Log-derivative formula
The \(\log\)-derivative formula states that for a function composition involving logarithms:
If we have a function \(f(\mathbf{x})\) (\(\mathbf{x} \in \mathbb{R}^n\)) and we’re looking for the gradient of \(\log(f(\mathbf{x}))\), we can apply the \(\log\)-derivative formula in the multivariable setting:
This is the vector calculus equivalent of the scalar \(\log\)-derivative formula. The gradient \(\nabla_{\mathbf{x}}\) gives us a vector of partial derivatives with respect to each component of \(\mathbf{x}\).
For each component \(x_i\) of the vector \(\mathbf{x}\), we have:
So the full gradient is:
This generalizes the scalar \(\log\)-derivative to the vector setting, maintaining the same core principle: the gradient of the logarithm of a function equals the gradient of the function divided by the function itself.