[R] Help in Recursive Function for steps reconstruction in Clusters Analysis - HCA: Single Link

[Cleber Borges]

For now, I've been able to think of a temporary solution... until I can 
think of a better teaching solution:

The same function f():

#####
f <- function(xmd) {
     rs <- NULL
     xmd1 <- 999999999
     xmd2 <- 999999999

     if (xmd[1] > 0) {
         xmd1 <- f(r$merge[ xmd[1], ])
     } else {
         xmd1 <- xmd[1]
     }

     if (xmd[2] > 0) {
         xmd2 <- f(r$merge[ xmd[2], ])
     } else {
         xmd2 <- xmd[2]
     }

     return( rbind( rs, rbind( xmd1, xmd2 ) ) )
}

####

Didactic example to illustrate the chain behavior of the "Single Link" 
Algorithm


n <- 100

d <- matrix( runif( n * 2 ), n, 2 )

r <- hclust( dist( d ), met='single' )

plot( d, cex=1, col=1 )

for( i in 1:( nrow( d ) - 1 ) ){

     p <- as.numeric( abs( f( r$merge[ i, ] ) ) )

     points( d[ p,], col=i, cex=2, pch=19 )

     cat(' Step: ', i, '\n' )

     flush.console()

     Sys.sleep( .2 )

}






Em 26/10/2024 16:14, Cleber Borges escreveu:
> Hello everybody,
>
> I'm trying to build a function to illustrate, in 2D, the sequence of 
> the "Single Link" algorithm (the purpose is merely didactic).
>
> The idea is to have the scatter of points on a graph.
> Iteratively, build the segments (with the "segments()" function, for 
> each step).
>
> I simulated a data set "d", and created an object "r" using the command:
>
> r <- hclust( dist( d ), met='single' )
>
> My problem:
>
> I created a recursive function, "f()", to find the shortest distances. 
> It is defined below.
> But the return has only one column, while I thought there would be two 
> columns.
>
>
> What I get:
> ###############
> # Testing f()
> f( c(2, 3) )   #### is the fourth step of hclust
>
>      [,1]
> xmd1   -6
> xmd2  -10
> xmd1   -3
> xmd1   -4
> xmd2   -5
>
> What I expected:
> ##################
>
> # xmd1   xmd2
> # -6     -10
> # -3      -5
> # -4      -5
>
>
> If anyone can help me with this recursive function, I would appreciate 
> it in advance.
> I've never used recursion before, so I don't have a good grasp of how 
> it works.
>
> Thank you very much in advance,
>
> Cleber.
>
>
>
> All details are below:
> ########################
>
>
> d <- scan()
>  0.986  0.900
> -1.331  1.503
> -0.220 -0.752
>  0.102 -0.071
> -0.171  0.018
> -0.782  0.490
>  1.154 -0.074
> -0.768 -1.529
>  1.761 -1.396
> -0.730  0.910
>
> d <- matrix( d, 10, byrow=T )
>
> r <- hclust( dist( d ), met='single' )
> r$merge
> #      [,1] [,2]
> # [1,]   -4   -5
> # [2,]   -6  -10
> # [3,]   -3    1
> # [4,]    2    3
> # [5,]   -2    4
> # [6,]   -8    5
> # [7,]   -1   -7
> # [8,]    6    7
> # [9,]   -9    8
>
>
> f <- function(xmd) {
>
>     rs <- NULL
>     xmd1 <- 999999999
>     xmd2 <- 999999999
>
>     if (xmd[1] > 0) {
>         xmd1 <- f(r$merge[ xmd[1], ])
>     } else {
>         xmd1 <- xmd[1]
>     }
>
>     if (xmd[2] > 0) {
>         xmd2 <- f(r$merge[ xmd[2], ])
>     } else {
>         xmd2 <- xmd[2]
>     }
>
>     return( rbind( rs, rbind( xmd1, xmd2 ) ) )
> }
>
>
>
> # Testing f()
> f( c(2, 3) )
>
> # result of f()
> #      [,1]
> # xmd1   -6
> # xmd2  -10
> # xmd1   -3
> # xmd1   -4
> # xmd2   -5
>
> # My expectative:
> # xmd1   xmd2
> # -6     -10
> # -3      -5
> # -4      -5
>
>
> # Testing f()
>  f( c( 6, 7) )
>
> # result of f()
> #      [,1]
> # xmd1   -8
> # xmd1   -2
> # xmd1   -6
> # xmd2  -10
> # xmd1   -3
> # xmd1   -4
> # xmd2   -5
> # xmd1   -1
> # xmd2   -7
>
>
> # My expectative:
> # xmd1   xmd2
> # -8     -10
> # -2     -10
> # -6     -10
> # -3      -5
> # -4      -5
> # -1      -7
>