tex/theory1_ex4.tex
author Eugen Sawin <sawine@me73.com>
Mon, 04 Jul 2011 16:41:40 +0200
changeset 9 15594952470f
parent 0 ea81d1d9d86f
permissions -rw-r--r--
KMP.
     1 \documentclass[a4paper, 10pt, pagesize, smallheadings]{article}  
     2 \usepackage{graphicx}
     3 %\usepackage[latin1]{inputenc}
     4 \usepackage{amsmath, amsthm, amssymb}
     5 \usepackage{typearea}
     6 \usepackage{algorithm}
     7 \usepackage{algorithmic}
     8 \usepackage{fullpage}
     9 \usepackage[all]{xy}
    10 \title{Theory I, Sheet 4 Solution}
    11 \author{Eugen Sawin}
    12 \renewcommand{\familydefault}{\sfdefault}
    13 \include{pythonlisting}
    14 
    15 
    16 \pagestyle{empty}
    17 \begin{document}
    18 \maketitle
    19 
    20 \section*{Exercise 4.1}
    21 \begin{figure}[h]
    22 \entrymodifiers={}
    23 \xymatrix
    24 {
    25 & & *++[o][F-]{50} \ar@{-}[dl] \ar@{-}[dr] & & & & & & *++[o][F-]{50} \ar@{-}[dl] \ar@{-}[dr]\\
    26 & *++[o][F-]{30} \ar@{-}[dl] \ar@{-}[dr] & & *++[o][F-]{60} \ar@{-}[dr] & \ar@{=>}[rr]^{delete(40)}_{upout(30)} & & & *++[o][F-]{30} \ar@{-}[dl] & & *++[o][F-]{60}\ar@{-}[dr]\\
    27 *++[o][F-]{20} & & *++[o][F-]{40} & & *++[o][F-]{70} & & *++[o][F-]{20} & & & & *++[o][F-]{70}\\
    28 }
    29 \end{figure}
    30 \begin{figure}[h]
    31 \xymatrix
    32 {
    33 & & *++[o][F-]{50} \ar@{-}[dl] \ar@{-}[dr] & & & & & & *++[o][F-]{50} \ar@{-}[dl] \ar@{-}[dr]\\
    34 & *++[o][F-]{30} \ar@{-}[dl] & & *++[o][F-]{60} \ar@{-}[dr] & \ar@{=>}[rr]^{delete(60)}_{upout(50)} & & & *++[o][F-]{30} \ar@{-}[dl] & & *++[o][F-]{70}\\
    35 *++[o][F-]{20} & & & & *++[o][F-]{70} & & *++[o][F-]{20} & & & & \\
    36 }
    37 \end{figure}
    38 There are no rotations required because the AVL-balance property is not violated.
    39 
    40 \section*{Exercise 4.2}
    41 We use the following Python function for the calculation of the hash code, which resembles the Java code presented in the lecture:
    42 \begin{python}
    43 def hash(string, m):
    44     return sum(map(ord, string)) % m   
    45 \end{python}
    46 We choose $m = 13$ and get following results:\\\\
    47 \begin{tabular}{ l  c  l}
    48   \textbf{string} & \textbf{hash} & \textbf{collisions}\\ \hline 
    49   \emph{Hashing} & 4 & \\
    50   \emph{Exercise} & 5 & \\ 
    51   \emph{THEORYI} & 2 & \\
    52   \emph{god} & 2 & \emph{THEORYI}\\
    53   \emph{science} & 2 & \emph{THEORYI, god}\\
    54 \end{tabular}\\\\
    55 As you can see there are conflicts between \emph{god, science} and \emph{THEORYI}.
    56 
    57 \section*{Exercise 4.3}
    58 Let $U = \{0,...,N-1\}$, where $N = 49$ and $m = 35$. Let $a_i = 42i$ and $b_i = 28i$.\\
    59 Let $H = \{h_i(k) = ((a_ik + b_i) \bmod N) \bmod m\}$ for $ i \in \{1,...,N(N-1)\}$ be a class of hash functions. We show that $H$ is \emph{not} universal by contradiction.\\\\
    60 We assume $H$ is universal.
    61 % \left\{ x \in X \mid x > \frac{1}{2}\right\} \]
    62 \[\iff \frac{|\{h \in H \mid h(x) = h(y) \land x \neq y\}|}{|H|} \leq \frac{1}{m}\]
    63 First we look at the mapping space of the inner modulo function for all pairs $(q, r)$ with \\$q = (a_ix + b_i) \bmod N$ and $r = (a_iy + b_i) \bmod N$ for all $x,y \in U$ and $x \neq y$.
    64 \[q = r \iff (a_ix + b_i) \bmod N = (a_iy + b_i) \bmod N\]
    65 \[\iff a_i(x - y) \bmod N = 0\]
    66 \[\iff \exists_{c \in \mathbb{N}}: a_i(x - y) = cN\]
    67 \[\implies a_i|N \land (x - y)|N\]
    68 \[\iff \exists_{l \in \mathbb{N}, 0<l\leq7}: (x - y) = 7l\]
    69 It follows that there are collisions for $(x - y) = 7l$ with $c = \frac{a_i \cdot 7l}{N} = \frac{42i \cdot 7l}{49} = 6il$\\
    70 To determine the number of inner collisions, we calculate $\sum_{0 < l \leq 7} (48 - 7l)$.
    71 \[\sum_{0 < l \leq 7} (48 - 7l) = 7(48 - \sum_{0 < l \leq 7}l) = 140\]
    72 We know that the number of total collisions of a hash function is greater or equal to the number of collisions of its inner modulo mapping space.
    73 \[ \implies \frac{140}{|H|} = \frac{140}{49 \cdot 48} > \frac{1}{35} \implies Contradiction!\]
    74 Even within the mapping space of the inner modulo function we get more collisions than accepted for a universal class of hash functions; it follows that $H$ is \emph{not} universal. \qed
    75 \end{document}