Sum of Flux.params

Specific Domains Machine Learning
1

I tried these two functions to get the l2 regularization term for a NN model:

function ℓ₂(a)
    s = zero(Float32)
    for i in a
        s += sum(i.^2)
    end
    s
end

function l₂(a)
    s = zero(Float32)
    Lₐ = length(a)
    for i in 1:Lₐ
        @inbounds Lᵢ = length(a[i])
        for j in 1:Lᵢ
            @inbounds s += a[i][j]^2    
        end
    end
    s
end

But I get different result, for the first function I get 469.8826f0 and for the second I get 469.87442f0.
Why this happens?
Moreover the second function is much slower than the first one. Why?
What is the best way to compute the sum of params of a model in julia?

Sorry for my english.

2

This can be done with a one-liner :slight_smile: . Regularisation · Flux has some good examples.