Evolving neural network as a decision support system — Controller for a game of “2048” case study

The paper proposes an approach to designing the neuro-genetic self-learning decision support system. The system is based on neural networks being adaptively learned by evolutionary mechanism, forming an evolved neural network. Presented learning algorithm enables for a selection of the neural network structure by establishing or removing of connections between the neurons, and then for a finding the beast suited values of the network weights and biases. The algorithm was validated on problem of learning to play the game of “2048”. The game has in fact very simple rules however it is very important to have a proper control strategy in order to gain high score. Moreover, existence of random factors makes it more difficult. The evolved neural controller is trying itself to discover and to learn the best strategy leading to the highest score. In the result, the controller has obtained a game score at a level similar to medium advanced human player. It is interesting that strategies developed by the controller are similar to the strategies that are applied by experienced human players. The paper provides an overview and an analysis of the impact of the main system factors on obtained results. Finally, the most important findings are indicated and future work is specified.