| 1 | #################################################################### |
| 2 | ## |
| 3 | ## File: aux.r |
| 4 | ## |
| 5 | ## Description: Miscelaneous functions for clustering with kcca |
| 6 | ## |
| 7 | ## Modified: june 2010 |
| 8 | ## |
| 9 | #################################################################### |
| 10 | |
| 11 | |
| 12 | ####################################################### |
| 13 | |
| 14 | # Transforms a matrix of data (one observation by row) |
| 15 | # into an array where position[ , , i] gives |
| 16 | # the smoothed modulus of the i-th cwt observation |
| 17 | |
| 18 | ######################################################## |
| 19 | |
| 20 | |
| 21 | ##NOTE: renvoie une matrice 3D |
| 22 | toCWT <- function(X, sw= 0, tw= 0, swabs= 0, |
| 23 | nvoice= 12, noctave= 5, |
| 24 | s0= 2, w0= 2*pi, lt= 24, dt= 0.5, |
| 25 | spectra = FALSE, smooth = TRUE, |
| 26 | scaled = FALSE, |
| 27 | scalevector) |
| 28 | { noctave <- adjust.noctave(lt, dt, s0, tw, noctave) |
| 29 | if(missing(scalevector)) |
| 30 | scalevector <- 2^(0:(noctave * nvoice) / nvoice) * s0 |
| 31 | res <- lapply(1:nrow(X), function(n) |
| 32 | { tsX <- ts( X[n,] ) |
| 33 | tsCent <- tsX - mean(tsX) |
| 34 | if(scaled) tsCent <- ts(scale(tsCent)) |
| 35 | tsCent.cwt <- cwt.ts(tsCent, s0, noctave, nvoice, w0) |
| 36 | tsCent.cwt |
| 37 | } ) |
| 38 | if( spectra ) res <- lapply(res, function(l) Mod(l)^2 ) |
| 39 | if( smooth ) res <- lapply(res, smCWT, swabs = swabs, |
| 40 | tw = tw, dt = dt, |
| 41 | scalevector = scalevector) |
| 42 | resArray <- array(NA, c(nrow(res[[1]]), ncol(res[[1]]), |
| 43 | length(res))) |
| 44 | for( l in 1:length(res) ) resArray[ , , l] <- res[[l]] |
| 45 | resArray |
| 46 | } |
| 47 | |
| 48 | |
| 49 | # =============================================================== |
| 50 | |
| 51 | smCWT <- function(CWT, sw= 0, tw= 0, swabs= 0, |
| 52 | nvoice= 12, noctave= 2, s0= 2, w0= 2*pi, |
| 53 | lt= 24, dt= 0.5, scalevector ) |
| 54 | { |
| 55 | # noctave <- adjust.noctave(lt, dt, s0, tw, noctave) |
| 56 | # scalevector <- 2^(0:(noctave * nvoice) / nvoice) * s0 |
| 57 | wsp <- Mod(CWT) |
| 58 | smwsp <- smooth.matrix(wsp, swabs) |
| 59 | smsmwsp <- smooth.time(smwsp, tw, dt, scalevector) |
| 60 | smsmwsp |
| 61 | } |
| 62 | |
| 63 | |
| 64 | # =============================================================== |
| 65 | |
| 66 | toDWT <- function(x, filter.number = 6, family = "DaubLeAsymm") |
| 67 | { x2 <- spline(x, n = 2^ceiling( log(length(x), 2) ), |
| 68 | method = 'natural')$y |
| 69 | Dx2 <- wd(x2, family = family, filter.number = filter.number)$D |
| 70 | Dx2 |
| 71 | } |
| 72 | |
| 73 | # =============================================================== |
| 74 | |
| 75 | contrib <- function(x) |
| 76 | { J <- log( length(x)+1, 2) |
| 77 | nrj <- numeric(J) |
| 78 | t0 <- 1 |
| 79 | t1 <- 0 |
| 80 | for( j in 1:J ) { |
| 81 | t1 <- t1 + 2^(J-j) |
| 82 | nrj[j] <- sqrt( sum( x[t0:t1]^2 ) ) |
| 83 | t0 <- t1 + 1 |
| 84 | } |
| 85 | return(nrj) |
| 86 | } |
| 87 | |
| 88 | |
| 89 | # ========================================= distance for coh === |
| 90 | |
| 91 | coherence <- function( x, y) |
| 92 | { J <- log(length(x) + 1, 2) |
| 93 | t0 <- 1 |
| 94 | sg2_x <- 0 |
| 95 | sg2_y <- 0 |
| 96 | sg_xy <- 0 |
| 97 | for(j in 0:(J - 1)) |
| 98 | { t1 <- t0 + 2^(J - j)/2 - 1 |
| 99 | tt <- t0:t1 |
| 100 | sg2_x <- sg2_x + mean(x[t0:t1]^2) |
| 101 | sg2_y <- sg2_y + mean(y[t0:t1]^2) |
| 102 | sg_xy <- sg_xy + mean(x[t0:t1] * y[t0:t1]) |
| 103 | t0 <- t1 + 1 |
| 104 | } |
| 105 | res <- sg_xy^2 / sg2_x / sg2_y |
| 106 | res |
| 107 | } |
| 108 | |
| 109 | |
| 110 | vect2mat <- function(vect){ |
| 111 | vect <- as.vector(vect) |
| 112 | matrix(vect[-(1:2)], delta, lscvect) |
| 113 | } |
| 114 | |
| 115 | |
| 116 | # ========================================= # myimg for graphics |
| 117 | jet.colors <- colorRampPalette(c("#00007F", "blue", "#007FFF", |
| 118 | "cyan", "#7FFF7F", "yellow", |
| 119 | "#FF7F00", "red", "#7F0000")) |
| 120 | |
| 121 | myimg <- function(MAT, x = 1:nrow(MAT), y = 1:col(MAT), ... ) |
| 122 | filled.contour( x = x, y = y, z = MAT, |
| 123 | xlab= 'Time', ylab= 'scale', |
| 124 | color.palette = jet.colors, |
| 125 | ... ) |
| 126 | |
| 127 | |