MIN Minimum Function

Section: Elementary Functions


Computes the minimum of an array along a given dimension, or alternately, computes two arrays (entry-wise) and keeps the smaller value for each array. As a result, the min function has a number of syntaxes. The first one computes the minimum of an array along a given dimension. The first general syntax for its use is either
   [y,n] = min(x,[],d)

where x is a multidimensional array of numerical type, in which case the output y is the minimum of x along dimension d. The second argument n is the index that results in the minimum. In the event that multiple minima are present with the same value, the index of the first minimum is used. The second general syntax for the use of the min function is

   [y,n] = min(x)

In this case, the minimum is taken along the first non-singleton dimension of x. For complex data types, the minimum is based on the magnitude of the numbers. NaNs are ignored in the calculations. The third general syntax for the use of the min function is as a comparison function for pairs of arrays. Here, the general syntax is

   y = min(x,z)

where x and z are either both numerical arrays of the same dimensions, or one of the two is a scalar. In the first case, the output is the same size as both arrays, and is defined elementwise by the smaller of the two arrays. In the second case, the output is defined elementwise by the smaller of the array entries and the scalar.

Function Internals

In the general version of the min function which is applied to a single array (using the min(x,[],d) or min(x) syntaxes), The output is computed via

and the output array n of indices is calculated via

In the two-array version (min(x,z)), the single output is computed as


The following piece of code demonstrates various uses of the minimum function. We start with the one-array version.
--> A = [5,1,3;3,2,1;0,3,1]

A = 
 5 1 3 
 3 2 1 
 0 3 1 

We first take the minimum along the columns, resulting in a row vector.

--> min(A)

ans = 
 0 1 1 

Next, we take the minimum along the rows, resulting in a column vector.

--> min(A,[],2)

ans = 

When the dimension argument is not supplied, min acts along the first non-singular dimension. For a row vector, this is the column direction:

--> min([5,3,2,9])

ans = 

For the two-argument version, we can compute the smaller of two arrays, as in this example:

--> a = int8(100*randn(4))

a = 
   27   61 -128  -22 
  -54  -74  -27  -61 
   66 -128  127   34 
  -13 -128   42   86 

--> b = int8(100*randn(4))

b = 
   56   33  -15   17 
  127  127   29  -97 
   99  -45  117 -128 
   56 -114  127  104 

--> min(a,b)

ans = 
   27   33 -128  -22 
  -54  -74  -27  -97 
   66 -128  117 -128 
  -13 -128   42   86 

Or alternately, we can compare an array with a scalar

--> a = randn(2)

a = 
   -0.2804    0.3739 
   -0.8883   -0.4260 

--> min(a,0)

ans = 
   -0.2804         0 
   -0.8883   -0.4260