R/spd-interp-functions.R

In areshenk/spdm: Functions for working with symmetric positive-definite matrices

# Interpolation functions
spd.interpolate.euclidean <- function(x, y, t, ...){
    return((1-t)*x + t*y)
}

spd.interpolate.logeuclidean <- function(x, y, t, ...){
    return(expm2((1-t)*logm2(x) + t*logm2(y)))
}

spd.interpolate.riemannian <- function(x, y, t, ...){

    yt <- spd.logmap(y, p = x)
    return(spd.expmap(t*yt, p = x))

    # x.inv.sqrt <- solve(sqrtm2(x))
    # x.sqrt     <- sqrtm2(x)
    # return(x.sqrt %*% mat.frac.pow(x.inv.sqrt %*% y %*% x.inv.sqrt, t) %*% x.sqrt)
}