Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

18267 1x1 NaN matrix eigenvalue #18526

Merged
merged 8 commits into from
Sep 27, 2016
Merged

18267 1x1 NaN matrix eigenvalue #18526

Show file tree
Hide file tree
Changes from 7 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
80addf6
just finding out where the error is going wrong
mcognetta Sep 15, 2016
48f4f69
added a check for a 1x1 NaN filled matrix when computing the eigenvalue.
mcognetta Sep 15, 2016
9ec6b2f
moved the check for 1x1 NaN matrix symmetry to the issymmetric function.
mcognetta Sep 15, 2016
76c40e8
added tests for computing the eigenvalues of a matrix with NaNs in it.
mcognetta Sep 18, 2016
4363a0e
updated tests so that all of NaN, NaN16, and NaN32 are tested
mcognetta Sep 18, 2016
573386b
missed a paren
mcognetta Sep 18, 2016
6ee5bae
missed a paren
mcognetta Sep 18, 2016
3f18aa4
changed the indicies of the loop in issymmetric to include the diagonal.
mcognetta Sep 26, 2016
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion base/linalg/eigen.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,7 @@ isposdef(A::Union{Eigen,GeneralizedEigen}) = isreal(A.values) && all(A.values .>

function eigfact!{T<:BlasReal}(A::StridedMatrix{T}; permute::Bool=true, scale::Bool=true)
n = size(A, 2)
n==0 && return Eigen(zeros(T, 0), zeros(T, 0, 0))
n == 0 && return Eigen(zeros(T, 0), zeros(T, 0, 0))
issymmetric(A) && return eigfact!(Symmetric(A))
A, WR, WI, VL, VR, _ = LAPACK.geevx!(permute ? (scale ? 'B' : 'P') : (scale ? 'S' : 'N'), 'N', 'V', 'N', A)
all(WI .== 0.) && return Eigen(WR, VR)
Expand Down
6 changes: 6 additions & 0 deletions base/linalg/generic.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -704,6 +704,12 @@ function issymmetric(A::AbstractMatrix)
if indsm != indsn
return false
end
if indsm == Base.OneTo(1)
if isnan(A[1,1])
return false
end
return true
end
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think that this is sufficient although it will fix the eig issue. For larger matrices, we'd have to check that whole diagonal so instead the loop ranges below should be made one larger to include the diagonal.

for i = first(indsn):last(indsn)-1, j = (i+1):last(indsn)
if A[i,j] != transpose(A[j,i])
return false
Expand Down
10 changes: 10 additions & 0 deletions test/linalg/eigen.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -95,6 +95,16 @@ for eltya in (Float32, Float64, Complex64, Complex128, Int)
@test v == f[:vectors]
end
end

#test eigenvalue computations with NaNs
for eltya in (NaN16, NaN32, NaN)
@test_throws(ArgumentError, eig(fill(eltya, 1, 1)))
@test_throws(ArgumentError, eig(fill(eltya, 2, 2)))
test_matrix = rand(typeof(eltya),3,3)
test_matrix[2,2] = eltya
@test_throws(ArgumentError, eig(test_matrix))
end

# test a matrix larger than 140-by-140 for #14174
let aa = rand(200, 200)
for atype in ("Array", "SubArray")
Expand Down