Why back propagation goes backward

Automated excerpt

So at every node vv, if we knew ∂f/∂v\partial f / \partial v, we could compute ∂f/∂θ1\partial f / \partial \theta_1. We want a forward propagating algorithm that can compute the partial derivative ∂f/∂θi\partial f / \partial \theta_i for an arbitrary weight θi\theta_i. In our setup, for every downstream node wjw_j that depends on a node vv, it is impossible to compute ∂wj/∂v\partial w_j / \partial v at node vv.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.