dsearch

Name

dsearch — binary search (aka dichotomous search in french)

Calling Sequence

[ind, occ, info] = dsearch(X, val [, ch ])

Parameters

X: a real vector or matrix
val: a real (row or column) vector with n components in strictly increasing order val(1) < val(2) < ... < val(n)
ch: (optional) a character "c" or "d" (default value "c")
ind: a real vector or matrix with the same dimensions than X
occ: a real vector with the same format than val (but with n-1 components in the case ch="c")
info: integer

Description

This function is useful to search in an ordered table and/or to count the number of components of a vector falling in some classes (a class being an interval or a value).

By default or when ch="c", this is the interval case, that is, for each X(i) search in which of the n-1 intervals it falls, the intervals being defined by:

 
I1 = [val(1), val(2)]
Ik = (val(k), val(k+1)] for 1 < k <= n-1 ;

and:

ind(i): is the interval number of X(i) (0 if X(i) is not in [val(1),val(n)])
occ(k): is the number of components of X which are in Ik
info: is the number of components of X which are not in [val(1),val(n)]

When ch="d" case, this is the discrete case, that is, for each X(i) search if it is equal to one val(k) and:

ind(i): is equal to the index of the component of val which matches X(i) (ind(i) = k if X(i)=val(k)) or 0 if X(i) is not in val.
occ(k): is the number of components of X equal to val(k)
info: is the number of components of X which are not in the set {val(1),...,val(n)}

Examples

 
// example #1 (elementary stat for U(0,1))
m = 50000 ; n = 10;
X = grand(m,1,"def");
val = linspace(0,1,n+1)';
[ind, occ] = dsearch(X, val);
clf() ; plot2d2(val, [occ/m;0])  // no normalisation : y must be near 1/n


// example #2 (elementary stat for B(N,p))
N = 8 ; p = 0.5; m = 50000;
X = grand(m,1,"bin",N,p); val = (0:N)';
[ind, occ] = dsearch(X, val, "d");
Pexp = occ/m; Pexa = binomial(p,N); 
clf() ; hm = 1.1*max(max(Pexa),max(Pexp));
plot2d3([val val+0.1], [Pexa' Pexp],[1 2],"111",  ...
        "Pexact@Pexp", [-1 0 N+1 hm],[0 N+2 0 6])
xtitle(  "binomial distribution B("+string(N)+","+string(p)+") :" ...
        +" exact probability versus experimental ones")


// example #3 (piecewise Hermite polynomial)
x = [0 ; 0.2 ; 0.35 ; 0.5 ; 0.65 ; 0.8 ;  1];
y = [0 ; 0.1 ;-0.1  ; 0   ; 0.4  ;-0.1 ;  0];
d = [1 ; 0   ; 0    ; 1   ; 0    ; 0   ; -1];
X = linspace(0, 1, 200)';
ind = dsearch(X, x);

// define Hermite base functions
deff("y=Ll(t,k,x)","y=(t-x(k+1))./(x(k)-x(k+1))")   // Lagrange left on Ik
deff("y=Lr(t,k,x)","y=(t-x(k))./(x(k+1)-x(k))")     // Lagrange right on Ik
deff("y=Hl(t,k,x)","y=(1-2*(t-x(k))./(x(k)-x(k+1))).*Ll(t,k,x).^2")
deff("y=Hr(t,k,x)","y=(1-2*(t-x(k+1))./(x(k+1)-x(k))).*Lr(t,k,x).^2")
deff("y=Kl(t,k,x)","y=(t-x(k)).*Ll(t,k,x).^2")
deff("y=Kr(t,k,x)","y=(t-x(k+1)).*Lr(t,k,x).^2")

// plot the curve
Y = y(ind).*Hl(X,ind) + y(ind+1).*Hr(X,ind) + d(ind).*Kl(X,ind) + d(ind+1).*Kr(X,ind);
clf(); plot2d(X,Y,2) ; plot2d(x,y,-9,"000") 
xtitle("an Hermite piecewise polynomial")
// NOTE : you can verify by adding these ones : 
// YY = interp(X,x,y,d); plot2d(X,YY,3,"000")

Authors

B.P.