Biome Simulation with Machine Learning

Rad uspored̄uje performanse neuronskih mreža i stabala odluke kao modela ponašanja za jedinke koje se nalaze u simuliranom 2D svijetu. Cijeli svijet i njegova pravila interakcije med̄u objektima unutar njega implementirana su i opisana u sklopu ovog rada. Modeli ponašanja optimiziraju se korištenjem generacijskog genetskog algoritma kojim se nastoji čim više približiti stvarnoj evoluciji iz prirode.Thesis compares the performances of neural networks and decision trees as behavior models for individuals which exist in a simulated 2D world. Whole world and its rules of interaction between the objects within it are implemented and described as a part of this thesis. Behavior models are optimized by using generational genetic algorithm which tries to mimic the real evolution found in nature