several fixes; still some issues db,bin != ascii,csv

[epclust.git] / TODO

diff --git a/TODO b/TODO
index f5e0015..53a82b3 100644 (file)
--- a/TODO
+++ b/TODO
@@ -26,3 +26,47 @@ utiliser Rcpp ?
 #fct qui pour deux series (ID, medoides) renvoie distance WER (Rwave ou à moi)
 #transformee croisee , smoothing lissage 3 composantes , + calcul pour WER
 #determiner nvoice noctave (entre octave + petit et + grand)
+
+#TODO: load some dataset ASCII CSV
+#data_bin_file <<- "/tmp/epclust_test.bin"
+#unlink(data_bin_file)
+
+#https://stat.ethz.ch/pipermail/r-help/2011-June/280133.html
+#randCov = function(d)
+#{
+#      x <- matrix(rnorm(d*d), nrow=d)
+#      x <- x / sqrt(rowSums(x^2))
+#      x %*% t(x)
+#}
+
+#TODO: soften condition clustering.R line 37 ?
+#regarder mapply et mcmapply pour le // (pas OK pour Windows ou GUI... mais ?)
+#TODO: map-reduce more appropriate R/clustering.R ligne 88
+#Alternative: use bigmemory to share series when CSV or matrix(...)
+
+#' @importFrom synchronicity boost.mutex lock unlock
+
+subtree: epclust, shared. This root folder should remain private
+
+#TODO: use dbs(),
+               #https://www.r-bloggers.com/debugging-parallel-code-with-dbs/
+               #http://gforge.se/2015/02/how-to-go-parallel-in-r-basics-tips/
+
+synchrones --> somme, pas moyenne
+
+PLOT:
+plot manifold 2D distances WER /
+fenetre tempo forme des courbes /
+medoids /
+gain en prevision: clust puis full --> enercast
+
+réduire taille 17519 trop long ?
+
+synchrone : sum
+cwt : trim R part
+// : clever by rows retenir cwt...
+
+Stockage matrices : en colonnes systématiquement ?
+
+TODO: revoir les arguments, simplifier (dans les clustering...),
+  permettre algos de clustering quelconques, args: medoids (curves puis dists), K