this is for holding javascript data
Vadim Kosoy edited subsection_Orthogonality_Theorem_There_is__.tex
about 8 years ago
Commit id: a21d30fe7f90e13e2a7eee0a9540a88b86fc843c
deletions | additions
diff --git a/subsection_Orthogonality_Theorem_There_is__.tex b/subsection_Orthogonality_Theorem_There_is__.tex
index aeb118e..98c2902 100644
--- a/subsection_Orthogonality_Theorem_There_is__.tex
+++ b/subsection_Orthogonality_Theorem_There_is__.tex
...
There is a variant of definition ~\ref{def:op} which is nearly equivalent and is often useful.
We can think of functions $f: \Supp \mu \rightarrow \Reals$ as vectors in a real
Hilbert inner product space with inner product $\Chev{f,g}:=\E_\mu[fg]$. Informally, we can think of $\Gamma$-schemes as a subspace (although a $\Gamma$-scheme is not even a function) and an $\mathcal{E}(\Gamma)$-optimal predictor for $(\mu,f)$ as the nearest point to $f$ in this subspace. Now, given
a Hilbert space $H$, an inner product $V$, a vector $f \in
H$, V$, an actual subspace
$V $W \subseteq
H$ V$ and $p = \Argmin{q \in
V} W} \Norm{q - f}^2$, we have $\forall v \in
V: W: \Chev{v,p-f}=0$.