3rd Year Engineering 'Machine Learning' project at University of Cambridge (CUED)
Predicting the dynamics of the cart-pole system by employing linear regression with non-linear Gaussian basis functions (to expand the dimensionality of the input state).
Control the system with linear and non-linear policies, which define the action to be taken to give rise to the desired behaviour of the pole being balanced vertically upwards (i.e. the unstable equilibrium).
The final controller consisted of a non-linear policy and two linear policies.