Commit 50c40a2c authored by Philipp Spohr's avatar Philipp Spohr
Browse files

removed thesis

parent 1bfd7af3
\tikzstyle{block} = [rectangle, draw, fill=blue!20,
text width=5em, text centered, rounded corners, minimum height=4em]
\tikzstyle{line} = [draw, -latex']
\begin{tikzpicture}[node distance = 1cm]
\node [block] (datamodel) {Data Modeling};
\node [block, left= of datamodel] (input) {Input};
\node [block, below= of datamodel] (reduce) {Reduction Rules};
\node [block, below right = of reduce] (solveILP) {ILP Solver};
\node [block, below left = of reduce] (solveHeuristic) {Heuristic Solver};
\node [block, below right= of solveHeuristic] (output) {Generate Output};
\path [line] (input) -- (datamodel);
\path [line] (datamodel) -- (reduce);
\path [line] (reduce) -- (solveILP);
\path [line] (reduce) -- (solveHeuristic);
\path [line] (solveILP) -- (output);
\path [line] (solveHeuristic) -- (output);
\caption{Overview of the Yoshiko Algorithm} \label{algOverview}
COMPLETE GRAPH, CHOOSE EDGES SO THAT sum of C(E) is MAX while satisfying Triangle inequalities > Fully Disjunct Clique-Graph
\ No newline at end of file
The Yoshiko Wrapper provides a clean and simple interface to generate the model.
\subparagraph{Mapping edge costs}
The user has the possibility to use a numeric Cytoscape column of the node table as a source for the edge-cost function $C$.
\subparagraph{Insertion and deletion cost}
The default values $C_I$ and $C_D$ can be set by the user with the default values being $C_I=-1$ and $C_D=1$.
It should be noted, that the insertion cost value is not normalized or in any way adjusted when a mapping is used. This means that the user needs to adjust this value wisely to fit the data.
As an example the user might have mapped the edge costs to a column containing values in the range of $10^6-10^7$.
The default insertion cost of $-1$ will be irrelevant in comparison and the algorithm will most likely insert all missing edges and generate one big cluster as a solution.
Overall the ratio $R = \frac{|C_I|}{C_D}$ should give you a rough idea how the algorithm will operate.\\
$R > 1$ means, that the algorithm is more likely to delete edges to generate cliques, a value of $R<1$ means insertions are more likely.
\subparagraph{Mapping permanent or forbidden edges}
The Yoshiko app has additional convenience functions:
The user can map edges to a boolean Cytoscape column to mark them as either
\textbf{forbidden} (meaning that those edges will never be part of the solution) or
\textbf{permanent} (meaning that those edges will always be part of the solution).
Marking an edge e as forbidden is equivalent to $C(e)=-\infty$,
marking an edge e as permanent is equivalent to $C(e)=\infty$.
This way the user is able to apply expert knowledge about the input instance to increase the quality of the solution.
The Yoshiko algorithm models the data as a complete graph
with an associated edge-cost function
$C: E \rightarrow \mathbb{R}
\cup \lbrace -\infty \rbrace
\cup \lbrace \infty \rbrace $.
As many input instances do not describe a full graph, missing edges and costs need to be modeled.
This is achieved by using default values for insertion or deletion.
A default insertion cost $C_I \in [-\infty,0]$ is used as $C(e)$ whenever the input instance does not contain an edge $e$.
A default deletion cost $C_D \in [0,\infty ]$ is used as $C(e)$ whenever the input instance does contain an edge $e$ that has no cost associated yet.
\ No newline at end of file
\author{Philipp Spohr}
\title{Rough Analysis of Cluster Editing Reduction Rules in Multi-Graphs}
\ No newline at end of file
\author{Philipp Spohr}
\title{Developing and evaluating a Cytoscape app for graph-based clustering}
This paper describes an implementation of the Yoshiko-alorithm,
which clusters data based on the weighted cluster-editing problem,
as a plugin for Cytoscape.
Cytoscape is a network visualization and analysis tool (insert info)
Yoshiko is based on work from (insert info)
\section{The Yoshiko-App for Cytoscape}
\subsection{Technical Details}
\subsubsection{Program Structure}
\subsubsection{Data Modeling}
\section{Evaluation of the Yoshiko Algorithm}
\subsection{Multigraph Analysis}
\subsection{Integration in other frameworks}
\ No newline at end of file
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