All bleeding edge research code is written in Matlab as many of my collaborateurs use it. Let me know if you want an implementation in a specific programming language and I'll help you get started as much as possible.
I have contributed to a linear programming API written in Java. SCPSolver is a front-end to various linear programming libraries intended for rapid prototyping and easy installation. If you want to solve small to medium-sized LPs in Java, SCPSolver might just be what you are looking for.
For my undergraduate research thesis, I have developed a Java API for a specific case of Distance Metric Learning, called Label Metric Learning. For the research thesis, I only tried this approach with support vector regression, but more recently, I thought about hooking it up to a deep neural net, but have not yet come around to it. Might be interesting to see what happens.