fix code bug #8
fix code bug #8
Conversation
test/runtests.jl
Outdated
| W = Float64.([w1 w2]) | ||
| H = Float64.([h1; h2]) |
There was a problem hiding this comment.
| W = Float64.([w1 w2]) | |
| H = Float64.([h1; h2]) | |
| W = [w1 w2] | |
| H = [h1; h2] |
(If you need these, it's probably a code bug that should be fixed.)
There was a problem hiding this comment.
This is normalization part W[:, j] = w/normw, input W is int, but w/normw is float number, this will through a bug.
test/runtests.jl
Outdated
| H_v = [Hn[i, :] for i in axes(Hn, 1)] | ||
|
|
||
| Q1, Q2, _, _, _, _ =build_Qs(W_v, H_v, 1, 2) | ||
| @test Q1 == Q1' |
There was a problem hiding this comment.
| @test Q1 == Q1' | |
| @test issymmetric(Q1) |
test/runtests.jl
Outdated
|
|
||
| Q1, Q2, _, _, _, _ =build_Qs(W_v, H_v, 1, 2) | ||
| @test Q1 == Q1' | ||
| @test Q2 == Q2 |
test/runtests.jl
Outdated
| @test Q1 == Q1' | ||
| @test Q2 == Q2 | ||
| F = eigen(Q1, Q2) | ||
| Fvals, Fvecs = F.values::Vector{eltype(Q2)}, F.vectors::Matrix{eltype(Q2)} |
There was a problem hiding this comment.
Is this to solve a type-stability issue? That's generally not something to worry about for tests, though of course it doesn't hurt.
But you can solve inference problems in eigen with Symmetric (or Hermitian if you're dealing with complex-valued matrices):
julia> @inferred eigen(Q1, Q2)
ERROR: return type GeneralizedEigen{Float64, Float64, Matrix{Float64}, Vector{Float64}} does not match inferred return type Union{GeneralizedEigen{ComplexF64, ComplexF64, Matrix{ComplexF64}, Vector{ComplexF64}}, GeneralizedEigen{Float64, Float64, Matrix{Float64}, Vector{Float64}}}
Stacktrace:
[1] error(s::String)
@ Base ./error.jl:35
[2] top-level scope
@ REPL[34]:1
julia> @inferred eigen(Symmetric(Q1), Symmetric(Q2))
GeneralizedEigen{Float64, Float64, Matrix{Float64}, Vector{Float64}}
values:
2-element Vector{Float64}:
4938.9585139612855
84532.04148603871
vectors:
2×2 Matrix{Float64}:
0.968769 -0.768621
-1.19147 -0.331201
test/runtests.jl
Outdated
| λ_min = τ/2-sqrt(τ^2/4-δ) | ||
|
|
||
| @test abs(λ_max - maximum(F.values))<=1e-12 | ||
| @test abs(λ_min - minimum(F.values))<=1e-12 |
There was a problem hiding this comment.
Interestingly, I got this:
julia> @test abs(λ_min - minimum(F.values))<=1e-12
Test Failed at REPL[44]:1
Expression: abs(λ_min - minimum(F.values)) <= 1.0e-12
Evaluated: 4.547473508864641e-12 <= 1.0e-12Different library versions and CPUs can have slight differences in roundoff error. You might need to make this slightly less stringent. 1e-10, perhaps?
There was a problem hiding this comment.
Notice that GitHub actions tests also failed, for the same reason.
test/runtests.jl
Outdated
|
|
||
| w = [u[1], u[2]] | ||
| @test norm(u[1].*W_v[1].+u[2].*W_v[2]) ≈ 1 | ||
| @test norm(Q1*w - maximum(F.values)*Q2*w) < 1e-12 |
There was a problem hiding this comment.
Likewise here I got
julia> @test norm(Q1*w - maximum(F.values)*Q2*w) < 1e-12
Test Failed at REPL[51]:1
Expression: norm(Q1 * w - maximum(F.values) * Q2 * w) < 1.0e-12
Evaluated: 1.4551915228366852e-11 < 1.0e-12There was a problem hiding this comment.
I cannot got this fail on RIS.
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #8 +/- ##
=======================================
Coverage 81.63% 81.63%
=======================================
Files 1 1
Lines 98 98
=======================================
Hits 80 80
Misses 18 18 ☔ View full report in Codecov by Sentry. |
fix code bug on$\xi$ and $\mathbf{u}$