[taliesinb]

Choose a path.

What are you interested in? The theory of information? Ideas about space that go behind the flat plane of a piece of paper? How to communicate and co-operate securely when people might be eavesdropping and impersonating your allies? How to model various kinds of learning? How one studies the interacting strategies of military, economic, or biological adversaries? How complexity arises in simple systems? How sounds and images can be compressed?

These are just examples, but each one of these topics will select a certain 'diet' of pure mathematics that will keep you well fed with cool ideas as well as 'in's to different fields.

Although it does work for some people to just pick up a book on linear algebra and read it from cover to cover, I've never found that a very interesting or self-motivating way to learn. I think its better to hang one's knowledge on a tree of interlinked explorations.

[DharmaSoldat]

I second that. My main side-project-esque interests involve AI and Computer Vision. Knowing that cleared up a lot of confusion as to what was 'best to learn'.