Fig.1

Concept

Softmax Jacobian

Softmax Jacobian is the matrix of partial derivatives of the softmax output with respect to its input logits. If p = softmax(z) is a length-K probability vector, the Jacobian is J = diag(p) - p p^T, so entry J_ij = p_i(δ_ij - p_j).

The rest of “Softmax Jacobian” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.

Log in to unlock

← Back to the library

Softmax Jacobian, explained · Fig. 1