epclust/tests/testthat/test.clusteringTask.R

   1 computeClusters
   2 computeSynchrones
   3
   4 # Cluster one full task (nb_curves / ntasks series)
   5 clusteringTask = function(indices,getSeries,getSeriesForSynchrones,synchrones_file,
   6     getCoefs,K1,K2,nb_series_per_chunk,ncores,to_file,ftype)
   7 {
   8     cl = parallel::makeCluster(ncores)
   9     repeat
  10     {
  11         nb_workers = max( 1, round( length(indices) / nb_series_per_chunk ) )
  12         indices_workers = lapply(seq_len(nb_workers), function(i) {
  13             upper_bound = ifelse( i<nb_workers,
  14                 min(nb_series_per_chunk*i,length(indices)), length(indices) )
  15             indices[(nb_series_per_chunk*(i-1)+1):upper_bound]
  16         })
  17         indices = unlist( parallel::parLapply(cl, indices_workers, function(inds)
  18             computeClusters1(inds, getCoefs, K1)) )
  19         if (length(indices_clust) == K1)
  20             break
  21     }
  22     parallel::stopCluster(cl)
  23     if (K2 == 0)
  24         return (indices)
  25     computeClusters2(indices, K2, getSeries, getSeriesForSynchrones, to_file,
  26                                      nb_series_per_chunk,ftype)
  27     vector("integer",0)
  28 }
  29
  30 # Apply the clustering algorithm (PAM) on a coeffs or distances matrix
  31 computeClusters1 = function(indices, getCoefs, K1)
  32 {
  33     coefs = getCoefs(indices)
  34     indices[ cluster::pam(coefs, K1, diss=FALSE)$id.med ]
  35 }
  36
  37 # Cluster a chunk of series inside one task (~max nb_series_per_chunk)
  38 computeClusters2 = function(indices, K2, getSeries, getSeriesForSynchrones, to_file,