This article presents a theoretical and conceptual proposal of a novel direct search method of optimization, Game of Patterns (GofP) method, so-called by the author, for solving unconstrained mixed integer nonlinear optimization problems. The GofP method is based on: a set of $\eta$ random patterns of search, which initially form the set of $\eta$ active players, namely, $\{\mathcal{H}_{\ell}^{[0]}\}_{\ell=1}^{\eta}$; and a game rule framework for solving our problems. At each \mbox{$k$th} game round, each \mbox{$\ell$th} player $\mathcal{H}_{\ell}^{[k]}$ will explore inside his own pattern in the current \mbox{$k$th} round, if the player is active. The strategy of each \mbox{$\ell$th} player is given by a random quantity $\mathcal{S}_{\ell}^{[k]}$, which will allow each player to explore inside his own pattern by a set of $\mathcal{S}_{\ell}^{[k]}$ trial points. At the beginning of each \mbox{$k$th} game round, each active \mbox{$\ell$th} player bets $\mathcal{S}_{\ell}^{[k]}$ according to his budget $M_{\ell}$ for each round. At the end of each \mbox{$k$th} round, the player that had identified the best objective function value is considered the winner of the \mbox{$k$th} round, therefore the rest of active players must pay off the winner their bets. This process is recurrently repeated until that had been disqualified \mbox{$\eta-1$} players from the game. It is worthwhile to point out that any player is disqualified if his balance account $b_{\ell}^{[k]}$ is less than his budget $M_{\ell}$ for each round.
Creative Commons Attribution-NonCommercial-NoDerivatives 4.0
/
Search inscriptions
