treelet (version 0.3-0)

Score_JTree: The function computes the scores for the best K basis, for each level of the tree

Description

calls three functions:Build_JTree,JTree_Scores,Get_C.The returned variables are computed on a maximum level tree

Usage

Score_JTree(X, K, TrainIdx, TestIdx)

Arguments

X

data matrix (each row is an observation)

K

number of vectors in final representation

TrainIdx

indices for train data (for building the tree)

TestIdx

indices for test data (for computing a score)

Value

• levels
  at each level
• scores
  what the score is at the corresponding level
• Zpos
  Zpos[dim-1,2] positions of the two clusters merged; number between 1 and d (the length of the comp vector)
• PCidx
  ID of principal components; [1 2] or [2 1]
• T
  Element T{k} in the list array contains a 2-by-2 rot matrix
• all_nodes
  node labels

Details

Remark: Compare this function with Run_JTree

Examples

N=1000;
dim_x=10;
sigma_x=1;

set.seed(100)
s1_2=290;
s2_2=300;
c1=-0.3;
c2=0.925;
u1=sqrt(s1_2)*matrix(rnorm(N),ncol=1);
u2=sqrt(s2_2)*matrix(rnorm(N),ncol=1);
u3=c1*u1+c2*u2;
s3_2=c1^2*s1_2+c2^2*s2_2;
Xraw=cbind(u1,u1,u1,u1,u2,u2,u2,u2,u3,u3);
#Xraw=cbind(rep(u1,4),rep(u2,4),u3,u3);
X=Xraw+sigma_x*matrix(rnorm(N*dim_x),ncol=dim_x);

###Just create a matrix X, whatever matrix you want


TrainIdx=sample(N,floor(3/4*N));#use 3/4 of the data for training data, randomize
TotalIdx=seq(1,N,by=1);
TestIdx=TotalIdx[-c(TrainIdx)]; # the others are testing data

K=5;# for example,,,,,,


out1=Score_JTree(X,K,TrainIdx,TestIdx);