Commit 95cb5ebd authored by robertchristensen's avatar robertchristensen

made suggested changes in lecture 2 notes.

parent be10f1cc
...@@ -133,10 +133,11 @@ We will define the function $\gt(x,y)$ to be ...@@ -133,10 +133,11 @@ We will define the function $\gt(x,y)$ to be
\begin{marginfigure} \begin{marginfigure}
\centering \centering
\includegraphics{figs/ge_matrix_2} \includegraphics{figs/ge_matrix_2}
\caption{Example matrix for $\gt(x,y)$ when $|x| = |y| = 5$ with fooling set elements of size 2 drawn.}
\end{marginfigure} \end{marginfigure}
\sunote{Better to have just the figure as a PDF and use the caption for the %\sunote{Better to have just the figure as a PDF and use the caption for the
text. This allows you to use the \gt command for the function name, to get a % text. This allows you to use the \gt command for the function name, to get a
consistent look.} % consistent look.}
According to the definition of $\gt(x,y)$ the lower triangle According to the definition of $\gt(x,y)$ the lower triangle
...@@ -159,26 +160,28 @@ follows directly. ...@@ -159,26 +160,28 @@ follows directly.
% recording ~20:20 % recording ~20:20
\paragraph{Example: \disjoint} \paragraph{Example: \disjoint}
\sunote{Didn't we define \disjoint in the last lecture notes?} %\sunote{Didn't we define \disjoint in the last lecture notes?}
%\ronote{In response, we did define \disjoint, but in in lecture 1
We define $x$ and $y$ to be a vector of bits % we did not reason about the definition with as much detail.}
which define set membership. %
%We define $x$ and $y$ to be a vector of bits
\begin{align*} %which define set membership.
& x,y \in [0,1]^n \\ %
\text{and } & x_i \Leftrightarrow i \in S_x \\ %\begin{align*}
& y_i \Leftrightarrow i \in S_y % & x,y \in [0,1]^n \\
\end{align*} %\text{and } & x_i \Leftrightarrow i \in S_x \\
% & y_i \Leftrightarrow i \in S_y
the function $\disjoint(x,y)$ is defined as follows %\end{align*}
%
\begin{equation*} %the function $\disjoint(x,y)$ is defined as follows
\disjoint(x,y) = \left\{ %
\begin{array}{l l} %\begin{equation*}
1 & \quad \text{if $(x,y)$ are disjoint: } \left\{ \forall i, x_i \land y_i = 0 \right\} \\ %\disjoint(x,y) = \left\{
0 & \quad \text{otherwise} % \begin{array}{l l}
\end{array} \right. % 1 & \quad \text{if $(x,y)$ are disjoint: } \left\{ \forall i, x_i \land y_i = 0 \right\} \\
\end{equation*} % 0 & \quad \text{otherwise}
% \end{array} \right.
%\end{equation*}
\begin{marginfigure} \begin{marginfigure}
...@@ -191,9 +194,9 @@ $1$ & $0$ & $1$ & $1$ & $0$ & $0$ \\ ...@@ -191,9 +194,9 @@ $1$ & $0$ & $1$ & $1$ & $0$ & $0$ \\
$1$ & $1$ & $1$ & $0$ & $0$ & $0$ \\ $1$ & $1$ & $1$ & $0$ & $0$ & $0$ \\
\end{tabular} \end{tabular}
\\ \vspace{1em} \\ \vspace{1em}
Matrix for $DISJ(x,y)$ when $n=4$ showing which \caption{Matrix for $DISJ(x,y)$ when $n=4$ showing which
combinations will result in a $1$ or combinations will result in a $1$ or
a $0$. a $0$.}
\end{marginfigure} \end{marginfigure}
Let us observe one of the items in the Let us observe one of the items in the
...@@ -201,7 +204,7 @@ fooling set using our example. Using the ...@@ -201,7 +204,7 @@ fooling set using our example. Using the
pair $(x_1, y_1) = (01, 10)$ and $(x_2, y_2) = (10, 01)$ pair $(x_1, y_1) = (01, 10)$ and $(x_2, y_2) = (10, 01)$
we can see this is part of the fooling set. This set we can see this is part of the fooling set. This set
is interesting because $(x_1,y_1) = (\overline{x_2}, \overline{y_2})$. is interesting because $(x_1,y_1) = (\overline{x_2}, \overline{y_2})$.
In fact, the fooling st for $DISJ(x,y)$ are the bit patterns with the bits flipped; In fact, the fooling st for $\disjoint(x,y)$ are the bit patterns with the bits flipped;
it is the set of all $A$ and $\overline{A}$. it is the set of all $A$ and $\overline{A}$.
% recording ~33:20 % recording ~33:20
...@@ -348,7 +351,31 @@ rest we do not care about. We are choosing some subset of the ...@@ -348,7 +351,31 @@ rest we do not care about. We are choosing some subset of the
$r$ bits to be fixed to $1$, the rest we can pick to be anything $r$ bits to be fixed to $1$, the rest we can pick to be anything
else. Doing this results in $2^n - 2^{n-r}$ else. Doing this results in $2^n - 2^{n-r}$
\sunote{See the note that I provided in the discussion group and insert it here} % \sunote{See the note that I provided in the discussion group and insert it here}
The statement that the rank of the IP matrix over GF($2$)$=2^n$ is
not true. A proof of this can be found in Sherstov's
notes\footnote{http://www.cs.ucla.edu/~sherstov/teaching2012-winter/docs/lecture03.pdf},
which we will briefly discuss here.
The first observation is that the rank argument works over any field. So
we can choose the field we wish to evaluate matrix rank over. In particular,
the rank over the reals is larger than the rank over any other field.
Recall that we would like to show that the inner product has deterministic
communication complexity $n$ via the rank method. So we need to show that
the rank of the induced matrix (where $M_{ij}=<\!\!x_i,y_j\!\!>$) is large:
remember that IP is a function from $n$ bit string to $\{0, 1\}$.
Firstly, consider the auxiliary matrix $M'=2M-1$ (where $1$ is the all ons matrix).
All this does is change $0$ to $-1$ and $1$ to $1$, which is more convenient
to work with. It is a nonsingular transformation, so $\text{rank}(M) \ge \text{rank}(M')-1$.
Now it is easy to see that $M'_{\text{IP}}$ on one-bit strings is the Hadamard
matrix\footnote{$ \begin{bmatrix} 1 & 1 \\ 1 & -1 \end{bmatrix}$} and the IP matrix for $n$ bit strings is merely the $n$-th order
version of this (or the $n$th tensor power of this matrix). But this matrix
has full rank: in particular its rank is $2^n$. Therefor the rank of the IP matrix
is at least $2^n-1$.
%recording ~1:00:00 %recording ~1:00:00
\section{Randomized Algorithms} \section{Randomized Algorithms}
......
<?xml version="1.0"?> <?xml version="1.0"?>
<!DOCTYPE ipe SYSTEM "ipe.dtd"> <!DOCTYPE ipe SYSTEM "ipe.dtd">
<ipe version="70005" creator="Ipe 7.1.5"> <ipe version="70005" creator="Ipe 7.1.5">
<info created="D:20140907182737" modified="D:20140908095409"/> <info created="D:20140907182737" modified="D:20140915144053"/>
<preamble>\usepackage{amsmath}</preamble> <preamble>\usepackage{amsmath}</preamble>
<ipestyle name="basic"> <ipestyle name="basic">
<symbol name="arrow/arc(spx)"> <symbol name="arrow/arc(spx)">
...@@ -233,9 +233,7 @@ h ...@@ -233,9 +233,7 @@ h
<page> <page>
<layer name="alpha"/> <layer name="alpha"/>
<view layers="alpha" active="alpha"/> <view layers="alpha" active="alpha"/>
<text layer="alpha" matrix="1 0 0 1 0.727189 24.315" transformations="translations" pos="64 624" stroke="black" type="minipage" width="96" height="66.157" depth="61.31" valign="top">Example matrix for $gt(x,y)$ when \\$|x| = |y| = 5$ with fooling set elements of size 2 drawn. <text layer="alpha" matrix="1 0 0 1 0.727189 -43.8865" transformations="translations" pos="64 624" stroke="black" type="minipage" width="96" height="32.346" depth="27.43" valign="top">$
\\ \\
$
\begin{bmatrix} \begin{bmatrix}
1 &amp; 0 &amp; 0 &amp; 0 &amp; 0 \\ 1 &amp; 0 &amp; 0 &amp; 0 &amp; 0 \\
1 &amp; 1 &amp; 0 &amp; 0 &amp; 0 \\ 1 &amp; 1 &amp; 0 &amp; 0 &amp; 0 \\
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment