Rad se bavi izradom agenta za igru Othello. Agent je implementiran korištenjem algoritma minimax i alfa-beta podrezivanja. Kao heuristička funkcija korištena je neuronska mreža trenirana genetskim algoritmom. Modeli su trenirani igranjem međusobno. Kao procjena koliko je model dobar odigran je velik broj partija protiv agenta koji igra nasumične poteze. Razvijena su dva modela različite kompleksnosti od kojih je jednostavniji pokazao malo bolje rezultate. This work deals with creating an agent for the game Othello. The agent is implemented using the minimax algorithm and alpha-beta pruning. A neural network trained with a genetic algorithm was used as a heuristic function. The models were trained playing versus themselves. To evaluate how good a model performs a large number of games were played versus an agent that moves randomly. As a result two models of different complexities were developed where the simpler one performed a bit better.
