tests/test_shoppingmall_small.R

In dfleis/morpca: Manifold Optimization for Robust PCA

#===== LIBRARIES =====#
#library(devtools)
#install_github("dfleis/morpca")
library(morpca)
library(matrixvis)

#library(Matrix)
library(R.matlab) # reading .mat files

#===== LOAD DATA =====#
shoppingmall_mat <- readMat("test/yi/data/shoppingmall.mat")

Y <- as.matrix(shoppingmall_mat$X)[,(81920 - 256*25 + 1):81920]
image_matrix(matrix(Y[1,], nrow = 256),
             col = colorRampPalette(c("black", "white"))(16))

#===== SET PARAMETERS =====#
set.seed(124)
r <- 3
gamma <- 0.1
retraction <- "o"
stepsize <- 0.7
maxiter <- 100

#===== OPTIMIZE =====#
pt <- proc.time()
L.opt <- morpca(Y = Y, r = r, gamma = gamma,
                sparsity   = matrix(1, nrow = nrow(Y), ncol = ncol(Y)),
                retraction = retraction,
                stepsize   = stepsize,
                maxiter    = maxiter,
                stepsout   = F,
                verbose    = T)
proc.time() - pt


#===== VISUALIZE IN/OUTPUTS =====#
frame.idx <- 2

L <- L.opt$solution

# inputs
input.tmp0 <- Y[frame.idx, 1:ncol(Y)]
input.tmp1 <- matrix(input.tmp0, nrow = 256, byrow = F)
input.plot <- t(apply(input.tmp1, 2, rev))
image(input.plot, col = colorRampPalette(c("black", "white"))(64),
      xaxt = 'n', yaxt = 'n', main = "input")

# outputs
output.tmp0 <- L[frame.idx, 1:ncol(L)]
output.tmp1 <- matrix(output.tmp0, nrow = 256, byrow = F)
output.plot <- t(apply(output.tmp1, 2, rev))
image(output.plot, col = colorRampPalette(c("black", "white"))(64),
      xaxt = 'n', yaxt = 'n', main = "output")

# difference
diff.plot <- input.plot - output.plot
image(diff.plot, col = colorRampPalette(c("black", "white"))(64),
      xaxt = 'n', yaxt = 'n', main = "difference")

obj <- L.opt$objective
plot(obj[2:length(obj)], type = 'o', pch = 21, bg = 'white',
     log = 'y', cex = 0.75, main = "Objective")
plot(abs(diff(obj[2:length(obj)])), type = 'o', pch = 21, bg = 'white',
     log = 'y', cex = 0.75, main = "Objective")