save state; test clustering not OK, all others OK
[epclust.git] / epclust / R / clustering.R
CommitLineData
4bcfdbee
BA
1#' @name clustering
2#' @rdname clustering
eef6f6c9 3#' @aliases clusteringTask1 clusteringTask2 computeClusters1 computeClusters2
4bcfdbee 4#'
492cd9e7 5#' @title Two-stage clustering, withing one task (see \code{claws()})
4bcfdbee 6#'
492cd9e7
BA
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
bf5c0844
BA
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
2b9f5356 14#' first clustering algorithm on a contributions matrix, while the latter clusters
0486fbad 15#' a set of series inside one task (~nb_items_clust1)
4bcfdbee
BA
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
4bcfdbee
BA
20#' @inheritParams computeSynchrones
21#' @inheritParams claws
22#'
0486fbad
BA
23#' @return For \code{clusteringTask1()}, the indices of the computed (K1) medoids.
24#' Indices are irrelevant for stage 2 clustering, thus \code{clusteringTask2()}
25#' outputs a big.matrix of medoids (of size LxK2, K2 = final number of clusters)
4bcfdbee
BA
26NULL
27
28#' @rdname clustering
29#' @export
0486fbad 30clusteringTask1 = function(indices, getContribs, K1, algoClust1, nb_items_clust1,
37c82bba 31 ncores_clust=1, verbose=FALSE, parll=TRUE)
5c652979 32{
492cd9e7 33 if (parll)
7b13d0c2 34 {
37c82bba 35 cl = parallel::makeCluster(ncores_clust, outfile = "")
492cd9e7 36 parallel::clusterExport(cl, varlist=c("getContribs","K1","verbose"), envir=environment())
7b13d0c2 37 }
492cd9e7
BA
38 while (length(indices) > K1)
39 {
0486fbad
BA
40 indices_workers = .spreadIndices(indices, nb_items_clust1)
41 if (verbose)
42 cat(paste("*** [iterated] Clustering task 1 on ",length(indices)," series\n", sep=""))
e161499b
BA
43 indices <-
44 if (parll)
45 {
46 unlist( parallel::parLapply(cl, indices_workers, function(inds) {
47 require("epclust", quietly=TRUE)
0486fbad 48 inds[ algoClust1(getContribs(inds), K1) ]
e161499b
BA
49 }) )
50 }
51 else
52 {
53 unlist( lapply(indices_workers, function(inds)
0486fbad 54 inds[ algoClust1(getContribs(inds), K1) ]
e161499b
BA
55 ) )
56 }
492cd9e7
BA
57 }
58 if (parll)
59 parallel::stopCluster(cl)
60
56857861 61 indices #medoids
5c652979
BA
62}
63
4bcfdbee
BA
64#' @rdname clustering
65#' @export
0486fbad
BA
66clusteringTask2 = function(medoids, K2, algoClust2, getRefSeries, nb_ref_curves,
67 nb_series_per_chunk, sync_mean, nbytes,endian,ncores_clust=1,verbose=FALSE,parll=TRUE)
5c652979 68{
e161499b 69 if (verbose)
0486fbad 70 cat(paste("*** Clustering task 2 on ",ncol(medoids)," synchrones\n", sep=""))
e161499b 71
0486fbad 72 if (ncol(medoids) <= K2)
bf5c0844 73 return (medoids)
0486fbad
BA
74 synchrones = computeSynchrones(medoids, getRefSeries, nb_ref_curves,
75 nb_series_per_chunk, sync_mean, ncores_clust, verbose, parll)
a174b8ea 76 distances = computeWerDists(synchrones, nbytes, endian, ncores_clust, verbose, parll)
e161499b 77 if (verbose)
0486fbad
BA
78 cat(paste(" algoClust2() on ",nrow(distances)," items\n", sep=""))
79 medoids[ algoClust2(distances,K2), ]
e161499b 80}
bf5c0844 81
4bcfdbee
BA
82#' computeSynchrones
83#'
84#' Compute the synchrones curves (sum of clusters elements) from a matrix of medoids,
85#' using L2 distances.
86#'
24ed5d83 87#' @param medoids big.matrix of medoids (curves of same length as initial series)
4bcfdbee
BA
88#' @param getRefSeries Function to retrieve initial series (e.g. in stage 2 after series
89#' have been replaced by stage-1 medoids)
492cd9e7 90#' @param nb_ref_curves How many reference series? (This number is known at this stage)
4bcfdbee
BA
91#' @inheritParams claws
92#'
eef6f6c9 93#' @return A big.matrix of size L x K1 where L = length of a serie
24ed5d83 94#'
4bcfdbee 95#' @export
0486fbad
BA
96computeSynchrones = function(medoids, getRefSeries, nb_ref_curves,
97 nb_series_per_chunk, sync_mean, ncores_clust=1,verbose=FALSE,parll=TRUE)
e205f218 98{
492cd9e7 99 computeSynchronesChunk = function(indices)
3eef8d3d 100 {
363ae134
BA
101 if (parll)
102 {
103 require("bigmemory", quietly=TRUE)
6ad3f3fd 104 requireNamespace("synchronicity", quietly=TRUE)
363ae134
BA
105 require("epclust", quietly=TRUE)
106 synchrones <- bigmemory::attach.big.matrix(synchrones_desc)
0486fbad
BA
107 if (sync_mean)
108 counts <- bigmemory::attach.big.matrix(counts_desc)
363ae134
BA
109 medoids <- bigmemory::attach.big.matrix(medoids_desc)
110 m <- synchronicity::attach.mutex(m_desc)
111 }
112
6ad3f3fd
BA
113 ref_series = getRefSeries(indices)
114 nb_series = nrow(ref_series)
115
0486fbad 116 # Get medoids indices for this chunk of series
2c14dbea 117 mi = computeMedoidsIndices(medoids@address, ref_series)
e161499b
BA
118
119 for (i in seq_len(nb_series))
56857861 120 {
492cd9e7
BA
121 if (parll)
122 synchronicity::lock(m)
eef6f6c9 123 synchrones[, mi[i] ] = synchrones[, mi[i] ] + ref_series[,i]
0486fbad
BA
124 if (sync_mean)
125 counts[ mi[i] ] = counts[ mi[i] ] + 1
492cd9e7
BA
126 if (parll)
127 synchronicity::unlock(m)
128 }
129 }
130
0486fbad 131 K = ncol(medoids) ; L = nrow(medoids)
492cd9e7 132 # Use bigmemory (shared==TRUE by default) + synchronicity to fill synchrones in //
24ed5d83 133 # TODO: if size > RAM (not our case), use file-backed big.matrix
eef6f6c9 134 synchrones = bigmemory::big.matrix(nrow=L, ncol=K, type="double", init=0.)
0486fbad
BA
135 if (sync_mean)
136 counts = bigmemory::big.matrix(nrow=K, ncol=1, type="double", init=0)
24ed5d83 137 # synchronicity is only for Linux & MacOS; on Windows: run sequentially
492cd9e7
BA
138 parll = (requireNamespace("synchronicity",quietly=TRUE)
139 && parll && Sys.info()['sysname'] != "Windows")
140 if (parll)
363ae134 141 {
492cd9e7 142 m <- synchronicity::boost.mutex()
363ae134
BA
143 m_desc <- synchronicity::describe(m)
144 synchrones_desc = bigmemory::describe(synchrones)
0486fbad
BA
145 if (sync_mean)
146 counts_desc = bigmemory::describe(counts)
363ae134 147 medoids_desc = bigmemory::describe(medoids)
24ed5d83 148 cl = parallel::makeCluster(ncores_clust)
0486fbad
BA
149 varlist=c("synchrones_desc","sync_mean","m_desc","medoids_desc","getRefSeries")
150 if (sync_mean)
151 varlist = c(varlist, "counts_desc")
152 parallel::clusterExport(cl, varlist, envir=environment())
24ed5d83
BA
153 }
154
0486fbad
BA
155 if (verbose)
156 {
157 if (verbose)
158 cat(paste("--- Compute ",K," synchrones with ",nb_ref_curves," series\n", sep=""))
159 }
492cd9e7 160 indices_workers = .spreadIndices(seq_len(nb_ref_curves), nb_series_per_chunk)
c45fd663 161 ignored <-
492cd9e7 162 if (parll)
e161499b 163 parallel::parLapply(cl, indices_workers, computeSynchronesChunk)
492cd9e7 164 else
c45fd663 165 lapply(indices_workers, computeSynchronesChunk)
492cd9e7 166
24ed5d83
BA
167 if (parll)
168 parallel::stopCluster(cl)
169
0486fbad
BA
170 if (!sync_mean)
171 return (synchrones)
172
173 #TODO: can we avoid this loop? ( synchrones = sweep(synchrones, 2, counts, '/') )
492cd9e7 174 for (i in seq_len(K))
eef6f6c9 175 synchrones[,i] = synchrones[,i] / counts[i]
3eef8d3d 176 #NOTE: odds for some clusters to be empty? (when series already come from stage 2)
8702eb86 177 # ...maybe; but let's hope resulting K1' be still quite bigger than K2
eef6f6c9 178 noNA_rows = sapply(seq_len(K), function(i) all(!is.nan(synchrones[,i])))
24ed5d83
BA
179 if (all(noNA_rows))
180 return (synchrones)
181 # Else: some clusters are empty, need to slice synchrones
eef6f6c9 182 bigmemory::as.big.matrix(synchrones[,noNA_rows])
e205f218 183}
1c6f223e 184
4bcfdbee
BA
185#' computeWerDists
186#'
187#' Compute the WER distances between the synchrones curves (in rows), which are
188#' returned (e.g.) by \code{computeSynchrones()}
189#'
24ed5d83
BA
190#' @param synchrones A big.matrix of synchrones, in rows. The series have same length
191#' as the series in the initial dataset
492cd9e7 192#' @inheritParams claws
4bcfdbee 193#'
777c4b02 194#' @return A matrix of size K1 x K1
24ed5d83 195#'
4bcfdbee 196#' @export
a174b8ea 197computeWerDists = function(synchrones, nbytes,endian,ncores_clust=1,verbose=FALSE,parll=TRUE)
d03c0621 198{
4bcfdbee
BA
199 n <- nrow(synchrones)
200 delta <- ncol(synchrones)
db6fc17d 201 #TODO: automatic tune of all these parameters ? (for other users)
d03c0621 202 nvoice <- 4
4bcfdbee 203 # noctave = 2^13 = 8192 half hours ~ 180 days ; ~log2(ncol(synchrones))
d7d55bc1
BA
204 noctave = 13
205 # 4 here represent 2^5 = 32 half-hours ~ 1 day
db6fc17d 206 #NOTE: default scalevector == 2^(0:(noctave * nvoice) / nvoice) * s0 (?)
24ed5d83 207 scalevector <- 2^(4:(noctave * nvoice) / nvoice + 1)
db6fc17d 208 #condition: ( log2(s0*w0/(2*pi)) - 1 ) * nvoice + 1.5 >= 1
a174b8ea
BA
209 s0 = 2
210 w0 = 2*pi
db6fc17d
BA
211 scaled=FALSE
212 s0log = as.integer( (log2( s0*w0/(2*pi) ) - 1) * nvoice + 1.5 )
213 totnoct = noctave + as.integer(s0log/nvoice) + 1
214
e161499b 215 Xwer_dist <- bigmemory::big.matrix(nrow=n, ncol=n, type="double")
e161499b 216
4204e877
BA
217 cwt_file = ".epclust_bin/cwt"
218 #TODO: args, nb_per_chunk, nbytes, endian
219
e161499b
BA
220 # Generate n(n-1)/2 pairs for WER distances computations
221 pairs = list()
4204e877
BA
222 V = seq_len(n)
223 for (i in 1:n)
e161499b
BA
224 {
225 V = V[-1]
4204e877
BA
226 pairs = c(pairs, lapply(V, function(v) c(i,v)))
227 }
a174b8ea 228
4204e877
BA
229 computeSaveCWT = function(index)
230 {
231 ts <- scale(ts(synchrones[index,]), center=TRUE, scale=scaled)
232 totts.cwt = Rwave::cwt(ts, totnoct, nvoice, w0, plot=FALSE)
233 ts.cwt = totts.cwt[,s0log:(s0log+noctave*nvoice)]
234 #Normalization
235 sqs <- sqrt(2^(0:(noctave*nvoice)/nvoice)*s0)
236 sqres <- sweep(ts.cwt,2,sqs,'*')
237 res <- sqres / max(Mod(sqres))
238 #TODO: serializer les CWT, les récupérer via getDataInFile ;
239 #--> OK, faut juste stocker comme séries simples de taille delta*ncol (53*17519)
a174b8ea 240 binarize(c(as.double(Re(res)),as.double(Im(res))), cwt_file, ncol(res), ",", nbytes, endian)
4204e877
BA
241 }
242
243 if (parll)
244 {
245 cl = parallel::makeCluster(ncores_clust)
246 synchrones_desc <- bigmemory::describe(synchrones)
247 Xwer_dist_desc <- bigmemory::describe(Xwer_dist)
248 parallel::clusterExport(cl, varlist=c("synchrones_desc","Xwer_dist_desc","totnoct",
249 "nvoice","w0","s0log","noctave","s0","verbose","getCWT"), envir=environment())
250 }
0486fbad
BA
251
252 if (verbose)
253 {
254 cat(paste("--- Compute WER dists\n", sep=""))
255 # precompute save all CWT........
256 }
4204e877
BA
257 #precompute and serialize all CWT
258 ignored <-
259 if (parll)
260 parallel::parLapply(cl, 1:n, computeSaveCWT)
261 else
262 lapply(1:n, computeSaveCWT)
263
264 getCWT = function(index)
265 {
266 #from cwt_file ...
a174b8ea 267 res <- getDataInFile(c(2*index-1,2*index), cwt_file, nbytes, endian)
eef6f6c9 268 ###############TODO:
4204e877 269 }
e161499b 270
777c4b02 271 # Distance between rows i and j
e161499b
BA
272 computeDistancesIJ = function(pair)
273 {
2c14dbea 274 if (parll)
363ae134 275 {
2c14dbea
BA
276 require("bigmemory", quietly=TRUE)
277 require("epclust", quietly=TRUE)
278 synchrones <- bigmemory::attach.big.matrix(synchrones_desc)
279 Xwer_dist <- bigmemory::attach.big.matrix(Xwer_dist_desc)
280 }
281
e161499b
BA
282 i = pair[1] ; j = pair[2]
283 if (verbose && j==i+1)
284 cat(paste(" Distances (",i,",",j,"), (",i,",",j+1,") ...\n", sep=""))
4204e877
BA
285 cwt_i <- getCWT(i)
286 cwt_j <- getCWT(j)
2c14dbea 287
363ae134
BA
288 num <- epclustFilter(Mod(cwt_i * Conj(cwt_j)))
289 WX <- epclustFilter(Mod(cwt_i * Conj(cwt_i)))
4204e877 290 WY <- epclustFilter(Mod(cwt_j * Conj(cwt_j)))
e161499b 291 wer2 <- sum(colSums(num)^2) / sum(colSums(WX) * colSums(WY))
2c14dbea 292 Xwer_dist[i,j] <- sqrt(delta * ncol(cwt_i) * max(1 - wer2, 0.)) #FIXME: wer2 should be < 1
e161499b
BA
293 Xwer_dist[j,i] <- Xwer_dist[i,j]
294 Xwer_dist[i,i] = 0.
295 }
296
0486fbad
BA
297 if (verbose)
298 {
299 cat(paste("--- Compute WER dists\n", sep=""))
300 }
e161499b 301 ignored <-
492cd9e7 302 if (parll)
e161499b 303 parallel::parLapply(cl, pairs, computeDistancesIJ)
492cd9e7 304 else
e161499b 305 lapply(pairs, computeDistancesIJ)
492cd9e7
BA
306
307 if (parll)
308 parallel::stopCluster(cl)
6ad3f3fd 309
492cd9e7 310 Xwer_dist[n,n] = 0.
777c4b02
BA
311 distances <- Xwer_dist[,]
312 rm(Xwer_dist) ; gc()
313 distances #~small matrix K1 x K1
492cd9e7
BA
314}
315
316# Helper function to divide indices into balanced sets
0486fbad 317.spreadIndices = function(indices, nb_per_set)
492cd9e7
BA
318{
319 L = length(indices)
0486fbad 320 nb_workers = floor( L / nb_per_set )
0fe757f7 321 rem = L %% nb_per_set
37c82bba 322 if (nb_workers == 0 || (nb_workers==1 && rem==0))
492cd9e7 323 {
0fe757f7 324 # L <= nb_per_set, simple case
492cd9e7
BA
325 indices_workers = list(indices)
326 }
327 else
328 {
329 indices_workers = lapply( seq_len(nb_workers), function(i)
0fe757f7 330 indices[(nb_per_set*(i-1)+1):(nb_per_set*i)] )
0486fbad
BA
331 # Spread the remaining load among the workers
332 rem = L %% nb_per_set
492cd9e7
BA
333 while (rem > 0)
334 {
335 index = rem%%nb_workers + 1
336 indices_workers[[index]] = c(indices_workers[[index]], indices[L-rem+1])
337 rem = rem - 1
d03c0621 338 }
1c6f223e 339 }
492cd9e7 340 indices_workers
1c6f223e 341}