I opened a github account and began a project on Krylov subspace methods in Java.  This project is for educational purposes and reproduces fundamental, published algorithms to make them available for experimentation.

An Introduction

I am an applied mathematician with interests in convex optimization and machine learning.  For some years I've been programming in Java and C++ for my job in private industry. 

Recently,  I've begun to learn test-driven design, and I'd like to see what this technique has to contribute to the object oriented design of numerical code.  I realize that currently native code is faster than Java in some numerical applications.  However,  Java is statically typed and Java IDEs allow refactoring and facilitate test-driven design.  This would seem to make Java an ideal prototyping language for object oriented numerics.  I hope that I can demonstrate its advantages.