MSc. Project Presentation : Study of autonomous agents' learning performance in crossing cellular automaton based highway using machine learning techniques

CANDIDATE:  JUNLIN HAO

ABSTRACT:

A model of cognitive agents learning to cross a cellular automaton based highway is introduced. The resulting data files are generated from this simulation model under different configurations of values of the model parameters. To find out the common features among the samples with various treatments, various unsupervised machine learning techniques including principal component analysis, K-means clustering analysis and Gaussian mixture model are used in the analysis. Moreover, the supervised machine learning techniques including the multivariate linear regression, the generalized linear regression and the artificial neural networks are employed to build regression models for the data and give predictions of cognitive agents’ decisions. Through the investigation of the simulation data, the relationship between the model parameters and autonomous agents’ decisions are well explored and more features of the model are revealed. Furthermore, the improvement of the assumption of the probability distribution of the autonomous agents’ decisions is recommended to more accurately understand the structure of the simulation model.