Theoretical neuroscience is an umbrella term encompassing many current approaches. Dynamical approach to neuroscience is a particular theoretical approach that utilizes time-smooth continuous equations to describe neural events as the neural systems evolve through state-space. They are structurally deterministic, but noise terms and random processes can be integrated into them. There's also symbolic dynamics which utilize the Markov partition.
A large part of dynamical systems approach is studying quantities geometrically (i.e you draw the average trajectories of your system and you begin to see structures that have functional meaning in the state-space of the system). You generally measure quantities of the system like: the lyapunov constant, the natural measure, the basin of attraciton, etc.