[epclust.git] / epclust / R / main.R

#' @include utils.R
#' @include clustering.R
NULL

#' Cluster power curves with PAM in parallel CLAWS: CLustering with wAvelets and Wer distanceS
#'
#' Groups electricity power curves (or any series of similar nature) by applying PAM
#' algorithm in parallel to chunks of size \code{nb_series_per_chunk}
#'
#' @param data Access to the data, which can be of one of the three following types:
#' \itemize{
#'   \item data.frame: each line contains its ID in the first cell, and all values after
#'   \item connection: any R connection object (e.g. a file) providing lines as described above
#'   \item function: a custom way to retrieve the curves; it has two arguments: the ranks to be
#'     retrieved, and the IDs - at least one of them must be present (priority: ranks).
#' }
#' @param K1 Number of super-consumers to be found after stage 1 (K1 << N)
#' @param K2 Number of clusters to be found after stage 2 (K2 << K1)
#' @param ntasks Number of tasks (parallel iterations to obtain K1 medoids); default: 1.
#'   Note: ntasks << N, so that N is "roughly divisible" by N (number of series)
#' @param nb_series_per_chunk (Maximum) number of series in each group, inside a task
#' @param min_series_per_chunk Minimum number of series in each group
#' @param wf Wavelet transform filter; see ?wt.filter. Default: haar
#' @param WER "end" to apply stage 2 after stage 1 has iterated and finished, or "mix"
#'   to apply it after every stage 1
#' @param ncores_tasks "MPI" number of parallel tasks (1 to disable: sequential tasks)
#' @param ncores_clust "OpenMP" number of parallel clusterings in one task
#' @param random Randomize chunks repartition
#' @param ... Other arguments to be passed to \code{data} function
#'
#' @return A data.frame of the final medoids curves (identifiers + values)
#'
#' @examples
#' getData = function(start, n) {
#'   con = dbConnect(drv = RSQLite::SQLite(), dbname = "mydata.sqlite")
#'   df = dbGetQuery(con, paste(
#'     "SELECT * FROM times_values GROUP BY id OFFSET ",start,
#'     "LIMIT ", n, " ORDER BY date", sep=""))
#'   return (df)
#' }
#' #####TODO: if DB, array rank --> ID at first retrieval, when computing coeffs; so:: NO use of IDs !
#'   #TODO: 3 examples, data.frame / binary file / DB sqLite
#'   + sampleCurves : wavBootstrap de package wmtsa
#' cl = epclust(getData, K1=200, K2=15, ntasks=1000, nb_series_per_chunk=5000, WER="mix")
#' @export
claws = function(getSeries, K1, K2,
	random=TRUE, #randomize series order?
	wf="haar", #stage 1
	WER="end", #stage 2
	ntasks=1, ncores_tasks=1, ncores_clust=4, #control parallelism
	nb_series_per_chunk=50*K1, min_series_per_chunk=5*K1, #chunk size
	sep=",", #ASCII input separator
	nbytes=4, endian=.Platform$endian) #serialization (write,read)
{
	# Check/transform arguments
	if (!is.matrix(getSeries) && !is.function(getSeries) &&
		!is(getSeries, "connection" && !is.character(getSeries)))
	{
		stop("'getSeries': matrix, function, file or valid connection (no NA)")
	}
	K1 = .toInteger(K1, function(x) x>=2)
	K2 = .toInteger(K2, function(x) x>=2)
	if (!is.logical(random))
		stop("'random': logical")
	tryCatch(
		{ignored <- wt.filter(wf)},
		error = function(e) stop("Invalid wavelet filter; see ?wavelets::wt.filter"))
	if (WER!="end" && WER!="mix")
		stop("WER takes values in {'end','mix'}")
	ntasks = .toInteger(ntasks, function(x) x>=1)
	ncores_tasks = .toInteger(ncores_tasks, function(x) x>=1)
	ncores_clust = .toInteger(ncores_clust, function(x) x>=1)
	nb_series_per_chunk = .toInteger(nb_series_per_chunk, function(x) x>=K1)
	min_series_per_chunk = .toInteger(K1, function(x) x>=K1 && x<=nb_series_per_chunk)
	if (!is.character(sep))
		stop("'sep': character")
	nbytes = .toInteger(nbytes, function(x) x==4 || x==8)

	# Serialize series if required, to always use a function
	bin_dir = "epclust.bin/"
	dir.create(bin_dir, showWarnings=FALSE, mode="0755")
	if (!is.function(getSeries))
	{
		series_file = paste(bin_dir,"data",sep="") ; unlink(series_file)
		serialize(getSeries, series_file, nb_series_per_chunk, sep, nbytes, endian)
		getSeries = function(indices) getDataInFile(indices, series_file, nbytes, endian)
	}

	# Serialize all wavelets coefficients (+ IDs) onto a file
	coefs_file = paste(bin_dir,"coefs",sep="") ; unlink(coefs_file)
	index = 1
	nb_curves = 0
	repeat
	{
		series = getSeries((index-1)+seq_len(nb_series_per_chunk))
		if (is.null(series))
			break
		coeffs_chunk = curvesToCoeffs(series, wf)
		serialize(coeffs_chunk, coefs_file, nb_series_per_chunk, sep, nbytes, endian)
		index = index + nb_series_per_chunk
		nb_curves = nb_curves + nrow(coeffs_chunk)
	}
	getCoefs = function(indices) getDataInFile(indices, coefs_file, nbytes, endian)

	if (nb_curves < min_series_per_chunk)
		stop("Not enough data: less rows than min_series_per_chunk!")
	nb_series_per_task = round(nb_curves / ntasks)
	if (nb_series_per_task < min_series_per_chunk)
		stop("Too many tasks: less series in one task than min_series_per_chunk!")

	# Cluster coefficients in parallel (by nb_series_per_chunk)
	indices_all = if (random) sample(nb_curves) else seq_len(nb_curves)
	indices_tasks = lapply(seq_len(ntasks), function(i) {
		upper_bound = ifelse( i<ntasks, min(nb_series_per_task*i,nb_curves), nb_curves )
		indices_all[((i-1)*nb_series_per_task+1):upper_bound]
	})
	cl = parallel::makeCluster(ncores_tasks)
	# 1000*K1 indices [if WER=="end"], or empty vector [if WER=="mix"] --> series on file
	indices = unlist( parallel::parLapply(cl, indices_tasks, function(inds) {
		indices_medoids = clusteringTask(inds,getCoefs,K1,nb_series_per_chunk,ncores_clust)
		if (WER=="mix")
		{
			medoids2 = computeClusters2(
				getSeries(indices_medoids), K2, getSeries, nb_series_per_chunk)
			serialize(medoids2, synchrones_file, nb_series_per_chunk, sep, nbytes, endian)
			return (vector("integer",0))
		}
		indices_medoids
	}) )
	parallel::stopCluster(cl)

	getSeriesForSynchrones = getSeries
	synchrones_file = paste(bin_dir,"synchrones",sep="") ; unlink(synchrones_file)
	if (WER=="mix")
	{
		indices = seq_len(ntasks*K2)
		#Now series must be retrieved from synchrones_file
		getSeries = function(inds) getDataInFile(inds, synchrones_file, nbytes, endian)
		#Coefs must be re-computed
		unlink(coefs_file)
		index = 1
		repeat
		{
			series = getSeries((index-1)+seq_len(nb_series_per_chunk))
			if (is.null(series))
				break
			coeffs_chunk = curvesToCoeffs(series, wf)
			serialize(coeffs_chunk, coefs_file, nb_series_per_chunk, sep, nbytes, endian)
			index = index + nb_series_per_chunk
		}
	}

	# Run step2 on resulting indices or series (from file)
	indices_medoids = clusteringTask(
		indices, getCoefs, K1, nb_series_per_chunk, ncores_tasks*ncores_clust)
	computeClusters2(getSeries(indices_medoids),K2,getSeriesForSynchrones,nb_series_per_chunk)
}

# helper
curvesToCoeffs = function(series, wf)
{
	L = length(series[1,])
	D = ceiling( log2(L) )
	nb_sample_points = 2^D
	apply(series, 1, function(x) {
		interpolated_curve = spline(1:L, x, n=nb_sample_points)$y
		W = wavelets::dwt(interpolated_curve, filter=wf, D)@W
		rev( sapply( W, function(v) ( sqrt( sum(v^2) ) ) ) )
	})
}

# helper
.toInteger <- function(x, condition)
{
	if (!is.integer(x))
		tryCatch(
			{x = as.integer(x)[1]},
			error = function(e) paste("Cannot convert argument",substitute(x),"to integer")
		)
	if (!condition(x))
		stop(paste("Argument",substitute(x),"does not verify condition",body(condition)))
	x
}
Commit	Line	Data
56857861 BA	1	#' @include utils.R
	2	#' @include clustering.R
	3	NULL
	4
	5	#' Cluster power curves with PAM in parallel CLAWS: CLustering with wAvelets and Wer distanceS
7f0781b7	6	#'
56857861	7	#' Groups electricity power curves (or any series of similar nature) by applying PAM
cea14f3a	8	#' algorithm in parallel to chunks of size \code{nb_series_per_chunk}
7f0781b7 BA	9	#'
	10	#' @param data Access to the data, which can be of one of the three following types:
	11	#' \itemize{
	12	#' \item data.frame: each line contains its ID in the first cell, and all values after
	13	#' \item connection: any R connection object (e.g. a file) providing lines as described above
c33af7e4 BA	14	#' \item function: a custom way to retrieve the curves; it has two arguments: the ranks to be
c33af7e4 BA	15	#' retrieved, and the IDs - at least one of them must be present (priority: ranks).
7f0781b7	16	#' }
1c6f223e BA	17	#' @param K1 Number of super-consumers to be found after stage 1 (K1 << N)
	18	#' @param K2 Number of clusters to be found after stage 2 (K2 << K1)
	19	#' @param ntasks Number of tasks (parallel iterations to obtain K1 medoids); default: 1.
	20	#' Note: ntasks << N, so that N is "roughly divisible" by N (number of series)
	21	#' @param nb_series_per_chunk (Maximum) number of series in each group, inside a task
cea14f3a	22	#' @param min_series_per_chunk Minimum number of series in each group
3465b246	23	#' @param wf Wavelet transform filter; see ?wt.filter. Default: haar
7f0781b7 BA	24	#' @param WER "end" to apply stage 2 after stage 1 has iterated and finished, or "mix"
7f0781b7 BA	25	#' to apply it after every stage 1
5c652979 BA	26	#' @param ncores_tasks "MPI" number of parallel tasks (1 to disable: sequential tasks)
	27	#' @param ncores_clust "OpenMP" number of parallel clusterings in one task
	28	#' @param random Randomize chunks repartition
0e2dce80	29	#' @param ... Other arguments to be passed to \code{data} function
7f0781b7 BA	30	#'
7f0781b7 BA	31	#' @return A data.frame of the final medoids curves (identifiers + values)
1c6f223e BA	32	#'
	33	#' @examples
	34	#' getData = function(start, n) {
	35	#' con = dbConnect(drv = RSQLite::SQLite(), dbname = "mydata.sqlite")
	36	#' df = dbGetQuery(con, paste(
	37	#' "SELECT * FROM times_values GROUP BY id OFFSET ",start,
	38	#' "LIMIT ", n, " ORDER BY date", sep=""))
	39	#' return (df)
	40	#' }
e205f218	41	#' #####TODO: if DB, array rank --> ID at first retrieval, when computing coeffs; so:: NO use of IDs !
5c652979 BA	42	#' #TODO: 3 examples, data.frame / binary file / DB sqLite
5c652979 BA	43	#' + sampleCurves : wavBootstrap de package wmtsa
1c6f223e BA	44	#' cl = epclust(getData, K1=200, K2=15, ntasks=1000, nb_series_per_chunk=5000, WER="mix")
1c6f223e BA	45	#' @export
56857861 BA	46	claws = function(getSeries, K1, K2,
	47	random=TRUE, #randomize series order?
	48	wf="haar", #stage 1
	49	WER="end", #stage 2
	50	ntasks=1, ncores_tasks=1, ncores_clust=4, #control parallelism
	51	nb_series_per_chunk=50K1, min_series_per_chunk=5K1, #chunk size
	52	sep=",", #ASCII input separator
	53	nbytes=4, endian=.Platform$endian) #serialization (write,read)
ac1d4231	54	{
0e2dce80	55	# Check/transform arguments
56857861 BA	56	if (!is.matrix(getSeries) && !is.function(getSeries) &&
56857861 BA	57	!is(getSeries, "connection" && !is.character(getSeries)))
0e2dce80	58	{
56857861	59	stop("'getSeries': matrix, function, file or valid connection (no NA)")
5c652979	60	}
56857861 BA	61	K1 = .toInteger(K1, function(x) x>=2)
	62	K2 = .toInteger(K2, function(x) x>=2)
	63	if (!is.logical(random))
	64	stop("'random': logical")
	65	tryCatch(
	66	{ignored <- wt.filter(wf)},
	67	error = function(e) stop("Invalid wavelet filter; see ?wavelets::wt.filter"))
7f0781b7 BA	68	if (WER!="end" && WER!="mix")
7f0781b7 BA	69	stop("WER takes values in {'end','mix'}")
56857861 BA	70	ntasks = .toInteger(ntasks, function(x) x>=1)
	71	ncores_tasks = .toInteger(ncores_tasks, function(x) x>=1)
	72	ncores_clust = .toInteger(ncores_clust, function(x) x>=1)
	73	nb_series_per_chunk = .toInteger(nb_series_per_chunk, function(x) x>=K1)
	74	min_series_per_chunk = .toInteger(K1, function(x) x>=K1 && x<=nb_series_per_chunk)
	75	if (!is.character(sep))
	76	stop("'sep': character")
	77	nbytes = .toInteger(nbytes, function(x) x==4 \|\| x==8)
	78
	79	# Serialize series if required, to always use a function
	80	bin_dir = "epclust.bin/"
	81	dir.create(bin_dir, showWarnings=FALSE, mode="0755")
	82	if (!is.function(getSeries))
	83	{
	84	series_file = paste(bin_dir,"data",sep="") ; unlink(series_file)
	85	serialize(getSeries, series_file, nb_series_per_chunk, sep, nbytes, endian)
	86	getSeries = function(indices) getDataInFile(indices, series_file, nbytes, endian)
	87	}
ac1d4231	88
7b13d0c2	89	# Serialize all wavelets coefficients (+ IDs) onto a file
56857861	90	coefs_file = paste(bin_dir,"coefs",sep="") ; unlink(coefs_file)
7f0781b7	91	index = 1
cea14f3a	92	nb_curves = 0
6ecf5c2d	93	repeat
ac1d4231	94	{
0e2dce80 BA	95	series = getSeries((index-1)+seq_len(nb_series_per_chunk))
0e2dce80 BA	96	if (is.null(series))
cea14f3a	97	break
0e2dce80	98	coeffs_chunk = curvesToCoeffs(series, wf)
56857861	99	serialize(coeffs_chunk, coefs_file, nb_series_per_chunk, sep, nbytes, endian)
cea14f3a	100	index = index + nb_series_per_chunk
5c652979	101	nb_curves = nb_curves + nrow(coeffs_chunk)
8e6accca	102	}
56857861	103	getCoefs = function(indices) getDataInFile(indices, coefs_file, nbytes, endian)
8e6accca	104
5c652979 BA	105	if (nb_curves < min_series_per_chunk)
	106	stop("Not enough data: less rows than min_series_per_chunk!")
	107	nb_series_per_task = round(nb_curves / ntasks)
	108	if (nb_series_per_task < min_series_per_chunk)
	109	stop("Too many tasks: less series in one task than min_series_per_chunk!")
ac1d4231	110
7b13d0c2	111	# Cluster coefficients in parallel (by nb_series_per_chunk)
56857861	112	indices_all = if (random) sample(nb_curves) else seq_len(nb_curves)
48108c39	113	indices_tasks = lapply(seq_len(ntasks), function(i) {
5c652979	114	upper_bound = ifelse( i<ntasks, min(nb_series_per_task*i,nb_curves), nb_curves )
56857861	115	indices_all[((i-1)*nb_series_per_task+1):upper_bound]
48108c39	116	})
e205f218	117	cl = parallel::makeCluster(ncores_tasks)
56857861	118	# 1000*K1 indices [if WER=="end"], or empty vector [if WER=="mix"] --> series on file
e205f218	119	indices = unlist( parallel::parLapply(cl, indices_tasks, function(inds) {
56857861 BA	120	indices_medoids = clusteringTask(inds,getCoefs,K1,nb_series_per_chunk,ncores_clust)
	121	if (WER=="mix")
	122	{
	123	medoids2 = computeClusters2(
	124	getSeries(indices_medoids), K2, getSeries, nb_series_per_chunk)
	125	serialize(medoids2, synchrones_file, nb_series_per_chunk, sep, nbytes, endian)
	126	return (vector("integer",0))
	127	}
	128	indices_medoids
e205f218 BA	129	}) )
e205f218 BA	130	parallel::stopCluster(cl)
3465b246	131
e205f218	132	getSeriesForSynchrones = getSeries
56857861	133	synchrones_file = paste(bin_dir,"synchrones",sep="") ; unlink(synchrones_file)
e205f218 BA	134	if (WER=="mix")
	135	{
	136	indices = seq_len(ntasks*K2)
	137	#Now series must be retrieved from synchrones_file
56857861	138	getSeries = function(inds) getDataInFile(inds, synchrones_file, nbytes, endian)
e205f218 BA	139	#Coefs must be re-computed
	140	unlink(coefs_file)
	141	index = 1
	142	repeat
	143	{
	144	series = getSeries((index-1)+seq_len(nb_series_per_chunk))
	145	if (is.null(series))
	146	break
	147	coeffs_chunk = curvesToCoeffs(series, wf)
56857861	148	serialize(coeffs_chunk, coefs_file, nb_series_per_chunk, sep, nbytes, endian)
e205f218 BA	149	index = index + nb_series_per_chunk
	150	}
	151	}
0e2dce80 BA	152
0e2dce80 BA	153	# Run step2 on resulting indices or series (from file)
56857861 BA	154	indices_medoids = clusteringTask(
	155	indices, getCoefs, K1, nb_series_per_chunk, ncores_tasks*ncores_clust)
	156	computeClusters2(getSeries(indices_medoids),K2,getSeriesForSynchrones,nb_series_per_chunk)
	157	}
	158
	159	# helper
	160	curvesToCoeffs = function(series, wf)
	161	{
	162	L = length(series[1,])
	163	D = ceiling( log2(L) )
	164	nb_sample_points = 2^D
	165	apply(series, 1, function(x) {
	166	interpolated_curve = spline(1:L, x, n=nb_sample_points)$y
	167	W = wavelets::dwt(interpolated_curve, filter=wf, D)@W
	168	rev( sapply( W, function(v) ( sqrt( sum(v^2) ) ) ) )
	169	})
	170	}
	171
	172	# helper
	173	.toInteger <- function(x, condition)
	174	{
	175	if (!is.integer(x))
	176	tryCatch(
	177	{x = as.integer(x)[1]},
	178	error = function(e) paste("Cannot convert argument",substitute(x),"to integer")
	179	)
	180	if (!condition(x))
	181	stop(paste("Argument",substitute(x),"does not verify condition",body(condition)))
	182	x
cea14f3a	183	}