Deconv not implemented for arrays

[orswan]

conv happily accepts 2d arrays:

> M = conv(ones(2,2),ones(2,2))
3×3 Matrix{Float64}:
 1.0  2.0  1.0
 2.0  4.0  2.0
 1.0  2.0  1.0

However, the analogous operation for deconv throws an error:

> deconv(M,ones(2,2))

MethodError: no method matching deconv(::Matrix{Float64}, ::Matrix{Float64})
...

Would it be possible to implement deconv for arrays?

[martinholters]

Can you specify what it should do, exactly? The vector-deconv does polynomial division:

DSP.jl/src/dspbase.jl

Lines 157 to 160 in 4ff6208

	deconv(b,a) -> c
	Construct vector `c` such that `b = conv(a,c) + r`.
	Equivalent to polynomial division.

That is quite brief, but means that c has length N=length(b)-length(a)+1 and the first N entries of r (if one computes it) are zero (up to rounding errors). I'm not sure how this generalizes to matrices. If a is separable, one could separate the deconvolution into a row-wise and a column-wise step, but for general a, I don't know.

That said, a deconvolution in the MMSE sense (i.e. find c such that sum(abs2, b - conv(a,c)) is minimized) might be useful and would also have a more obvious generalization to matrices, but it's not deconv.

[mbaz]

This link may provide some pointers: https://math.stackexchange.com/questions/190214/exact-deconvolution-of-two-matrices-using-numerical-techniques