vectors equality test failure - decimals and sign issue on zeros
1 view (last 30 days)
Show older comments
I shall test two vectors of same dimension for their equality: namely, x0 and xhat as defined below (they belong to different functions, xhat is return value of inner called function).
--- x0 ---
support = randsample(n,k);
x0 = zeros(n,1);
x0(support) = randn(k,1);
--- xhat ---
b = A*x0;
cvx_quiet(true);
cvx_begin
variable x(n)
minimize( norm(x,1) )
subject to
A*x == b
cvx_end
xhat = x;
To test equality I use:
if isequal(xhat, x0, 1e-6)
or
if isequal(xhat, x0)
condition which never verifies because x0 contains '0' or other (4 decimals, signed), while xhat contains '0.0000', '-0.0000', or other (4 decimals, signed). The algorithm works properly as values (real) in magnitude are the same in both vectors, and at the same indexes, therefore the condition for 'equality' is satisfied in facts, but I miss the way to code it in matlab.
How can I 'turn' xhat into x0, so to get isequal() function answering TRUE ?
Many thanks!
Answers (1)
Geoff Hayes
on 13 Aug 2016
VMat - see http://matlab.wikia.com/wiki/FAQ#Why_is_0.3_-_0.2_-_0.1_.28or_similar.29_not_equal_to_zero.3F for some details on comparing two floating-point numbers. When doing so, you shouldn't be using equality. Instead, compare the absolute difference to a tolerance. For example,
tol = 1.0e-6;
areEqual = all(abs(xhat-x0) < tol);
will return true if the pairwise absolute difference between each element of the two arrays is less than the tolerance.
1 Comment
Guillaume
on 14 Aug 2016
@Vmat,
I guess that this tolerance comparison is what you were trying to do with your isequal(xhat, x0, 1e-6) but that is not how isequal works at all. There is no tolerance argument for isequal. As per the doc, the above simply checks that both xhat and x0 are equal to 1e-6.
See Also
Categories
Find more on Hypothesis Tests in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!