Geek asks and answers

Rounding numbers with a custom threshold

I want to be able to "round" a number up if it has passed a threshold (not 0.5) and otherwise round down.

here is some shitty code that I came up with. Is there a built-in function in matlab for this, or a more elegant solution (possibly vectorized)?

function [ rounded_numbers ] = custom_round( input_numbers, threshold )
%CUSTOM_ROUND rounds between 0 and 1 with threshold threshold

  [input_rows, input_cols] = size(input_numbers);
  rounded_numbers = zeros(input_rows, input_cols);

  for i = 1:length(input_numbers)
    if input_numbers(i) > threshold
      rounded_numbers(i) = 1;
    else
      rounded_numbers(i) = 0;
    end
  end
end

thank

+4
source share
3 answers

Here's a solution where we round off from zero if the number has passed the threshold

in = [0.2,-3.3,4.1];
th = 0.2;

%# get the fractional part of the number
frac = mod(in,1); %# positive for negative in

%# find the sign so we know whether to round
%# to plus or minus inf
sig = sign(in);

%# identify which way to round
upIdx = frac>th; %# at threshold, we round down

%# round towards inf if up
out = abs(in);
out(upIdx) = ceil(out(upIdx));
out(~upIdx) = floor(out(~upIdx));
%# re-set the sign
out= out.*sig
out =
 0    -4     4

Note. If the numbers are only between 0 and 1, this is even simpler:

%# this does exactly what your code does
out = double(in>th);
+2
source

just use

round(x - treshold + 0.5)

test case:

>> x = -10:0.3:10
ans =
    -2   -1.7  -1.4  -1.1  -0.8  -0.5  -0.2  0.1    0.4   0.7    1    1.3   1.6   1.9


>> treshold = 0.8; % round everything up for which holds mod(x,1) >= treshold
>> y = round(x-treshold+0.5)

ans =
    -2    -2    -2    -1    -1    -1    -1     0     0     0     1     1     1     2

negative numbers are also rounded correctly, except for the border: -0.8 is rounded to -1 instead of 0, but the same behavior as with normal rounding: round (-0.5) returns -1

+9

, 0 1. [0,1).

.

function [result] = custom_round( num, threshold )

if ( threshold < 0 ) || ( threshold >= 1 )
  error( 'threshold input must be in the range [0,1)' );
end

fractional = num - floor( num );
idx1 = fractional > threshold;
idx2 = fractional <= threshold;
difference = 1 - fractional;
result = num + ( difference .* idx1 ) - ( fractional .* idx2 );

end


>> custom_round( [0.25 0.5 0.75 1], 0.3 )
ans =
     0     1     1     1

>> custom_round( [0.25 0.5 0.75 1], 0.8 )
ans =
     0     0     0     1

>> custom_round( [10.25 10.5 10.75 11], 0.8 )
ans =
    10    10    10    11
+1
source

More articles:


All Articles