Yeah, maybe something with differential geometry and ML would be fun. I’ve also built a system for composing relations and generating gradients from a computat

Yeah, maybe something with differential geometry and ML would be fun. I’ve also built a system for composing relations and generating gradients from a computational graph...(for CAD design generation) Maybe I should look at working on a framework project. Whatever the theano folks are doing nowadays - or the like.