| 1 | library("Rwave") |
| 2 | |
| 3 | #Entrée : courbes synchrones, soit après étape 1 itérée, soit après chaqure étape 1 |
| 4 | step2 = function(conso) |
| 5 | { |
| 6 | n <- nrow(conso) |
| 7 | delta <- ncol(conso) |
| 8 | #TODO: automatic tune of all these parameters ? (for other users) |
| 9 | nvoice <- 4 |
| 10 | # noctave = 2^13 = 8192 half hours ~ 180 days ; ~log2(ncol(conso)) |
| 11 | noctave = 13 |
| 12 | # 4 here represent 2^5 = 32 half-hours ~ 1 day |
| 13 | #NOTE: default scalevector == 2^(0:(noctave * nvoice) / nvoice) * s0 (?) |
| 14 | scalevector <- 2^(4:(noctave * nvoice) / nvoice) * 2 |
| 15 | #condition: ( log2(s0*w0/(2*pi)) - 1 ) * nvoice + 1.5 >= 1 |
| 16 | s0=2 |
| 17 | w0=2*pi |
| 18 | scaled=FALSE |
| 19 | s0log = as.integer( (log2( s0*w0/(2*pi) ) - 1) * nvoice + 1.5 ) |
| 20 | totnoct = noctave + as.integer(s0log/nvoice) + 1 |
| 21 | |
| 22 | # (normalized) observations node with CWT |
| 23 | Xcwt4 <- lapply(seq_len(n), function(i) { |
| 24 | ts <- scale(ts(conso[i,]), center=TRUE, scale=scaled) |
| 25 | totts.cwt = Rwave::cwt(ts,totnoct,nvoice,w0,plot=0) |
| 26 | ts.cwt = totts.cwt[,s0log:(s0log+noctave*nvoice)] |
| 27 | #Normalization |
| 28 | sqs <- sqrt(2^(0:(noctave*nvoice)/nvoice)*s0) |
| 29 | sqres <- sweep(ts.cwt,MARGIN=2,sqs,'*') |
| 30 | sqres / max(Mod(sqres)) |
| 31 | }) |
| 32 | |
| 33 | Xwer_dist <- matrix(0., n, n) |
| 34 | fcoefs = rep(1/3, 3) #moving average on 3 values (TODO: very slow! correct?!) |
| 35 | for (i in 1:(n-1)) |
| 36 | { |
| 37 | for (j in (i+1):n) |
| 38 | { |
| 39 | #TODO: later, compute CWT here (because not enough storage space for 32M series) |
| 40 | # 'circular=TRUE' is wrong, should just take values on the sides; to rewrite in C |
| 41 | num <- filter(Mod(Xcwt4[[i]] * Conj(Xcwt4[[j]])), fcoefs, circular=TRUE) |
| 42 | WX <- filter(Mod(Xcwt4[[i]] * Conj(Xcwt4[[i]])), fcoefs, circular=TRUE) |
| 43 | WY <- filter(Mod(Xcwt4[[j]] * Conj(Xcwt4[[j]])), fcoefs, circular=TRUE) |
| 44 | wer2 <- sum(colSums(num)^2) / sum( sum(colSums(WX) * colSums(WY)) ) |
| 45 | Xwer_dist[i,j] <- sqrt(delta * ncol(Xcwt4[[1]]) * (1 - wer2)) |
| 46 | Xwer_dist[j,i] <- Xwer_dist[i,j] |
| 47 | } |
| 48 | } |
| 49 | diag(Xwer_dist) <- numeric(n) |
| 50 | Xwer_dist |
| 51 | } |