Kalkulator Backpropagation

Hitung forward pass, loss, dan backward pass (gradient) MLP step by step.
δ^L = (ŷ−y)⊙σ′(z^L)  |  δ^l = (W^l+1T·δ^l+1)⊙σ′(z^l)  |  ΔW^l = δ^l·(a^l-1)^T