\relax 
\@writefile{toc}{\contentsline {section}{\numberline {1}Influence maximization}{1}}
\newlabel{subsec:p1a}{{1.1}{1}}
\@writefile{toc}{\contentsline {subsection}{\numberline {1.1}Example where $f(S_2) < f(T)$}{1}}
\@writefile{lof}{\contentsline {figure}{\numberline {1}{\ignorespaces Example where $f(S_2) < f(T)$.}}{1}}
\newlabel{fig:p1a}{{1}{1}}
\@writefile{lof}{\contentsline {figure}{\numberline {2}{\ignorespaces Example where $f(S_3) < 0.8 \cdot f(T)$.}}{2}}
\newlabel{fig:p1b}{{2}{2}}
\@writefile{toc}{\contentsline {subsection}{\numberline {1.2}Example where $f(S_3) \leq 0.8 f(T)$}{2}}
\newlabel{eqn:p2b_ineq}{{2}{2}}
\@writefile{lof}{\contentsline {figure}{\numberline {3}{\ignorespaces A possible influence set structure that results in a bad data-dependent bound.}}{3}}
\newlabel{fig:p1d}{{3}{3}}
\newlabel{subsec:optgreedy}{{1.3}{3}}
\@writefile{toc}{\contentsline {subsection}{\numberline {1.3}Sufficient property for $f(S_i) = f(T)$}{3}}
\@writefile{toc}{\contentsline {subsection}{\numberline {1.4}Bad data-dependent bound}{3}}
\@writefile{lof}{\contentsline {figure}{\numberline {4}{\ignorespaces Simulation (blue dots) of the probability distribution $P(x) = \frac  {\alpha -1}{x_\mathrm  {min}}\left (\frac  {x}{x_\mathrm  {min}}\right )^{-\alpha }$ (red dashed line).}}{4}}
\newlabel{fig:p2b}{{4}{4}}
\@writefile{toc}{\contentsline {section}{\numberline {2}Empirical power laws}{4}}
\@writefile{toc}{\contentsline {subsection}{\numberline {2.1}Complementary CDF}{4}}
\@writefile{lof}{\contentsline {figure}{\numberline {5}{\ignorespaces Simulation (blue dots) of the probability distribution $P(x) = \frac  {\alpha -1}{x_\mathrm  {min}}\left (\frac  {x}{x_\mathrm  {min}}\right )^{-\alpha }$ (red dashed line).}}{5}}
\newlabel{fig:p2c}{{5}{5}}
\@writefile{toc}{\contentsline {subsection}{\numberline {2.2}Simulated dataset with $\alpha =2$ and $x_\mathrm  {min}=1$}{5}}
\newlabel{subsec:fithist}{{2.3}{5}}
\@writefile{toc}{\contentsline {subsection}{\numberline {2.3}Least squares fit for $\alpha $ on PDF}{5}}
\newlabel{subsec:fitccdf}{{2.4}{6}}
\@writefile{toc}{\contentsline {subsection}{\numberline {2.4}Least squares fit for $\alpha $ on CCDF}{6}}
\newlabel{subsec:mle}{{2.5}{6}}
\@writefile{toc}{\contentsline {subsection}{\numberline {2.5}MLE estimate of $\alpha $}{6}}
\newlabel{eqn:mlealpha}{{8}{6}}
\@writefile{toc}{\contentsline {subsection}{\numberline {2.6}Comparison of estimation methods}{6}}
\@writefile{lof}{\contentsline {figure}{\numberline {6}{\ignorespaces Word frequency distribution from Moby Dick and Don Quijote sources. The MLE fit (red dashed line) uses the formulas from the continuous power-law distribution.}}{8}}
\newlabel{fig:p3a}{{6}{8}}
\@writefile{toc}{\contentsline {section}{\numberline {3}Combination of exponentials for generating power laws}{8}}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.1}Empirical power laws in word-frequency distributions}{8}}
\@writefile{lof}{\contentsline {figure}{\numberline {7}{\ignorespaces Word frequency distribution from Moby Dick and Don Quijote sources. The parameter $x_\mathrm  {min}$ was fixed at $1$, while the MLE curve was computed as a function of $\alpha $ with the normalization constant appropriate for a discrete power-law distribution. It can be seen that the MLE fit on the discrete model (black curve) performs better than the one based on the continuous model (red curve).}}{9}}
\newlabel{fig:p3a_discrete}{{7}{9}}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.2}Theoretical power laws in monkey novels}{11}}
\@writefile{toc}{\contentsline {subsubsection}{\numberline {3.2.1}Distinct words of length $y$}{11}}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.3}Relative frequency of a particular word of length $y$}{11}}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.4}Probability of a word having relative frequency $x$}{11}}
\@writefile{toc}{\contentsline {section}{\numberline {4}Exploring robustness of networks}{11}}
\@writefile{toc}{\contentsline {subsection}{\numberline {4.1}Diameter robustness}{11}}
\@writefile{toc}{\contentsline {subsection}{\numberline {4.2}$|\mathrm  {LCC}|$ robustness}{12}}
\@writefile{lof}{\contentsline {figure}{\numberline {8}{\ignorespaces The graph diameter as a function of the number of nodes removed. Two deletion policies are considered: ``Attack''-mode (circles) where nodes with highest degree are targeted first; ``Failure''-mode (triangles) where nodes are removed at random. [a] The variation in graph diameter in the initial $2\%$ removal. [b] The variation in graph diameter as up to $50\%$ of the nodes are removed.}}{13}}
\newlabel{fig:p4a}{{8}{13}}
\@writefile{lof}{\contentsline {figure}{\numberline {9}{\ignorespaces The size of the largest connected component as a function of the number of nodes removed. Two deletion policies are considered: ``Attack''-mode (circles) where nodes with highest degree are targeted first; ``Failure''-mode (triangles) where nodes are removed at random.}}{14}}
\newlabel{fig:p4b}{{9}{14}}