In this chapter we move from previous tabular representation for state/action value to function representation. That is to say, we use a function to fit the true expression of the state/action value function. Such a function can be predifined, e.g., a linear function, or a neural network.
The reason why we move from tabular-based representation to function-based are:

