on ",nrow(medoids)," lines\n", sep=""))
  75
  76     if (nrow(medoids) <= K2)
  77         return (medoids)
  78     synchrones = computeSynchrones(medoids,
  79         getRefSeries, nb_ref_curves, nb_series_per_chunk, ncores_clust, verbose, parll)
  80     distances = computeWerDists(synchrones, ncores_clust, verbose, parll)
  81     medoids[ computeClusters2(distances,K2,verbose), ]
  82 }
  83
  84 #' @rdname clustering
  85 #' @export
  86 computeClusters1 = function(contribs, K1, verbose=FALSE)
  87 {
  88     if (verbose)
  89         cat(paste("   computeClusters1() on ",nrow(contribs)," lines\n", sep=""))
  90     cluster::pam(contribs, K1, diss=FALSE)$id.med
  91 }
  92
  93 #' @rdname clustering
  94 #' @export
  95 computeClusters2 = function(distances, K2, verbose=FALSE)
  96 {
  97     if (verbose)
  98         cat(paste("   computeClusters2() on ",nrow(distances)," lines\n", sep=""))
  99     cluster::pam(distances, K2, diss=TRUE)$id.med
 100 }
 101
 102 #' computeSynchrones
 103 #'
 104 #' Compute the synchrones curves (sum of clusters elements) from a matrix of medoids,
 105 #' using L2 distances.
 106 #'
 107 #' @param medoids big.matrix of medoids (curves of same length as initial series)
 108 #' @param getRefSeries Function to retrieve initial series (e.g. in stage 2 after series
 109 #'   have been replaced by stage-1 medoids)
 110 #' @param nb_ref_curves How many reference series? (This number is known at this stage)
 111 #' @inheritParams claws
 112 #'
 113 #' @return A big.matrix of size K1 x L where L = data_length
 114 #'
 115 #' @export
 116 computeSynchrones = function(medoids, getRefSeries,
 117     nb_ref_curves, nb_series_per_chunk, ncores_clust=1,verbose=FALSE,parll=TRUE)
 118 {
 119     if (verbose)
 120         cat(paste("--- Compute synchrones\n", sep=""))
 121
 122     computeSynchronesChunk = function(indices)
 123     {
 124         ref_series = getRefSeries(indices)
 125         nb_series = nrow(ref_series)
 126
 127         if (parll)
 128         {
 129             require("bigmemory", quietly=TRUE)
 130             require("synchronicity", quietly=TRUE)
 131             require("epclust", quietly=TRUE)
 132             synchrones <- bigmemory::attach.big.matrix(synchrones_desc)
 133             medoids <- bigmemory::attach.big.matrix(medoids_desc)
 134             m <- synchronicity::attach.mutex(m_desc)
 135         }
 136
 137         #get medoids indices for this chunk of series
 138         mi = computeMedoidsIndices(medoids@address, ref_series)
 139 #       #R-equivalent, requiring a matrix (thus potentially breaking "fit-in-memory" hope)
 140 #       mat_meds = medoids[,]
 141 #       mi = rep(NA,nb_series)
 142 #       for (i in 1:nb_series)
 143 #           mi[i] <- which.min( rowSums( sweep(mat_meds, 2, ref_series[i,], '-')^2 ) )
 144 #       rm(mat_meds); gc()
 145
 146         for (i in seq_len(nb_series))
 147         {
 148             if (parll)
 149                 synchronicity::lock(m)
 150             synchrones[mi[i],] = synchrones[mi[i],] + ref_series[i,]
 151 #TODO: remove counts
 152             counts[mi[i],1] = counts[mi[i],1] + 1
 153             if (parll)
 154                 synchronicity::unlock(m)
 155         }
 156     }
 157
 158     K = nrow(medoids) ; L = ncol(medoids)
 159     # Use bigmemory (shared==TRUE by default) + synchronicity to fill synchrones in //
 160     # TODO: if size > RAM (not our case), use file-backed big.matrix
 161     synchrones = bigmemory::big.matrix(nrow=K, ncol=L, type="double", init=0.)
 162     counts = bigmemory::big.matrix(nrow=K, ncol=1, type="double", init=0)
 163     # synchronicity is only for Linux & MacOS; on Windows: run sequentially
 164     parll = (requireNamespace("synchronicity",quietly=TRUE)
 165         && parll && Sys.info()['sysname'] != "Windows")
 166     if (parll)
 167     {
 168         m <- synchronicity::boost.mutex()
 169         m_desc <- synchronicity::describe(m)
 170         synchrones_desc = bigmemory::describe(synchrones)
 171         medoids_desc = bigmemory::describe(medoids)
 172
 173         cl = parallel::makeCluster(ncores_clust)
 174         parallel::clusterExport(cl,
 175             varlist=c("synchrones_desc","counts","verbose","m_desc","medoids_desc","getRefSeries"),
 176             envir=environment())
 177     }
 178
 179     indices_workers = .spreadIndices(seq_len(nb_ref_curves), nb_series_per_chunk)
 180     ignored <-
 181         if (parll)
 182             parallel::parLapply(cl, indices_workers, computeSynchronesChunk)
 183         else
 184             lapply(indices_workers, computeSynchronesChunk)
 185
 186     if (parll)
 187         parallel::stopCluster(cl)
 188
 189     #TODO: can we avoid this loop? ( synchrones = sweep(synchrones, 1, counts, '/') )
 190     for (i in seq_len(K))
 191         synchrones[i,] = synchrones[i,] / counts[i,1]
 192     #NOTE: odds for some clusters to be empty? (when series already come from stage 2)
 193     #      ...maybe; but let's hope resulting K1' be still quite bigger than K2
 194     noNA_rows = sapply(seq_len(K), function(i) all(!is.nan(synchrones[i,])))
 195     if (all(noNA_rows))
 196         return (synchrones)
 197     # Else: some clusters are empty, need to slice synchrones
 198     synchrones[noNA_rows,]
 199 }
 200
 201 #' computeWerDists
 202 #'
 203 #' Compute the WER distances between the synchrones curves (in rows), which are
 204 #' returned (e.g.) by \code{computeSynchrones()}
 205 #'
 206 #' @param synchrones A big.matrix of synchrones, in rows. The series have same length
 207 #'   as the series in the initial dataset
 208 #' @inheritParams claws
 209 #'
 210 #' @return A matrix of size K1 x K1
 211 #'
 212 #' @export
 213 computeWerDists = function(synchrones, ncores_clust=1,verbose=FALSE,parll=TRUE)
 214 {
 215     if (verbose)
 216         cat(paste("--- Compute WER dists\n", sep=""))
 217
 218     n <- nrow(synchrones)
 219     delta <- ncol(synchrones)
 220     #TODO: automatic tune of all these parameters ? (for other users)
 221     nvoice   <- 4
 222     # noctave = 2^13 = 8192 half hours ~ 180 days ; ~log2(ncol(synchrones))
 223     noctave = 13
 224     # 4 here represent 2^5 = 32 half-hours ~ 1 day
 225     #NOTE: default scalevector == 2^(0:(noctave * nvoice) / nvoice) * s0 (?)
 226     scalevector  <- 2^(4:(noctave * nvoice) / nvoice + 1)
 227     #condition: ( log2(s0*w0/(2*pi)) - 1 ) * nvoice + 1.5 >= 1
 228     s0=2
 229     w0=2*pi
 230     scaled=FALSE
 231     s0log = as.integer( (log2( s0*w0/(2*pi) ) - 1) * nvoice + 1.5 )
 232     totnoct = noctave + as.integer(s0log/nvoice) + 1
 233
 234     Xwer_dist <- bigmemory::big.matrix(nrow=n, ncol=n, type="double")
 235     fcoefs = rep(1/3, 3) #moving average on 3 values
 236
 237     # Generate n(n-1)/2 pairs for WER distances computations
 238     pairs = list()
 239     V = seq_len(n)
 240     for (i in 1:n)
 241     {
 242         V = V[-1]
 243         pairs = c(pairs, lapply(V, function(v) c(i,v)))
 244     }
 245
 246     # Distance between rows i and j
 247     computeDistancesIJ = function(pair)
 248     {
 249         if (parll)
 250         {
 251             require("bigmemory", quietly=TRUE)
 252             require("epclust", quietly=TRUE)
 253             synchrones <- bigmemory::attach.big.matrix(synchrones_desc)
 254             Xwer_dist <- bigmemory::attach.big.matrix(Xwer_dist_desc)
 255         }
 256
 257         computeCWT = function(index)
 258         {
 259             ts <- scale(ts(synchrones[index,]), center=TRUE, scale=scaled)
 260             totts.cwt = Rwave::cwt(ts, totnoct, nvoice, w0, plot=FALSE)
 261             ts.cwt = totts.cwt[,s0log:(s0log+noctave*nvoice)]
 262             #Normalization
 263             sqs <- sqrt(2^(0:(noctave*nvoice)/nvoice)*s0)
 264             sqres <- sweep(ts.cwt,2,sqs,'*')
 265             sqres / max(Mod(sqres))
 266         }
 267 #browser()
 268         i = pair[1] ; j = pair[2]
 269         if (verbose && j==i+1)
 270             cat(paste("   Distances (",i,",",j,"), (",i,",",j+1,") ...\n", sep=""))
 271 print(system.time( {        cwt_i <- computeCWT(i)
 272         cwt_j <- computeCWT(j) } ))
 273
 274 print(system.time( {
 275         num <- epclustFilter(Mod(cwt_i * Conj(cwt_j)))
 276         WX  <- epclustFilter(Mod(cwt_i * Conj(cwt_i)))
 277         WY  <- epclustFilter(Mod(cwt_j * Conj(cwt_j)))
 278         wer2 <- sum(colSums(num)^2) / sum(colSums(WX) * colSums(WY))
 279         Xwer_dist[i,j] <- sqrt(delta * ncol(cwt_i) * max(1 - wer2, 0.)) #FIXME: wer2 should be < 1
 280         Xwer_dist[j,i] <- Xwer_dist[i,j]
 281 } ) )
 282         Xwer_dist[i,i] = 0.
 283     }
 284
 285     if (parll)
 286     {
 287         cl = parallel::makeCluster(ncores_clust)
 288         synchrones_desc <- bigmemory::describe(synchrones)
 289         Xwer_dist_desc <- bigmemory::describe(Xwer_dist)
 290
 291         parallel::clusterExport(cl, varlist=c("synchrones_desc","Xwer_dist_desc","totnoct",
 292             "nvoice","w0","s0log","noctave","s0","verbose"), envir=environment())
 293     }
 294 browser()
 295     ignored <-
 296         if (parll)
 297             parallel::parLapply(cl, pairs, computeDistancesIJ)
 298         else
 299             lapply(pairs, computeDistancesIJ)
 300
 301     if (parll)
 302         parallel::stopCluster(cl)
 303 #browser()
 304     Xwer_dist[n,n] = 0.
 305     distances <- Xwer_dist[,]
 306     rm(Xwer_dist) ; gc()
 307     distances #~small matrix K1 x K1
 308 }
 309
 310 # Helper function to divide indices into balanced sets
 311 .spreadIndices = function(indices, nb_per_chunk)
 312 {
 313     L = length(indices)
 314     nb_workers = floor( L / nb_per_chunk )
 315     if (nb_workers == 0)
 316     {
 317         # L < nb_series_per_chunk, simple case
 318         indices_workers = list(indices)
 319     }
 320     else
 321     {
 322         indices_workers = lapply( seq_len(nb_workers), function(i)
 323             indices[(nb_per_chunk*(i-1)+1):(nb_per_chunk*i)] )
 324         # Spread the remaining load among the workers
 325         rem = L %% nb_per_chunk
 326         while (rem > 0)
 327         {
 328             index = rem%%nb_workers + 1
 329             indices_workers[[index]] = c(indices_workers[[index]], indices[L-rem+1])
 330             rem = rem - 1
 331         }
 332     }
 333     indices_workers
 334 }