find function and vectors gives Empty matrix: 1-by-0

close all
clear variables
clc
w = [-65:0.1:10];%Error membership x axis range
Negbig_error = trapmf(w,[-65 -65 -62 -10]);
ant1 = Negbig_error(find(w == -25.8))

w ranges from -65 to 10 in my code with step of 0.1 but matlab cannot find y axis intercept when w = -25.8

returns error code: Empty matrix: 1-by-0

Answer 1

[Michael Haderlein]

That's a floating point arithmetics problem. 0.1 cannot be represented exactly. Creating the array your way, you get something like

>> -65+0.1*392

ans =

    -25.8000

but that's not exactly -25.8:

>> (-65+0.1*392)+25.8

ans =

     3.5527e-15

If I remember correctly, Matlab actually creates such arrays a bit more intelligently than sketched here, but the general issue remains. If you want to search for w==-25.8, you have to search for something close to this value, such as

>> w=-65:.1:10;
>> Negbig_error = trapmf(w,[-65 -65 -62 -10]);
>> w(abs(w+25.8)<.01)

ans =

    -25.8000

>> Negbig_error(abs(w+25.8)<.01)

ans =

      0.3038

find() is not necessary here, so I skipped it.

[Guillaume]

A fairly good write-up on floating point comparison here

[James Tursa]

FYI, the exact decimal conversions of the IEEE double floating point bit patterns for the numbers involved:

>> w = [-65:0.1:10];
>> num2strexact(w(393))

ans =

-2.580000000000000426325641456060111522674560546875e1

>> num2strexact(-25.8)

ans =

-2.5800000000000000710542735760100185871124267578125e1

You can find num2strexact here:

http://www.mathworks.com/matlabcentral/fileexchange/22239-num2strexact--exact-version-of-num2str-

[Amit Aryal]

Negbig_error(abs(w+25.8)<.01) (line 3)

Michael Thank you,

Please can you answer details of line 3 of your code. I don't understand why you don't use find function. 0 Comments Comment on this Answer Amit Aryal

[Michael Haderlein]

It's just logical indexing. If you have an array, say [10 20 30 40 50], you can pick elements either by their position number (myarray(2) will return 20 and myarray([2 4]) will return [20 40]) or you can pick them with logicals: myarray([false true false true false]) will also return [20 40]. That's a convenient way to select elements based on their value, e.g. myarray<30 will be [true true false false false]. Use this as logical index and you'll get the values of all elements fulfilling the condition.

[Stephen23]

@Amit Aryal: logical indexing is much faster than using find, and can be calcualted directly in many situations, e.g. to remove all negative values from a vector:

A = [1,-3,2,4,-9,];
A(A<0) = []

There are several different indexing methods in MATLAB, and they are each useful in different situations. You can learn about them here:

http://www.mathworks.com/help/matlab/math/matrix-indexing.html