This question is inspired by the following problem
If we know that the basis $B=(\mathbf{u},\mathbf{v})$, where $$\mathbf{u}= \begin{pmatrix}1 \\ 2i\end{pmatrix},\;\;\mathbf{v}=\begin{pmatrix}-i \\ 1\end{pmatrix}$$ is orthonormal under some inner product in the space $\mathbb C^2$, find $<(x_1,x_2),(y_1,y_2)>$
This following method works:
Define $A = (\mathbf{u},\mathbf{v})$ and compute $A^* A$ and you get the matrix of the inner product.
Question: Why does this work?
My progress: So far I've noticed that $A$ is the Gram matrix under the dot product, if that's of any relevance.
Thank you.
I got it, the result is not true in general and is an incredible coincidence. In general, we want to find a matrix $S$ such that $$A^* S A = I$$ (this equation gives exactly the four necessary conditions for the basis to be orthonormal). Thus $$ S = (A^*)^{-1} A^{-1}$$ Here comes the fun part. Obviously, in this case, $(A^*)^{-1} A^{-1} = A^* A$. How can that be? It is a result of three coincidences:
How all this happened AND got noticed by accident by someone (who erroneously thought $A^* A$ is the correct solution) is beyond me.