A Jacobian matrix of a map f=(f1,…,fm):U→Rm (with U⊆Rn) at a point a∈U is the m×n matrix
Jf(a)=(∂xj∂fi(a))1≤i≤m,1≤j≤n,
provided these partial derivatives exist.
When f is differentiable at a, the Fréchet derivative Df(a) is a linear map, and Jf(a) is the matrix that represents Df(a) in the standard bases of Rn and Rm. For scalar-valued f, Jf(a) is closely related (up to transpose conventions) to the gradient.