drop enercast submodule; drop Rcpp requirement; fix doc, complete code, fix fix fix
[epclust.git] / epclust / R / main.R
... / ...
CommitLineData
1#' CLAWS: CLustering with wAvelets and Wer distanceS
2#'
3#' Cluster electricity power curves (or any series of similar nature) by applying a
4#' two stage procedure in parallel (see details).
5#' Input series must be sampled on the same time grid, no missing values.
6#'
7#' Summary of the function execution flow:
8#' \enumerate{
9#' \item Compute and serialize all contributions, obtained through discrete wavelet
10#' decomposition (see Antoniadis & al. [2013])
11#' \item Divide series into \code{ntasks} groups to process in parallel. In each task:
12#' \enumerate{
13#' \item iterate the first clustering algorithm on its aggregated outputs,
14#' on inputs of size \code{nb_series_per_chunk}
15#' \item optionally, if WER=="mix":
16#' a) compute the K1 synchrones curves,
17#' a) compute WER distances (K1xK1 matrix) between medoids and
18#' b) apply the second clustering algorithm (output: K2 indices)
19#' }
20#' \item Launch a final task on the aggregated outputs of all previous tasks:
21#' ntasks*K1 if WER=="end", ntasks*K2 otherwise
22#' \item Compute synchrones (sum of series within each final group)
23#' }
24#' \cr
25#' The main argument -- \code{series} -- has a quite misleading name, since it can be
26#' either a [big.]matrix, a CSV file, a connection or a user function to retrieve series.
27#' When \code{series} is given as a function, it must take a single argument,
28#' 'indices', integer vector equal to the indices of the curves to retrieve;
29#' see SQLite example.
30#' WARNING: the return value must be a matrix (in columns), or NULL if no matches.
31#' \cr
32#' Note: Since we don't make assumptions on initial data, there is a possibility that
33#' even when serialized, contributions do not fit in RAM. For example,
34#' 30e6 series of length 100,000 would lead to a +4Go contribution matrix. Therefore,
35#' it's safer to place these in (binary) files; that's what we do.
36#'
37#' @param series Access to the (time-)series, which can be of one of the three
38#' following types:
39#' \itemize{
40#' \item [big.]matrix: each column contains the (time-ordered) values of one time-serie
41#' \item connection: any R connection object providing lines as described above
42#' \item character: name of a CSV file containing series in rows (no header)
43#' \item function: a custom way to retrieve the curves; it has only one argument:
44#' the indices of the series to be retrieved. See SQLite example
45#' }
46#' @param K1 Number of clusters to be found after stage 1 (K1 << N [number of series])
47#' @param K2 Number of clusters to be found after stage 2 (K2 << K1)
48#' @param nb_series_per_chunk (Maximum) number of series to retrieve in one batch
49#' @param nb_items_clust (~Maximum) number of items in clustering algorithm 1 input
50#' @param algoClust1 Clustering algorithm for stage 1. A function which takes (data, K)
51#' as argument where data is a matrix in columns and K the desired number of clusters,
52#' and outputs K medoids ranks. Default: PAM. In our method, this function is called
53#' on iterated medoids during stage 1
54#' @param algoClust2 Clustering algorithm for stage 2. A function which takes (dists, K)
55#' as argument where dists is a matrix of distances and K the desired number of clusters,
56#' and outputs K medoids ranks. Default: PAM. In our method, this function is called
57#' on a matrix of K1 x K1 (WER) distances computed between medoids after algorithm 1
58#' @param wav_filt Wavelet transform filter; see ?wavelets::wt.filter
59#' @param contrib_type Type of contribution: "relative", "logit" or "absolute" (any prefix)
60#' @param WER "end" to apply stage 2 after stage 1 has fully iterated, or "mix" to apply
61#' stage 2 at the end of each task
62#' @param smooth_lvl Smoothing level: odd integer, 1 == no smoothing. 3 seems good
63#' @param nvoice Number of voices within each octave for CWT computations
64#' @param random TRUE (default) for random chunks repartition
65#' @param ntasks Number of tasks (parallel iterations to obtain K1 [if WER=="end"]
66#' or K2 [if WER=="mix"] medoids); default: 1.
67#' Note: ntasks << N (number of series), so that N is "roughly divisible" by ntasks
68#' @param ncores_tasks Number of parallel tasks (1 to disable: sequential tasks)
69#' @param ncores_clust Number of parallel clusterings in one task (3 should be a minimum)
70#' @param sep Separator in CSV input file (if any provided)
71#' @param nbytes Number of bytes to serialize a floating-point number; 4 or 8
72#' @param endian Endianness for (de)serialization ("little" or "big")
73#' @param verbose Level of verbosity (0/FALSE for nothing or 1/TRUE for all; devel stage)
74#' @param parll TRUE to fully parallelize; otherwise run sequentially (debug, comparison)
75#'
76#' @return A list with
77#' \itemize{
78#' medoids: a matrix of the final K2 medoids curves, in columns
79#' ranks: corresponding indices in the dataset
80#' synchrones: a matrix of the K2 sum of series within each final group
81#' }
82#'
83#' @references Clustering functional data using Wavelets [2013];
84#' A. Antoniadis, X. Brossat, J. Cugliari & J.-M. Poggi.
85#' Inter. J. of Wavelets, Multiresolution and Information Procesing,
86#' vol. 11, No 1, pp.1-30. doi:10.1142/S0219691313500033
87#'
88#' @examples
89#' \dontrun{
90#' # WER distances computations are too long for CRAN (for now)
91#'
92#' # Random series around cos(x,2x,3x)/sin(x,2x,3x)
93#' x <- seq(0,500,0.05)
94#' L <- length(x) #10001
95#' ref_series <- matrix( c(cos(x),cos(2*x),cos(3*x),sin(x),sin(2*x),sin(3*x)), ncol=6 )
96#' library(wmtsa)
97#' series <- do.call( cbind, lapply( 1:6, function(i)
98#' do.call(cbind, wmtsa::wavBootstrap(ref_series[,i], n.realization=400)) ) )
99#' #dim(series) #c(2400,10001)
100#' res_ascii <- claws(series, K1=60, K2=6, 200, verbose=TRUE)
101#'
102#' # Same example, from CSV file
103#' csv_file <- "/tmp/epclust_series.csv"
104#' write.table(series, csv_file, sep=",", row.names=FALSE, col.names=FALSE)
105#' res_csv <- claws(csv_file, K1=60, K2=6, 200)
106#'
107#' # Same example, from binary file
108#' bin_file <- "/tmp/epclust_series.bin"
109#' nbytes <- 8
110#' endian <- "little"
111#' binarize(csv_file, bin_file, 500, nbytes, endian)
112#' getSeries <- function(indices) getDataInFile(indices, bin_file, nbytes, endian)
113#' res_bin <- claws(getSeries, K1=60, K2=6, 200)
114#' unlink(csv_file)
115#' unlink(bin_file)
116#'
117#' # Same example, from SQLite database
118#' library(DBI)
119#' series_db <- dbConnect(RSQLite::SQLite(), "file::memory:")
120#' # Prepare data.frame in DB-format
121#' n <- nrow(series)
122#' time_values <- data.frame(
123#' id <- rep(1:n,each=L),
124#' time <- rep( as.POSIXct(1800*(0:n),"GMT",origin="2001-01-01"), L ),
125#' value <- as.double(t(series)) )
126#' dbWriteTable(series_db, "times_values", times_values)
127#' # Fill associative array, map index to identifier
128#' indexToID_inDB <- as.character(
129#' dbGetQuery(series_db, 'SELECT DISTINCT id FROM time_values')[,"id"] )
130#' serie_length <- as.integer( dbGetQuery(series_db,
131#' paste("SELECT COUNT * FROM time_values WHERE id == ",indexToID_inDB[1],sep="")) )
132#' getSeries <- function(indices) {
133#' request <- "SELECT id,value FROM times_values WHERE id in ("
134#' for (i in indices)
135#' request <- paste(request, indexToID_inDB[i], ",", sep="")
136#' request <- paste(request, ")", sep="")
137#' df_series <- dbGetQuery(series_db, request)
138#' if (length(df_series) >= 1)
139#' as.matrix(df_series[,"value"], nrow=serie_length)
140#' else
141#' NULL
142#' }
143#' res_db <- claws(getSeries, K1=60, K2=6, 200))
144#' dbDisconnect(series_db)
145#'
146#' # All results should be the same:
147#' library(digest)
148#' digest::sha1(res_ascii)
149#' digest::sha1(res_csv)
150#' digest::sha1(res_bin)
151#' digest::sha1(res_db)
152#' }
153#' @export
154claws <- function(series, K1, K2, nb_series_per_chunk, nb_items_clust=7*K1,
155 algoClust1=function(data,K) cluster::pam(t(data),K,diss=FALSE,pamonce=1)$id.med,
156 algoClust2=function(dists,K) cluster::pam(dists,K,diss=TRUE,pamonce=1)$id.med,
157 wav_filt="d8", contrib_type="absolute", WER="end", smooth_lvl=3, nvoice=4,
158 random=TRUE, ntasks=1, ncores_tasks=1, ncores_clust=3, sep=",", nbytes=4,
159 endian=.Platform$endian, verbose=FALSE, parll=TRUE)
160{
161 # Check/transform arguments
162 if (!is.matrix(series) && !bigmemory::is.big.matrix(series)
163 && !is.function(series)
164 && !methods::is(series,"connection") && !is.character(series))
165 {
166 stop("'series': [big]matrix, function, file or valid connection (no NA)")
167 }
168 K1 <- .toInteger(K1, function(x) x>=2)
169 K2 <- .toInteger(K2, function(x) x>=2)
170 nb_series_per_chunk <- .toInteger(nb_series_per_chunk, function(x) x>=1)
171 nb_items_clust <- .toInteger(nb_items_clust, function(x) x>K1)
172 random <- .toLogical(random)
173 tryCatch({ignored <- wavelets::wt.filter(wav_filt)},
174 error=function(e) stop("Invalid wavelet filter; see ?wavelets::wt.filter") )
175 ctypes <- c("relative","absolute","logit")
176 contrib_type <- ctypes[ pmatch(contrib_type,ctypes) ]
177 if (is.na(contrib_type))
178 stop("'contrib_type' in {'relative','absolute','logit'}")
179 if (WER!="end" && WER!="mix")
180 stop("'WER': in {'end','mix'}")
181 random <- .toLogical(random)
182 ntasks <- .toInteger(ntasks, function(x) x>=1)
183 ncores_tasks <- .toInteger(ncores_tasks, function(x) x>=1)
184 ncores_clust <- .toInteger(ncores_clust, function(x) x>=1)
185 if (!is.character(sep))
186 stop("'sep': character")
187 nbytes <- .toInteger(nbytes, function(x) x==4 || x==8)
188 verbose <- .toLogical(verbose)
189 parll <- .toLogical(parll)
190
191 # Binarize series if it is not a function; the aim is to always use a function,
192 # to uniformize treatments. An equally good alternative would be to use a file-backed
193 # bigmemory::big.matrix, but it would break the "all-is-function" pattern.
194 if (!is.function(series))
195 {
196 if (verbose)
197 cat("...Serialize time-series (or retrieve past binary file)\n")
198 series_file <- ".series.epclust.bin"
199 if (!file.exists(series_file))
200 binarize(series, series_file, nb_series_per_chunk, sep, nbytes, endian)
201 getSeries <- function(inds) getDataInFile(inds, series_file, nbytes, endian)
202 }
203 else
204 getSeries <- series
205
206 # Serialize all computed wavelets contributions into a file
207 contribs_file <- ".contribs.epclust.bin"
208 index <- 1
209 nb_curves <- 0
210 if (verbose)
211 cat("...Compute contributions and serialize them (or retrieve past binary file)\n")
212 if (!file.exists(contribs_file))
213 {
214 nb_curves <- binarizeTransform(getSeries,
215 function(curves) curvesToContribs(curves, wav_filt, contrib_type),
216 contribs_file, nb_series_per_chunk, nbytes, endian)
217 }
218 else
219 {
220 # TODO: duplicate from getDataInFile() in de_serialize.R
221 contribs_size <- file.info(contribs_file)$size #number of bytes in the file
222 contrib_length <- readBin(contribs_file, "integer", n=1, size=8, endian=endian)
223 nb_curves <- (contribs_size-8) / (nbytes*contrib_length)
224 }
225 getContribs <- function(indices) getDataInFile(indices, contribs_file, nbytes, endian)
226
227 # A few sanity checks: do not continue if too few data available.
228 if (nb_curves < K2)
229 stop("Not enough data: less series than final number of clusters")
230 nb_series_per_task <- round(nb_curves / ntasks)
231 if (nb_series_per_task < K2)
232 stop("Too many tasks: less series in one task than final number of clusters")
233
234 # Generate a random permutation of 1:N (if random==TRUE);
235 # otherwise just use arrival (storage) order.
236 indices_all <- if (random) sample(nb_curves) else seq_len(nb_curves)
237 # Split (all) indices into ntasks groups of ~same size
238 indices_tasks <- lapply(seq_len(ntasks), function(i) {
239 upper_bound <- ifelse( i<ntasks, min(nb_series_per_task*i,nb_curves), nb_curves )
240 indices_all[((i-1)*nb_series_per_task+1):upper_bound]
241 })
242
243 if (parll && ntasks>1)
244 {
245 # Initialize parallel runs: outfile="" allow to output verbose traces in the console
246 # under Linux. All necessary variables are passed to the workers.
247 cl <- parallel::makeCluster(ncores_tasks, outfile="")
248 varlist <- c("ncores_clust","verbose","parll", #task 1 & 2
249 "K1","getContribs","algoClust1","nb_items_clust") #task 1
250 if (WER=="mix")
251 {
252 # Add variables for task 2
253 varlist <- c(varlist, "K2","getSeries","algoClust2","nb_series_per_chunk",
254 "smooth_lvl","nvoice","nbytes","endian")
255 }
256 parallel::clusterExport(cl, varlist, envir <- environment())
257 }
258
259 # This function achieves one complete clustering task, divided in stage 1 + stage 2.
260 # stage 1: n indices --> clusteringTask1(...) --> K1 medoids (indices)
261 # stage 2: K1 indices --> K1xK1 WER distances --> clusteringTask2(...) --> K2 medoids,
262 # where n == N / ntasks, N being the total number of curves.
263 runTwoStepClustering <- function(inds)
264 {
265 # When running in parallel, the environment is blank: we need to load the required
266 # packages, and pass useful variables.
267 if (parll && ntasks>1)
268 require("epclust", quietly=TRUE)
269 indices_medoids <- clusteringTask1(inds, getContribs, K1, algoClust1,
270 nb_items_clust, ncores_clust, verbose, parll)
271 if (WER=="mix")
272 {
273 indices_medoids <- clusteringTask2(indices_medoids, getSeries, K2, algoClust2,
274 nb_series_per_chunk,smooth_lvl,nvoice,nbytes,endian,ncores_clust,verbose,parll)
275 }
276 indices_medoids
277 }
278
279 if (verbose)
280 {
281 message <- paste("...Run ",ntasks," x stage 1", sep="")
282 if (WER=="mix")
283 message <- paste(message," + stage 2", sep="")
284 cat(paste(message,"\n", sep=""))
285 }
286
287 # As explained above, we obtain after all runs ntasks*[K1 or K2] medoids indices,
288 # depending wether WER=="end" or "mix", respectively.
289 indices_medoids_all <-
290 if (parll && ntasks>1)
291 unlist( parallel::parLapply(cl, indices_tasks, runTwoStepClustering) )
292 else
293 unlist( lapply(indices_tasks, runTwoStepClustering) )
294
295 if (parll && ntasks>1)
296 parallel::stopCluster(cl)
297
298 # For the last stage, ncores_tasks*(ncores_clusts+1) cores should be available:
299 # - ntasks for level 1 parallelism
300 # - ntasks*ncores_clust for level 2 parallelism,
301 # but since an extension MPI <--> tasks / OpenMP <--> sub-tasks is on the way,
302 # it's better to just re-use ncores_clust
303 ncores_last_stage <- ncores_clust
304
305 # Run last clustering tasks to obtain only K2 medoids indices
306 if (verbose)
307 cat("...Run final // stage 1 + stage 2\n")
308 indices_medoids <- clusteringTask1(indices_medoids_all, getContribs, K1, algoClust1,
309 nb_items_clust, ncores_tasks*ncores_clust, verbose, parll)
310 indices_medoids <- clusteringTask2(indices_medoids, getContribs, K2, algoClust2,
311 nb_series_per_chunk,smooth_lvl,nvoice,nbytes,endian,ncores_last_stage,verbose,parll)
312
313 # Compute synchrones, that is to say the cumulated power consumptions for each of the K2
314 # final groups.
315 medoids <- getSeries(indices_medoids)
316 synchrones <- computeSynchrones(medoids, getSeries, nb_curves, nb_series_per_chunk,
317 ncores_last_stage, verbose, parll)
318
319 # NOTE: no need to use big.matrix here, since there are only K2 << K1 << N remaining curves
320 list("medoids"=medoids, "ranks"=indices_medoids, "synchrones"=synchrones)
321}