simulation2.R

(for data generation) input parameters:

B <- 200 # No. of simulation repetition \n
n <- 400 # sample size
pn <- 400 # No. of predictors
prior_pi <- 0.4 # prior probabilities for two populations prior_pi and 1-prior_pi
K <- 50 # No. of basis
ck <- readRDS('E:/PhD/assistant works/functional classification/code/ck.rds') # keep fix for all simulations !!
m <- 100 # spaced times
phi <- 0 # measurement errors
tau <- 0 # covariance structure difference between two populations
rho <- 0.2 # controls the correlation among the functional predictors
delta <- 2 # controls the signal strength
rns <- c(1,3,5) # corresponding to the subsettings in each setting

ifGaussian <- FALSE # FALSE - use non-Gaussian ksi_tilde
df <- 3 # for non-Gaussian case

• makeCluster(No. of multiprocess)

outputs:

file_name_X # RDS data - dimension [400, 400, 100] - [n, pn, m] containing training and testing sets
file_name_Y # RDS data - dimension [400]

fclassification.R

(for classification of simulated data) input parameters:

B <- 200

setting <- 8 # setting number(in total 8 general settings)
delta <- 2 # same above
rn <- 5 # same above
phi <- 0 # same above
tau <- 0 # same above
(these 5 parameters are encoded in the file name of simulated data)

• makeCluster(No. of multiprocess)

outputs:

file_name_evatxt # evaluation table
file_name_w_opt # optimal w
file_name_w_select # optimal w before last projection step, used for calculating FNR FPR

fclassification_realdata.R

(for classification of real data with 5-fold, 10-fold and 20-fold) input parameters:

file_name # pre-processed real data
K_fold # No. of k-fold
B # No. of repetition

• makeCluster(No. of multiprocess)

outputs:

(same as above)

fclassification_realdata_leaveoneout.R

(for classification of real data with leave-one-out) input parameters:

file_name # pre-processed real data
K_fold <- 121 # leave-one-out
B # No. of repetition

• makeCluster(No. of multiprocess)

outputs:

(same as above)