Superfast Likelihood Inference for Stationary Gaussian Time Series
In SuperGauss: Superfast Likelihood Inference for Stationary Gaussian Time Series

\newcommand{\bm}[1]{\boldsymbol{#1}} \newcommand{\rv}[3][1]{#2_{#1},\ldots,#2_{#3}} \newcommand{\X}{\bm{X}} \newcommand{\cov}{\mathrm{cov}} \newcommand{\dt}{\Delta t} \newcommand{\msd}{\mathrm{\scriptsize MSD}} \newcommand{\acf}{\mathrm{\scriptsize ACF}} \newcommand{\dX}{\Delta\X} \newcommand{\VH}{\bm{V}_H}

This vignette illustrates the basic functionality of the SuperGauss package by simulating a few stochastic processes and estimating their parameters from regularly spaced data.

Simulation of Fractional Brownian Motion

A one-dimensional fractional Brownian motion (fBM) $X_t = X(t)$ is a continuous Gaussian process with $E[X_t] = 0$ and $\cov(X_t, X_s) = \tfrac 1 2 (|t|^{2H} + |s|^{2H} - |t-s|^{2H})$, for $0 < H < 1$. fBM is not stationary but has stationary increments, such that $(X_{t+\dt} - X_t) \stackrel{D}{=} (X_{s+\dt} - X_s)$ for any $s,t$. As such, its covariance function is completely determined its mean squared displacement (MSD) $$ \msd_X(t) = E[(X_t - X_0)^2] = |t|^{2H}. $$ When the Hurst parameter $H = \tfrac 1 2$, fBM reduces to ordinary Brownian motion.

require(SuperGauss)

N <- 5000 # number of observations
dT <- 1/60 # time between observations (seconds)
H <- .3 # Hurst parameter

tseq <- (0:N)*dT # times at which to sample fBM
npaths <- 5 # number of fBM paths to generate

# to generate fbm, generate its increments, which are stationary
msd <- fbm_msd(tseq = tseq[-1], H = H)
acf <- msd2acf(msd = msd) # convert msd to acf

# superfast method
system.time({
  dX <- rnormtz(n = npaths, acf = acf, fft = TRUE)
})
# fast method (about 3x as slow)
system.time({
  rnormtz(n = npaths, acf = acf, fft = FALSE)
})
# unstructured variance method (much slower)
system.time({
  matrix(rnorm(N*npaths), npaths, N) %*% chol(toeplitz(acf))
})

The following R code generates r npaths independent fBM realizations of length $N = r N$ with $H = r H$. The timing of the "superfast" method [@wood.chan94] provided in this package is compared to that of a "fast" method [e.g., @brockwell.davis91] and to the usual method (Cholesky decomposition of an unstructured variance matrix).

# convert increments to position measurements
Xt <- apply(rbind(0, dX), 2, cumsum)

# plot
clrs <- c("black", "red", "blue", "orange", "green2")
par(mar = c(4.1,4.1,.5,.5))
plot(0, type = "n", xlim = range(tseq), ylim = range(Xt),
     xlab = "Time (s)", ylab = "Position (m)")
for(ii in 1:npaths) {
  lines(tseq, Xt[,ii], col = clrs[ii], lwd = 2)
}

Inference for the Hurst Parameter

Suppose that $\X = (\rv [0] X N)$ are equally spaced observations of an fBM process with $X_i = X(i \dt)$, and let $\dX = (\rv [0] {\Delta X} {N-1})$ denote the corresponding increments, $\Delta X_i = X_{i+1} - X_i$. Then the loglikelihood function for $H$ is $$ \ell(H \mid \dX) = -\tfrac 1 2 \big(\dX' \VH^{-1} \dX + \log |\VH|\big), $$ where $V_H$ is a Toeplitz matrix, $$ \VH = [\cov(\Delta X_i, \Delta X_j)]{0 \le i,j < N} = \begin{bmatrix} \gamma_0 & \gamma_1 & \cdots & \gamma{N-1} \ \gamma_1 & \gamma_0 & \cdots & \gamma_{N-2} \ \vdots & \vdots & \ddots & \vdots \ \gamma_{N-1} & \gamma_{N-2} & \cdots & \gamma_0 \end{bmatrix}. $$ Thus, each evaluation of the loglikelihood requires the inverse and log-determinant of a Toeplitz matrix, which scales as $\mathcal O(N^2)$ with the Durbin-Levinson algorithm. The SuperGauss package implements an extended version of the Generalized Schur algorithm of @ammar.gragg88, which scales these computations as $\mathcal O(N \log^2 N)$. With careful memory management and extensive use of the FFTW library [@frigo.johnson05], the SuperGauss implementation crosses over Durbin-Levinson at around $N = 300$.

The `Toeplitz` Matrix Class

The bulk of the likelihood calculations in SuperGauss are handled by the Toeplitz matrix class. A Toeplitz object is created as follows:

# allocate and assign in one step
Tz <- Toeplitz$new(acf = acf)
Tz

# allocate memory only
Tz <- Toeplitz$new(N = N)
Tz
Tz$set_acf(acf = acf) # assign later

Its primary methods are illustrated below:

all(acf == Tz$get_acf()) # extract acf

# matrix multiplication
z <- rnorm(N)
x1 <- toeplitz(acf) %*% z # regular way
x2 <- Tz$prod(z) # with Toeplitz class
x3 <- Tz %*% z # with Toeplitz class overloading the `%*%` operator
range(x1-x2)
range(x2-x3)

# system of equations
y1 <- solve(toeplitz(acf), z) # regular way
y2 <- Tz$solve(z) # with Toeplitz class
y2 <- solve(Tz, z) # same thing but overloading `solve()`
range(y1-y2)

# log-determinant
ld1 <- determinant(toeplitz(acf))$mod
ld2 <- Tz$log_det() # with Toeplitz class
ld2 <- determinant(Tz) # same thing but overloading `determinant()`
                         # note: no $mod
c(ld1, ld2)

Maximum Likelihood Calculation

The following code shows how to obtain the maximum likelihood of $H$ and its standard error for a given fBM path. The log-PDF of the Gaussian with Toeplitz variance matrix is obtained either with SuperGauss::dnormtz(), or using the NormalToeplitz class. The advantage of the latter is that it does not reallocate memory for the underlying Toeplitz object at every likelihood evaulation.

For speed comparisons, the optimization underlying the MLE calculation is done both using the superfast Generalized Schur algorithm and the fast Durbin-Levinson algorithm.

dX <- diff(Xt[,1]) # obtain the increments of a given path
N <- length(dX)

# autocorrelation of fBM increments
fbm_acf <- function(H) {
  msd <- fbm_msd(1:N*dT, H = H)
  msd2acf(msd)
}

# loglikelihood using generalized Schur algorithm
NTz <- NormalToeplitz$new(N = N) # pre-allocate memory
loglik_GS <- function(H) {
  NTz$logdens(z = dX, acf = fbm_acf(H))
}

# loglikelihood using Durbin-Levinson algorithm
loglik_DL <- function(H) {
  dnormtz(X = dX, acf = fbm_acf(H), method = "ltz", log = TRUE)
}

# superfast method
system.time({
  GS_mle <- optimize(loglik_GS, interval = c(.01, .99), maximum = TRUE)
})
# fast method (about 10x slower)
system.time({
  DL_mle <- optimize(loglik_DL, interval = c(.01, .99), maximum = TRUE)
})
c(GS = GS_mle$max, DL = DL_mle$max)

# standard error calculation
require(numDeriv)
Hmle <- GS_mle$max
Hse <- -hessian(func = loglik_GS, x = Hmle) # observed Fisher Information
Hse <- sqrt(1/Hse[1])
c(mle = Hmle, se = Hse)

Caution with `R6` Classes

In order to effectively manage memory in the underlying C++ code, the Toeplitz class is implemented using R6 classes. Among other things, this means that when a Toeplitz object is passed to a function, the function does not make a copy of it: all modifications to the object inside the object are reflected on the object outside the function as well, as in the following example:

T1 <- Toeplitz$new(N = N)
T2 <- T1 # shallow copy: both of these point to the same memory location

# affects both objects
T1$set_acf(fbm_acf(.5))
T1
T2

fbm_logdet <- function(H) {
  T1$set_acf(acf = fbm_acf(H))
  T1$log_det()
}

# affects both objects
fbm_logdet(H = .3)
T1
T2

To avoid this behavior, it is necessary to make a deep copy of the object:

T3 <- T1$clone(deep = TRUE)
T1
T3

# only affect T1
fbm_logdet(H = .7)
T1
T3

Superfast Newton-Raphson

In addition to the superfast algorithm for Gaussian likelihood evaluations, SuperGauss provides such algorithms for the loglikelihood gradient and Hessian functions, leading to superfast versions of many inference algorithms such as Newton-Raphson and Hamiltonian Monte Carlo. These are provided by the NormalToeplitz$grad() and NormalToeplitz$hess() methods. Both of these methods optionally return the lower order derivatives as well, reusing common computations to improve performance. An example of Newton-Raphson is given below using the two-parameter exponential autocorrelation model $$ \acf_X(t \mid \lambda, \sigma) = \sigma^2 \exp(- |t/\lambda|). $$ The example uses stats::nlm() for optimization, which requires the derivatives to be passsed as attributes to the (negative) loglikelihood.

# autocorrelation function
exp_acf <- function(t, lambda, sigma) sigma^2 * exp(-abs(t/lambda))
# gradient, returned as a 2-column matrix
exp_acf_grad <- function(t, lambda, sigma) {
  ea <- exp_acf(t, lambda, 1)
  cbind(abs(t)*(sigma/lambda)^2 * ea, # d_acf/d_lambda
        2*sigma * ea) # d_acf/d_sigma
}
# Hessian, returned as an array of size length(t) x 2 x 2
exp_acf_hess <- function(t, lambda, sigma) {
  ea <- exp_acf(t, lambda, 1)
  sl2 <- sigma/lambda^2
  hess <- array(NA, dim = c(length(t), 2, 2))
  hess[,1,1] <- sl2^2*(t^2 - 2*abs(t)*lambda) * ea # d2_acf/d_lambda^2
  hess[,1,2] <- 2*sl2 * abs(t) * ea # d2_acf/(d_lambda d_sigma)
  hess[,2,1] <- hess[,1,2] # d2_acf/(d_sigma d_lambda)
  hess[,2,2] <- 2 * ea # d2_acf/d_sigma^2
  hess
}

# simulate data
lambda <- runif(1, .5, 2)
sigma <- runif(1, .5, 2)
tseq <- (1:N-1)*dT
acf <- exp_acf(t = tseq, lambda = lambda, sigma = sigma)
Xt <- rnormtz(acf = acf)

NTz <- NormalToeplitz$new(N = N) # storage space

# negative loglikelihood function of theta = (lambda, sigma)
# include attributes for gradient and Hessian
exp_negloglik <- function(theta) {
  lambda <- theta[1]
  sigma <- theta[2]
  # acf, its gradient, and Hessian
  acf <- exp_acf(tseq, lambda, sigma)
  dacf <- exp_acf_grad(tseq, lambda, sigma)
  d2acf <- exp_acf_hess(tseq, lambda, sigma)
  # derivatives of NormalToeplitz up to order 2
  derivs <- NTz$hess(z = Xt,
                     dz = matrix(0, N, 2),
                     d2z = array(0, dim = c(N, 2, 2)),
                     acf = acf,
                     dacf = dacf,
                     d2acf = d2acf,
                     full_out = TRUE)
  # negative loglikelihood with derivatives as attributes
  nll <- -1 * derivs$ldens
  attr(nll, "gradient") <- -1 * derivs$grad
  attr(nll, "hessian") <- -1 * derivs$hess
  nll
}

# optimization
system.time({
  mle_fit <- nlm(f = exp_negloglik, p = c(1,1), hessian = TRUE)
})

# display estimates with standard errors
rbind(true = c(lambda = lambda, sigma = sigma),
      est = mle_fit$estimate,
      se = sqrt(diag(solve(mle_fit$hessian))))