----- "Kosumo" <mas.kos...@gmail.com> a écrit :
> Dear all,
>
> I got confused with the "equals" method applied to OpenMapRealMatrix
> instance.
> I have this following code:
>
> RealMatrix raw = vectorGenerator.getMatrix();
> RealMatrix test =
> TestHelper.parseMatrixFromFile("resources/raw_matrix.txt");
>
> both the raw and test matrices are the instance of OpenMapRealMatrix.
> The
> test matrix is generated by reading the matrix value from a text
> file.
> Both of them are intended to have the same value as follow:
>
> 0.0000 0.0000 1.0000 1.0000 0.0000 0.0000 0.0000 0.0000
> 0.0000
> 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000
> 1.0000
> 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000
> 0.0000
> 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000
> 0.0000
> 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000
> 1.0000
> 1.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000
> 0.0000
> 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
> 1.0000
> 1.0000 0.0000 1.0000 1.0000 0.0000 0.0000 0.0000 0.0000
> 0.0000
> 1.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000
> 0.0000
> 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000
> 1.0000
> 0.0000 0.0000 0.0000 0.0000 0.0000 2.0000 0.0000 0.0000
> 1.0000
> 1.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 1.0000
> 0.0000
> 0.0000 0.0000 0.0000 1.0000 1.0000 0.0000 0.0000 0.0000
> 0.0000
>
> I want to test if they really have the same value. I tried to use
> "equals"
> method, Eclipse indicates that 'equals' is method from Object class.
It should not. There is an equals method in AbstractRealMatrix overriding
the method from Object. Of course, this method still takes an Object as
a parameter, as specified in the top level method.
This implementation of the equals method compares all elements with a loop
using the "!=" operator applied tou primitive double values.
> I suppose it will generate false since both matrices are not the same
> object. However when I execute this statement:
>
> raw.equals(test)
>
> it generate "true".
Since all numbers here seem to be exact integer values and since integer values
lower than 10^15 can be represented exactly in IEEE754 double, the true result
is
expected.
>
> In other case where the value of both matrix:
> 0.0000 0.0000 0.1306 0.1633 0.0000 0.0000 0.0000 0.0000
> 0.0000
> 0.0000 0.0000 0.0000 0.0000 0.0000 0.1306 0.0000 0.0000
> 0.1306
> 0.0000 0.2177 0.0000 0.0000 0.0000 0.0000 0.0000 0.2177
> 0.0000
> 0.0000 0.2177 0.0000 0.0000 0.0000 0.0000 0.1633 0.0000
> 0.0000
> 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1633 0.0000
> 0.1306
> 0.1306 0.0000 0.0000 0.0000 0.0000 0.1306 0.0000 0.0000
> 0.0000
> 0.0000 0.0171 0.0102 0.0128 0.0256 0.0102 0.0128 0.0171
> 0.0102
> 0.0954 0.0000 0.0954 0.1193 0.0000 0.0000 0.0000 0.0000
> 0.0000
> 0.1306 0.0000 0.1306 0.0000 0.0000 0.0000 0.0000 0.0000
> 0.0000
> 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1633 0.0000
> 0.1306
> 0.0000 0.0000 0.0000 0.0000 0.0000 0.2613 0.0000 0.0000
> 0.1306
> 0.0954 0.0000 0.0954 0.0000 0.0000 0.0000 0.0000 0.1590
> 0.0000
> 0.0000 0.0000 0.0000 0.1633 0.3266 0.0000 0.0000 0.0000
> 0.0000
>
> the above statement give me "false" value.
I think the "false" result is also normal. The numbers generated from
your vectorGenerator are probably different from the ones read in the
file, due to limited precision. If for example the numbers are written
the file as in this message (i.e. with four decimal places), then the
difference may go up to 5.0e-5. Increasing the accuracy in the file may
help but really opens a can of worms because some numbers which can be
represented perfectly in decimal (like 0.1) cannot be stored perfectly
in memory (in the IEEE754 format), so you would have to make sure you
use enough decimal places (16 is a good start) in the text representation
and hope the conversion operations that occurs at write and read time are
accurate enough ...
When double values are used in a computer, equality shoud be used
with caution.
If you intend to make sure two matrices are "close enough", a better way
is to use a check like:
if (m1.subtract(m2).getNorm() < threshold) {
// the matrices m1 and m2 are close
}
The value of the threshold to use depend on the values stored in your matrices.
If for example all matrix elements are between 0.0 and 1.0, you could use
something like 100 * MathUtils.EPSILON * m1.getColumndimension().
Luc
>
> Can anyone explain about this? I tried to look at the API. I found
> that OpenMapRealMatrix is subclass of AbstractRealMatrix.
> AbstractRealMatrix has an 'equals' method, but I can't access it
> since
> Eclipse still indicate that the 'equals' method is the one from
> Object
> class.
> How can I use 'equals' method from AbstractRealMatrix? Is it possible
> to use
> it for comparing value from two matrices?
> Thanks in advance,
>
> Kosumo
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscr...@commons.apache.org
For additional commands, e-mail: user-h...@commons.apache.org
