Houjun Liu

LU-Factorization

Elimination Matricies can be used to derive a LU factorization:

First, this gives an upper triangular matrix

\begin{equation} U = M_{n-1, n-1} \dots M_{22} M_{11} A \end{equation}

We can also create the inverses of each of these:

\begin{equation} A = L_{11} L_{22} \dots L_{n-1,n-1} \cdot M_{n-1,n-1} \dots M_{22} \cdot M_{11} \cdot A \end{equation}

The first half \(L_{j}\) composes a lower triangular matrix; the second half \(M_{j}\) which composes a upper triangular matrix.


Then, this helps solve:

\begin{equation} Ac = b \end{equation}

Because we can factor first to:

\begin{equation} \qty(LU) c = b \end{equation}

Then, this makes it really easy to solve, because we can.

\begin{equation} \hat{c} = Uc \end{equation}

then, we can solve \(L\hat{c} = b\) using forward substitution; then we can solve \(Uc = \hat{c}\) using back substitution.

Notice that for every new \(b\), we don’t need to perform Gaussian elimination.