Aim of amap package is to propose to end users (statistician and bioolgists)
powerfull analysis tools. Statisticians requires many options and new
algorithms as described in \cite{caussinu+ruiz2} whereas biologists
needs very optimised standard algorithms. This package will be based on R
language.
\section{Overall description}
Amap package will be a standard R package, using convention
described in \cite{R:writtingRExt}.
It will provides
\begin{itemize}
\item implementation of new algorithms of robust pca
described in \cite{caussinu+ruiz2},
\item implementation of new algorithm of optimal partition described in
\cite{mpetijean},
\item a k-means algorithm based on a rank-metric,
\item a hierarchical clustering optimised for multi-processor
servers, with the minimum memory allocation.
\item the possibility to use rank based metrics with hierarchical clustering
\end{itemize}
\section{Requirements}
\subsection{Scope}
\req{Amap package must implement algorithms described in \cite{caussinu+ruiz2}.}
\req{Amap package must implement algorithms described in \cite{mpetitjean}.}
\req{Amap package must implement a k-means algorithm based on a rank-metric.}
\req{Amap package must implement a rank based metrics with hierarchical clustering.}
\subsection{Performances}
\req{All algorithm must be coded in fortran, C or C++; R code will
be limited to data manipulation}
\req{Hierarchical clustering must compute a set of 15000 individuals with 200
variables on a dual Xeon 3.2~Ghz in less than
half an hour with less than 1 Go memory used.}
\section{Methods, tools and techniques}
\req{Amap package will be a standard R package, using convention
described in \cite{R:writtingRExt}. All function must have a user
manual}
\section{Qualification}
Amap package will provides a set of unitary tests, in order to check both
exactitude of algorithms, and non regression.\\
\req{All algorithms will be ``proven'' by comparison to another implementation
with some references data tests}
\req{All functions have an example detailed in documentation, that command R CMD check is able to run without error. Output of theses examples will be archived
in order to check non regression.}
\req{An R script will test all functions with the combinaison of
all parameters. Output of this script will be archived in order to check