There are several approaches to theoretical neuroscience. I am currently taking the physics/mathematics approach: modelling the currents in neurons and coupling several neurons together through differential equations.
In Computer Science, the tendency is more towards machine learning: Bayesian statistics, artificial neural networks, signal processing, robotics, artificial intelligence. It really depends on what level you want to engage the field. For instance, if you want to study vision science, you'll want to learn more about vision and color perception and some basic physics/optics. But I also think neurochemistry and neurophysics is important, just for breadth.
Computer scientists often use the leaky integrate and fire model or rate coding models. There's something called the neural engineering framework that seeks to make networks of neurons do standard logical operations. One of their projects is SPAUN, a whole brain simulation.