grad requires a scalar-valued function

[fatihtalu]

Does not it have to be the solution [ [2 0] [0 2] ] ?

[CarloLucibello]

as the error says, autograd requires scalar valued functions, and df is not. To obtain the hessian for your example you can use

julia> g=grad(f)
(::gradfun) (generic function with 1 method)

julia> ddf(w)= [grad(x->g(x)[1])(w) grad(x->g(x)[2])(w)]
ddf (generic function with 1 method)

julia> ddf(w)
2×2 Array{Int64,2}:
 2  0
 0  2

Also see #54

[fatihtalu]

Perfect! exactly as I want.
Thanks CarloLucibello.