Receiving the Error: "MethodError: no method matching eps(::Int64)" When Using NonlinearProblem Solver

New to Julia
, , , ,
3

I agree with lupemba. There is almost always surprises with a Vector{Real}. I don’t see exactly what is the problem in your code, but note that a Vector{Real} can contain both Float64 and Int, as well as other things:

julia> v = Real[2.34, 0, pi]
3-element Vector{Real}:
 2.34
 0
 π = 3.1415926535897...

julia> typeof.(v)
3-element Vector{DataType}:
 Float64
 Int64
 Irrational{:π}

I guess somehow an Int has sneaked into one of your vectors.

I realize that you use Real to accommodate the Dual type of ForwardDiff, but it is also generally a bad idea to have structs with abstract data types like AbstractVector/AbstractArray. The reason is mainly that it is slow.

If you need temporary storage inside your objective function which is compatible with ForwardDiff, there are tools for it in Home · PreallocationTools.jl.

1 Like