Files
tunmnlu/task_2/others-answer/omsa-main/ISYE-8803-OAN/exam2/.Rhistory
louiscklaw 9035c1312b update,
2025-02-01 02:09:32 +08:00

513 lines
10 KiB
R

n2 <- 100
A <- as.matrix(non_random_miss)
r <- 2
decomp <- svd(A)
U <- decomp$u
s <- decomp$d
V <- decomp$v
s[(r+1): length(s)] <- 0
S <-diag(s)
#svd decomp
X <- U %*% S %*% t(V)
X0 <- X
missing_observations <-which(A==0)
A <- ifelse(A == 0, TRUE, FALSE)
X[A]<-0
m <- sum(sum(A==0))
# Initialization
Y <- matrix(0,n1,n2)
delta <- (n1*n2)/m
lambda <- 5
tau <- delta*lambda
# Iterations
vec <- rep(0,500)
err <- rep(0,500)
for (i in 1:500){
decomp <- svd(Y)
U <- decomp$u
S <- decomp$d
V <- decomp$v
S_t <- S-tau
S_t[S_t<0] <- 0
Z_nonrandom <- U%*%diag(S_t)%*%t(V)
P <- X-Z_nonrandom
P[A] <- 0
Y0 <- Y
Y <- Z_nonrandom + delta*P
vec[i] <- sum(sum((Y-Y0)^2))
err[i] <- sum(sum((X0-Z_nonrandom)^2))/sum(sum((X0)^2))
}
plot(vec,type="l")
plot(err,type="l")
Ar <- as.vector(A)
Xr <- as.vector(X0)
Xr_nonrandom <- Xr[Ar]
Zr <- as.vector(Z_nonrandom)
Zr_nonrandom <- Zr[Ar]
#dev.off()
par(mfrow=c(1,2))
plot(Xr_nonrandom,type="l")
lines(Zr_nonrandom, col="red",lty=2)
plot(Xr_nonrandom-Zr_nonrandom,type="l")
#dev.off()
par(mfrow=c(1,2))
imagesc(Z_nonrandom, main ="M")
imagesc(original, main ="M0")
par(mfrow=c(1,3))
imagesc(Z_random, main ="RandomMiss")
imagesc(Z_nonrandom, main ="NonRandomMiss")
imagesc(original, main="Original")
n1 <- 100
n2 <- 100
A <- as.matrix(random_miss)
r <- 2
decomp <- svd(A)
U <- decomp$u
s <- decomp$d
V <- decomp$v
s[(r+1): length(s)] <- 0
S <-diag(s)
#svd decomp
X <- U %*% S %*% t(V)
X0 <- X
missing_observations <-which(A==0)
A <- ifelse(A == 0, TRUE, FALSE)
X[A]<-0
m <- sum(sum(A==0))
# Initialization
Y <- matrix(0,n1,n2)
delta <- (n1*n2)/m
lambda <- 5
tau <- delta*lambda
# Iterations
vec <- rep(0,500)
err <- rep(0,500)
for (i in 1:500){
decomp <- svd(Y)
U <- decomp$u
S <- decomp$d
V <- decomp$v
S_t <- S-tau
S_t[S_t<0] <- 0
Z_random <- U%*%diag(S_t)%*%t(V)
P <- X-Z_random
P[A] <- 0
Y0 <- Y
Y <- Z_random + delta*P
vec[i] <- sum(sum((Y-Y0)^2))
err[i] <- sum(sum((X0-Z_random)^2))/sum(sum((X0)^2))
}
plot(vec,type="l")
plot(err,type="l")
Ar <- as.vector(A)
Xr <- as.vector(X0)
Xr_random <- Xr[Ar]
Zr <- as.vector(Z_random)
Zr_random <- Zr[Ar]
#dev.off()
par(mfrow=c(1,2))
plot(Xr_random,type="l")
lines(Zr_random, col="red",lty=2)
plot(Xr_random-Zr_random,type="l")
#dev.off()
par(mfrow=c(1,2))
imagesc(Z_random, main ="M")
imagesc(original, main ="M0")
n1 <- 100
n2 <- 100
A <- as.matrix(non_random_miss)
r <- 2
decomp <- svd(A)
U <- decomp$u
s <- decomp$d
V <- decomp$v
s[(r+1): length(s)] <- 0
S <-diag(s)
#svd decomp
X <- U %*% S %*% t(V)
X0 <- X
missing_observations <-which(A==0)
A <- ifelse(A == 0, TRUE, FALSE)
X[A]<-0
m <- sum(sum(A==0))
# Initialization
Y <- matrix(0,n1,n2)
delta <- (n1*n2)/m
lambda <- 5
tau <- delta*lambda
# Iterations
vec <- rep(0,500)
err <- rep(0,500)
for (i in 1:500){
decomp <- svd(Y)
U <- decomp$u
S <- decomp$d
V <- decomp$v
S_t <- S-tau
S_t[S_t<0] <- 0
Z_nonrandom <- U%*%diag(S_t)%*%t(V)
P <- X-Z_nonrandom
P[A] <- 0
Y0 <- Y
Y <- Z_nonrandom + delta*P
vec[i] <- sum(sum((Y-Y0)^2))
err[i] <- sum(sum((X0-Z_nonrandom)^2))/sum(sum((X0)^2))
}
plot(vec,type="l")
plot(err,type="l")
Ar <- as.vector(A)
Xr <- as.vector(X0)
Xr_nonrandom <- Xr[Ar]
Zr <- as.vector(Z_nonrandom)
Zr_nonrandom <- Zr[Ar]
#dev.off()
par(mfrow=c(1,2))
plot(Xr_nonrandom,type="l")
lines(Zr_nonrandom, col="red",lty=2)
plot(Xr_nonrandom-Zr_nonrandom,type="l")
#dev.off()
par(mfrow=c(1,2))
imagesc(Z_nonrandom, main ="M")
imagesc(original, main ="M0")
A\
A
Z_nonrandom
knitr::opts_chunk$set(echo = TRUE)
library(R.matlab)
library(matlab)
n1 <- 50
n2 <- 50
A <- matrix(sample(-50:50,n1*n2, replace=TRUE),n1,n2)
r <- 2
decomp <- svd(A)
U <- decomp$u
s <- decomp$d
V <- decomp$v
s[(r+1): length(s)] <- 0
S <-diag(s)
X <- U %*% S %*% t(V)
X0 <- X
# Removing 20% of the observations
A <- matrix(runif(n1*n2, min = 0, max = 1),n1,n2)>0.5
X[A] <- 0
m <- sum(sum(A==0))
m
# Initialization
Y <- matrix(0,n1,n2)
delta <- (n1*n2)/m
tau <- 250
# Iterations
vec <- rep(0,500)
err <- rep(0,500)
for (i in 1:500){
decomp <- svd(Y)
U <- decomp$u
S <- decomp$d
V <- decomp$v
S_t <- S-tau
S_t[S_t<0] <- 0
Z <- U%*%diag(S_t)%*%t(V)
P <- X-Z
P[A] <- 0
Y0 <- Y
Y <- Y0 + delta*P
vec[i] <- sum(sum((Y-Y0)^2))
err[i] <- sum(sum((X0-Z)^2))/sum(sum((X0)^2))
}
#plot(vec,type="l")
#plot(err,type="l")
Ar <- as.vector(A)
Xr <- as.vector(X0)
Xr <- Xr[Ar]
Zr <- as.vector(Z)
Zr <- Zr[Ar]
#dev.off()
par(mfrow=c(1,2))
plot(Xr,type="l")
lines(Zr, col="red",lty=2)
plot(Xr-Zr,type="l")
#dev.off()
par(mfrow=c(1,2))
imagesc(Z, main ="M")
imagesc(X0, main ="M0")
A
Z
#Read in the matrix files
library(R.matlab)
library(matlab)
#Read in the matrix files
random_miss <- read.csv('./RandomMiss.csv',header=FALSE)
colnames(random_miss) <- c(paste0("X_", 1:100))
random_miss <- as.matrix(random_miss)
non_random_miss <- read.csv('./NonRandomMiss.csv',header=FALSE)
colnames(non_random_miss) <- c(paste0("X_", 1:100))
non_random_miss <- as.matrix(non_random_miss)
original <- read.csv('./Original.csv',header=FALSE)
colnames(original) <- c(paste0("X_", 1:100))
original <- as.matrix(original)
n1 <- 100
n2 <- 100
A <- as.matrix(random_miss)
r <- 2
decomp <- svd(A)
U <- decomp$u
s <- decomp$d
V <- decomp$v
s[(r+1): length(s)] <- 0
S <-diag(s)
#svd decomp
X <- U %*% S %*% t(V)
X0 <- X
missing_observations <-which(A==0)
A <- ifelse(A == 0, TRUE, FALSE)
X[A]<-0
m <- sum(sum(A==0))
# Initialization
Y <- matrix(0,n1,n2)
delta <- (n1*n2)/m
lambda <- 5
tau <- delta*lambda
# Iterations
vec <- rep(0,500)
err <- rep(0,500)
for (i in 1:500){
decomp <- svd(Y)
U <- decomp$u
S <- decomp$d
V <- decomp$v
S_t <- S-tau
S_t[S_t<0] <- 0
Z_random <- U%*%diag(S_t)%*%t(V)
P <- X-Z_random
P[A] <- 0
Y0 <- Y
Y <- Z_random + delta*P
vec[i] <- sum(sum((Y-Y0)^2))
err[i] <- sum(sum((X0-Z_random)^2))/sum(sum((X0)^2))
}
plot(vec,type="l")
plot(err,type="l")
Ar <- as.vector(A)
Xr <- as.vector(X0)
Xr_random <- Xr[Ar]
Zr <- as.vector(Z_random)
Zr_random <- Zr[Ar]
#dev.off()
par(mfrow=c(1,2))
plot(Xr_random,type="l")
lines(Zr_random, col="red",lty=2)
plot(Xr_random-Zr_random,type="l")
#dev.off()
par(mfrow=c(1,2))
imagesc(Z_random, main ="M")
imagesc(original, main ="M0")
A
Z
Z_random
n1 <- 100
n2 <- 100
A <- as.matrix(non_random_miss)
r <- 2
decomp <- svd(A)
U <- decomp$u
s <- decomp$d
V <- decomp$v
s[(r+1): length(s)] <- 0
S <-diag(s)
#svd decomp
X <- U %*% S %*% t(V)
X0 <- X
missing_observations <-which(A==0)
A <- ifelse(A == 0, TRUE, FALSE)
X[A]<-0
m <- sum(sum(A==0))
# Initialization
Y <- matrix(0,n1,n2)
delta <- (n1*n2)/m
lambda <- 5
tau <- delta*lambda
# Iterations
vec <- rep(0,500)
err <- rep(0,500)
for (i in 1:500){
decomp <- svd(Y)
U <- decomp$u
S <- decomp$d
V <- decomp$v
S_t <- S-tau
S_t[S_t<0] <- 0
Z_nonrandom <- U%*%diag(S_t)%*%t(V)
P <- X-Z_nonrandom
P[A] <- 0
Y0 <- Y
Y <- Z_nonrandom + delta*P
vec[i] <- sum(sum((Y-Y0)^2))
err[i] <- sum(sum((X0-Z_nonrandom)^2))/sum(sum((X0)^2))
}
plot(vec,type="l")
plot(err,type="l")
Ar <- as.vector(A)
Xr <- as.vector(X0)
Xr_nonrandom <- Xr[Ar]
Zr <- as.vector(Z_nonrandom)
Zr_nonrandom <- Zr[Ar]
#dev.off()
par(mfrow=c(1,2))
plot(Xr_nonrandom,type="l")
lines(Zr_nonrandom, col="red",lty=2)
plot(Xr_nonrandom-Zr_nonrandom,type="l")
#dev.off()
par(mfrow=c(1,2))
imagesc(Z_nonrandom, main ="M")
imagesc(original, main ="M0")
n1 <- 100
n2 <- 100
A <- as.matrix(random_miss)
r <- 2
decomp <- svd(A)
U <- decomp$u
s <- decomp$d
V <- decomp$v
s[(r+1): length(s)] <- 0
S <-diag(s)
#svd decomp
X <- U %*% S %*% t(V)
X0 <- X
missing_observations <-which(A==0)
A <- ifelse(A == 0, TRUE, FALSE)
X[A]<-0
m <- sum(sum(A==0))
# Initialization
Y <- matrix(0,n1,n2)
delta <- (n1*n2)/m
lambda <- 5
tau <- delta*lambda
# Iterations
vec <- rep(0,500)
err <- rep(0,500)
for (i in 1:500){
decomp <- svd(Y)
U <- decomp$u
S <- decomp$d
V <- decomp$v
S_t <- S-tau
S_t[S_t<0] <- 0
Z_random <- U%*%diag(S_t)%*%t(V)
P <- X-Z_random
P[A] <- 0
Y0 <- Y
Y <- Z_random + delta*P
vec[i] <- sum(sum((Y-Y0)^2))
err[i] <- sum(sum((X0-Z_random)^2))/sum(sum((X0)^2))
}
plot(vec,type="l")
plot(err,type="l")
Ar <- as.vector(A)
Xr <- as.vector(X0)
Xr_random <- Xr[Ar]
Zr <- as.vector(Z_random)
Zr_random <- Zr[Ar]
#dev.off()
par(mfrow=c(1,2))
plot(Xr_random,type="l")
lines(Zr_random, col="red",lty=2)
plot(Xr_random-Zr_random,type="l")
#dev.off()
par(mfrow=c(1,2))
imagesc(Z_random, main ="M")
imagesc(original, main ="M0")
#Read in the matrix files
library(R.matlab)
library(matlab)
#Read in the matrix files
random_miss <- read.csv('./RandomMiss.csv',header=FALSE)
colnames(random_miss) <- c(paste0("X_", 1:100))
random_miss <- as.matrix(random_miss)
non_random_miss <- read.csv('./NonRandomMiss.csv',header=FALSE)
colnames(non_random_miss) <- c(paste0("X_", 1:100))
non_random_miss <- as.matrix(non_random_miss)
original <- read.csv('./Original.csv',header=FALSE)
colnames(original) <- c(paste0("X_", 1:100))
original <- as.matrix(original)
#Read in the matrix files
random_miss <- read.csv('./RandomMiss.csv',header=FALSE)
colnames(random_miss) <- c(paste0("X_", 1:100))
random_miss <- as.matrix(random_miss)
non_random_miss <- read.csv('./NonRandomMiss.csv',header=FALSE)
colnames(non_random_miss) <- c(paste0("X_", 1:100))
non_random_miss <- as.matrix(non_random_miss)
original <- read.csv('./Original.csv',header=FALSE)
colnames(original) <- c(paste0("X_", 1:100))
original <- as.matrix(original)
n1 <- 100
n2 <- 100
A <- as.matrix(random_miss)
r <- 2
decomp <- svd(A)
U <- decomp$u
s <- decomp$d
V <- decomp$v
s[(r+1): length(s)] <- 0
S <-diag(s)
#svd decomp
X <- U %*% S %*% t(V)
X0 <- X
missing_observations <-which(A==0)
A <- ifelse(A == 0, TRUE, FALSE)
X[A]<-0
m <- sum(sum(A==0))
# Initialization
Y <- matrix(0,n1,n2)
delta <- (n1*n2)/m
lambda <- 5
tau <- delta*lambda
# Iterations
vec <- rep(0,500)
err <- rep(0,500)
for (i in 1:500){
decomp <- svd(Y)
U <- decomp$u
S <- decomp$d
V <- decomp$v
S_t <- S-tau
S_t[S_t<0] <- 0
Z_random <- U%*%diag(S_t)%*%t(V)
P <- X-Z_random
P[A] <- 0
Y0 <- Y
Y <- Z_random + delta*P
vec[i] <- sum(sum((Y-Y0)^2))
err[i] <- sum(sum((X0-Z_random)^2))/sum(sum((X0)^2))
}
plot(vec,type="l")
plot(err,type="l")
Ar <- as.vector(A)
Xr <- as.vector(X0)
Xr_random <- Xr[Ar]
Zr <- as.vector(Z_random)
Zr_random <- Zr[Ar]
#dev.off()
par(mfrow=c(1,2))
plot(Xr_random,type="l")
lines(Zr_random, col="red",lty=2)
plot(Xr_random-Zr_random,type="l")
#dev.off()
par(mfrow=c(1,2))
imagesc(Z_random, main ="M")
imagesc(original, main ="M0")
A
Z
View(Z_random)
#Read in the matrix files
library(R.matlab)
library(matlab)
#Read in the matrix files
library(R.matlab)
library(matlab)
#Read in the matrix files
library(R.matlab)
library(matlab)
#Read in the matrix files
library(R.matlab)
library(matlab)