Solving Least-Squares Problems with Gradient Descent: the Least Mean-Square Algorithm

Which of the following are disadvantages of directly computing the solution to the MMSE filtering or least squares problem using ${\bf w} = {\bf R}^{-1}{\bf p}$? Assume the dimension of ${\bf w}$ is $N$ and that $L$ data samples are available. Select all that apply.

The LMS algorithm updates the current estimate of the weights with a step in a direction given by the negative gradient of the instantaneous squared error ${\bf e}^2[n]$.

The instantaneous gradient used in the LMS algorithm is the same as the gradient of the MSE.

Define the error as $e[n] = d[n]-{\bf x}^T[n]{\bf w}$. Which of the following is the gradient of $e^2[n]$?