tensor_predictors/sim/sim_1e_normal.R

232 lines
8.8 KiB
R

library(tensorPredictors)
# library(RGCCA) # for `mgcca`
### Load modified version which _does not_ create a clusster in case of
### `n_cores == 1` allowing huge speed improvements! (at least on Ubuntu 22.04 LTS)
### Moreover, it is compatible with `Rscript`
devtools::load_all("~/Work/tensorPredictors/References/Software/TGCCA-modified", export_all = FALSE)
setwd("~/Work/tensorPredictors/sim/")
base.name <- format(Sys.time(), "sim_1e_normal-%Y%m%dT%H%M")
# Source utility function used in most simulations (extracted for convenience)
source("./sim_utils.R")
# Set PRNG seed for reproducability
# Note: `0x` is the HEX number prefix and the trailing `L` stands for "long"
# which is `R`s way if indicating an integer.
set.seed(0x1eL, "Mersenne-Twister", "Inversion", "Rejection")
### Simulation configuration
reps <- 100 # number of simulation replications
sample.sizes <- c(100, 200, 300, 500, 750) # sample sizes `n`
dimX <- c(5, 5) # predictor `X` dimension
dimF <- c(2, 2)
Sigma <- 0.5^abs(outer(1:prod(dimX), 1:prod(dimX), `-`)) # AR(0.5)
# # define projections onto tri-diagonal matrixes
# proj.betas <- Map(.projRank, rep(1L, length(dimX))) # wrong assumption of low rank betas
# proj.Omegas <- Map(.projBand, Map(c, dimX, dimX), 1L, 1L) # wrong assumption of band Scatter matrices
# data sampling routine
sample.data <- function(sample.size, dimX, dimF, B.true, Sigma) {
y <- rnorm(sample.size)
# the true F (in vectorized form)
vecF <- rbind(1, sin(y), cos(y), sin(y) * cos(y))
# sample vectorized X as a multi-variate normal (in vectorized form)
vecX <- B.true %*% vecF + t(chol(Sigma)) %*% matrix(rnorm(prod(sample.size, dimX)), prod(dimX))
X <- array(vecX, c(dimX, sample.size))
list(X, vecF, y, length(dim(X)))
}
# wrong assumption about the function `F(y)`
F.wrong <- function(y) array(rbind(1, y, y, y^2), c(2, 2, length(y)))
# Create a CSV logger to write simulation results to
log.file <- paste(base.name, "csv", sep = ".")
logger <- CSV.logger(
file.name = log.file,
header = c("sample.size", "rep", outer(
c("time", "dist.subspace", "dist.projection"), # measures
c("gmlm", "pca", "hopca", "tsir", "mgcca"), # methods
paste, sep = "."
))
)
# Miss-specify true beta as _not_ a kronecker product
B.true <- diag(1, prod(dimX), prod(dimF))
### for each sample size
for (sample.size in sample.sizes) {
# repeate every simulation
for (rep in seq_len(reps)) {
# Sample training data
c(X, F.true, y, sample.axis) %<-% sample.data(sample.size, dimX, dimF, B.true, Sigma)
# fit different models
# Wrapped in try-catch clock to ensure the simulation continues,
# if an error occures continue with nest resplication and log an error message
tryCatch({
time.gmlm <- system.time(
fit.gmlm <- gmlm_tensor_normal(X, F.wrong(y), sample.axis = sample.axis)
)["user.self"]
time.pca <- system.time(
fit.pca <- prcomp(mat(X, sample.axis), rank. = prod(dimF))
)["user.self"]
time.hopca <- system.time(
fit.hopca <- HOPCA(X, npc = dimF, sample.axis = sample.axis)
)["user.self"]
time.tsir <- system.time(
fit.tsir <- TSIR(X, y, dimF, sample.axis = sample.axis)
)["user.self"]
# `mgcca` expects the first axis to be the sample axis
X1 <- aperm(X, c(sample.axis, seq_along(dim(X))[-sample.axis]))
time.mgcca <- system.time(
fit.mgcca <- mgcca(
list(X1, y),
quiet = TRUE,
scheme = "factorial",
ncomp = c(prod(dimF), 1),
ranks = c(prod(dimF), 1)
)
)["user.self"]
}, error = function(ex) {
print(ex)
})
# Compute true reduction matrix
B.gmlm <- with(fit.gmlm, Reduce(`%x%`, rev(betas)))
B.pca <- fit.pca$rotation
B.hopca <- Reduce(`%x%`, rev(fit.hopca))
B.tsir <- Reduce(`%x%`, rev(fit.tsir))
B.mgcca <- fit.mgcca$astar[[1]]
# Subspace Distances: Normalized `|| P_A - P_B ||_F` where
# `P_A = A (A' A)^-1 A'` and the normalization means that with
# respect to the dimensions of `A, B` the subspace distance is in the
# range `[0, 1]`.
dist.subspace.gmlm <- dist.subspace(B.true, B.gmlm, normalize = TRUE)
dist.subspace.pca <- dist.subspace(B.true, B.pca, normalize = TRUE)
dist.subspace.hopca <- dist.subspace(B.true, B.hopca, normalize = TRUE)
dist.subspace.tsir <- dist.subspace(B.true, B.tsir, normalize = TRUE)
dist.subspace.mgcca <- dist.subspace(B.true, B.mgcca, normalize = TRUE)
# Projection Distances: Spectral norm (2-norm) `|| P_A - P_B ||_2`.
dist.projection.gmlm <- dist.projection(B.true, B.gmlm)
dist.projection.pca <- dist.projection(B.true, B.pca)
dist.projection.hopca <- dist.projection(B.true, B.hopca)
dist.projection.tsir <- dist.projection(B.true, B.tsir)
dist.projection.mgcca <- dist.projection(B.true, B.mgcca)
# Call CSV logger writing results to file
logger()
# print progress
cat(sprintf("sample size (%d): %d/%d - rep: %d/%d\n",
sample.size, which(sample.size == sample.sizes),
length(sample.sizes), rep, reps))
}
}
### read simulation results generate plots
if (!interactive()) { pdf(file = paste(base.name, "pdf", sep = ".")) }
sim <- read.csv(log.file)
# Remain sample size grouping variable to avoid conflicts
aggr.mean <- aggregate(sim, list(sampleSize = sim$sample.size), mean)
aggr.median <- aggregate(sim, list(sampleSize = sim$sample.size), median)
aggr.sd <- aggregate(sim, list(sampleSize = sim$sample.size), sd)
aggr.min <- aggregate(sim, list(sampleSize = sim$sample.size), min)
aggr.max <- aggregate(sim, list(sampleSize = sim$sample.size), max)
par(pch = 16, bty = "n", lty = "solid", lwd = 1.5)
cols <- c(gmlm = "blue", pca = "darkcyan", hopca = "red", tsir = "darkgreen",
mgcca = "purple")
with(aggr.mean, {
plot(range(sampleSize), c(0, 1), type = "n",
main = "Subspace Distance",
xlab = "Sample Size",
ylab = "Distance"
)
for (dist.name in ls(pattern = "^dist.subspace")) {
mean <- get(dist.name)
median <- aggr.median[aggr.sd$sampleSize == sampleSize, dist.name]
sd <- aggr.sd[aggr.sd$sampleSize == sampleSize, dist.name]
min <- aggr.min[aggr.sd$sampleSize == sampleSize, dist.name]
max <- aggr.max[aggr.sd$sampleSize == sampleSize, dist.name]
method <- tail(strsplit(dist.name, ".", fixed = TRUE)[[1]], 1)
col <- cols[method]
lines(sampleSize, mean, type = "o", col = col, lty = 1, lwd = 1)
lines(sampleSize, mean + sd, col = col, lty = 2, lwd = 0.8)
lines(sampleSize, mean - sd, col = col, lty = 2, lwd = 0.8)
lines(sampleSize, median, col = col, lty = 1, lwd = 1)
lines(sampleSize, min, col = col, lty = 3, lwd = 0.6)
lines(sampleSize, max, col = col, lty = 3, lwd = 0.6)
}
legend("topright", col = cols, lty = 1, legend = names(cols),
bty = "n", lwd = par("lwd"), pch = par("pch"))
})
with(aggr.mean, {
plot(range(sampleSize), c(0, 1), type = "n",
main = "Projection Distance",
xlab = "Sample Size",
ylab = "Distance"
)
for (dist.name in ls(pattern = "^dist.projection")) {
mean <- get(dist.name)
median <- aggr.median[aggr.sd$sampleSize == sampleSize, dist.name]
sd <- aggr.sd[aggr.sd$sampleSize == sampleSize, dist.name]
min <- aggr.min[aggr.sd$sampleSize == sampleSize, dist.name]
max <- aggr.max[aggr.sd$sampleSize == sampleSize, dist.name]
method <- tail(strsplit(dist.name, ".", fixed = TRUE)[[1]], 1)
col <- cols[method]
lines(sampleSize, mean, type = "o", col = col, lty = 1, lwd = 1)
lines(sampleSize, mean + sd, col = col, lty = 2, lwd = 0.8)
lines(sampleSize, mean - sd, col = col, lty = 2, lwd = 0.8)
lines(sampleSize, median, col = col, lty = 1, lwd = 1)
lines(sampleSize, min, col = col, lty = 3, lwd = 0.6)
lines(sampleSize, max, col = col, lty = 3, lwd = 0.6)
}
legend("topright", col = cols, lty = 1, legend = names(cols),
bty = "n", lwd = par("lwd"), pch = par("pch"))
})
# sample.axis <- 1L
# F.wrong <- array(outer(y, c(0, 1, 1, 2, 1, 2, 2, 3), `^`), c(10, 2, 2, 2))
# F.true <- array(c(1, sin(y), cos(y)))
# 1 s c
# s ss cs
# c sc cc
# s ss cs
# ss sss css
# cs scs ccs
# c sc cc
# sc ssc csc
# cc scc ccc
# osc <-
# I <- array(0, c(3, 3, 3))
# slice.index()