Initialized solution is set to dimension of constant vector rather than the number of columns
Closed this issue · 1 comments
vp314 commented
Minimum Working Example
using RLinearAlgebra
A = RLinearAlgebra.generate_matrix(100, 20, 20, 10.0)
x_oracle = randn(20)
b = A * x_oracle
sol = LinearSolver(TypeRPM(SamplerKaczmarzCYC()))
x = solve(sol, A, b)
Output
DimensionMismatch("second dimension of A, 20, does not match length of x, 100")
Source of the Issue
In src/solvers.jl
(Line 61 to 64), the initial value of x
is set to a zero vector that is the dimension of b
, the constant vector:
#Stopping threshold
x_init = copy(b)
x_init .= 0.0
residual = norm(A*x_init - b)
The issue is that the dimension of b
(e.g., 100) and the number of columns of A
(e.g., 20).
vp314 commented
I was not working on the most recent version of the master branch.