\frametitle{Goal}
\begin{block}{Problem}
\begin{itemize}
\tightlist
\item
\emph{Goal} : \emph{minimize packet loss ratio} (\(=\) maximize nb of
received \texttt{Ack}) in a \emph{finite-space discrete-time Decision
Making Problem}.
\item
\emph{Solution ?} \textbf{Multi-Armed Bandit algorithms},
\textbf{decentralized} and used \textbf{independently} by each dynamic
device.
\end{itemize}
\pause
\end{block}
\begin{block}{\emph{Decentralized} reinforcement learning optimization!}
\begin{itemize}
\tightlist
\item
Max transmission rate \(\equiv\) \textbf{max cumulated rewards}
\hspace*{0.25\textwidth}\(\max\limits_{\text{algorithm}\;A} \;\; \sum\limits_{t=1}^{T} \sum\limits_{j=1}^M r^j_{A(t)}\).
\item
Each player wants to \textbf{maximize its cumulated reward},
\item
With no central control, and no exchange of information,
\item
Only possible if : each player converges to one of the \(M\) best
arms, orthogonally (without collisions).
\end{itemize}
\end{block}