# Cheating the inverse…well sort of.

If a multiply two real numbers I get the product. If I divide the product by one of those real numbers I get the other.

By my reasoning since, ${ |A|^{-1} * |b|= |Solution|}$ , if I know the solution vector ( Using RRef) and divide by ${|b| }$ I should get ${|A|^{-1}}$.

But my online adventures tell me you can't divide vectors.

So what gives? Can I deduce the inverse of a matrix If I know ${|b|}$ and the solution vector or can't I?

• No, unless the matrices are 1-by-1, there are always many matrices $A$ such that $Ax = b$ for given vectors $x$ and $b$. (The division $b/x$ is not well-defined.) – Shaun Ault Nov 4 '13 at 13:19
• You cannot. You can if (for an $n\times n$) you know this for $n$ linearly independent vectors. – André Nicolas Nov 4 '13 at 13:20
• More concretely, consider the vector $v = ( 1; 1 )^T$. Then $( 1; 1) \cdot v = (2; 0) \cdot v = 2$. – Johannes Kloos Nov 4 '13 at 13:21
• I already have |A|.(presumably) as I completed RRef to get the solution vector. What I want is the INVERSE of |A|. Does it matter? – Chris Nov 4 '13 at 13:26
• Thanks for the input. Guess I'll have to go into row operators for Rref for my next adventure. – Chris Nov 4 '13 at 13:37

It works like that if and only if $A$ is invertible. If $A$ is invertible and we're trying to solve $A\mathbf{x}=\mathbf{b}$, then the solution is unique and is $A^{-1}\mathbf{b}$.

For example, if we want to solve $$\overbrace{\begin{bmatrix} -3 & -2 \\ -1 & 0 \\ \end{bmatrix}}^A \overbrace{\begin{bmatrix} x_1 \\ x_2 \end{bmatrix}}^{\mathbf{x}}=\overbrace{\begin{bmatrix} 2 \\ 3 \end{bmatrix}}^{\mathbf{b}}$$ then the unique solution is $$\mathbf{x}=\overbrace{\begin{bmatrix} 0 & -1 \\ -1/2 & 3/2 \\ \end{bmatrix}}^{A^{-1}} \overbrace{\begin{bmatrix} 2 \\ 3 \end{bmatrix}}^{\mathbf{b}}=\begin{bmatrix} -3 \\ 7/2 \\ \end{bmatrix}.$$

But it's typically not the easiest approach:

• Computing $A^{-1}$ could be more involved than finding $\mathbf{x}$ the usual way (row operations and back substitution).

• $A$ might not be invertible, but there still might be solutions to $A\mathbf{x}=\mathbf{b}$.