我正在尝试在Julia(0.6.2)中实现max-plus半环代数,继本文之后:
http://www.mit.edu/~kepner/pubs/JuliaSemiring_HPEC2013_Paper.pdf
直接从上述文件中得出的max-plus数字的类型定义和所有相关运算符的定义如下:
# define max-plus number type
immutable MPNumber{T} <: Number
val::T
end
+(a::MPNumber, b::MPNumber) = MPNumber(max(a.val, b.val))
*(a::MPNumber, b::MPNumber) = MPNumber(a.val + b.val)
show(io::IO, k::MPNumber) = show(io, k.val)
zero{T}(::MPNumber{T}) = MPNumber(typemin(T))
one{T}(::MPNumber{T}) = MPNumber(zero(T))
promote_rule{T<:Number}(::Type{MPNumber}, ::Type{T}) = MPNumber
mparray(A::Array) = map(MPNumber, A)
array{T}(A::Array{MPNumber{T}}) = map(x->x.val, A)
max-plus加法和乘法的基本测试工作很好,例如
MPNumber(1) + MPNumber(1)
正如预期的那样,给出MPNumber(1)
。如文中所示,Max-plus矩阵取幂也像魅力一样:
A = mparray(Array([[1, 2] [3, 4]]))
A*A
给出
2×2 Array{MPNumber{Int64},2}:
MPNumber{Int64}(5) MPNumber{Int64}(7)
MPNumber{Int64}(6) MPNumber{Int64}(8)
然而,当我尝试将max-plus向量与max-plus矩阵相乘时,
A = mparray(Array([[1, 2] [3, 4]]))
x = mparray([1, 2])
A*x
我收到以下错误:
MethodError: Cannot `convert` an object of type Int64 to an object of type MPNumber{Int64}
This may have arisen from a call to the constructor MPNumber{Int64}(...) since type constructors fall back to convert methods.
Stacktrace:
[1] generic_matvecmul!(::Array{MPNumber{Int64},1}, ::Char, ::Array{MPNumber{Int64},2}, ::Array{MPNumber{Int64},1}) at ./linalg/matmul.jl:434
[2] Ac_mul_B(::Array{MPNumber{Int64},1}, ::Array{MPNumber{Int64},2}) at ./linalg/rowvector.jl:227
我还是朱莉娅的新手,所以我很难弄清楚这里出了什么问题以及如何修复它。任何帮助将不胜感激。
答案 0 :(得分:7)
这是在Julia 0.6.2下测试的。 使用以下代码:
public interface IMessageManager
{
void AddSuccess(string message, int? dismissAfter = null);
void AddError(string message, int? dismissAfter = null);
void AddInfo(string message, int? dismissAfter = null);
void AddWarning(string message, int? dismissAfter = null);
Dictionary<string, IEnumerable<FlashMessage>> GetAlerts();
}
(我建议您从这部分开始,因为转换和促销更加棘手,不需要用于您的目的)。
注意到关键部分是我通过在struct MPNumber{T} <: Number
val::T
end
Base.:+(a::MPNumber, b::MPNumber) = MPNumber(max(a.val, b.val))
Base.:*(a::MPNumber, b::MPNumber) = MPNumber(a.val + b.val)
Base.show(io::IO, k::MPNumber) = show(io, k.val)
Base.zero(::MPNumber{T}) where T = MPNumber(typemin(T))
Base.one(::MPNumber{T}) where T = MPNumber(zero(T))
Base.zero(::Type{MPNumber{T}}) where T = MPNumber(typemin(T))
Base.one(::Type{MPNumber{T}}) where T = MPNumber(zero(T))
mparray(A::Array) = map(MPNumber, A)
前面加Base
来扩展函数。对于Base.
和+
,您需要在函数名称之前另外使用*
。
现在你可以运行你想要的一切:
:
希望。
编辑:我在您的代码中对julia> MPNumber(1) + MPNumber(1)
1
julia> A = mparray(Array([[1, 2] [3, 4]]))
2×2 Array{MPNumber{Int64},2}:
1 3
2 4
julia> A*A
2×2 Array{MPNumber{Int64},2}:
5 7
6 8
julia> A = mparray(Array([[1, 2] [3, 4]]))
2×2 Array{MPNumber{Int64},2}:
1 3
2 4
julia> x = mparray([1, 2])
2-element Array{MPNumber{Int64},1}:
1
2
julia> A*x
2-element Array{MPNumber{Int64},1}:
5
6
和zero
定义做了更完整的规定,并遵循标准要求。