I have about 400 ~ 500 lines of what I would call average open source code contribution. Does open source code contribution matter for graduate admissions? I am interested in applying to a MS in computer science and am interested in machine learning. I have a great plan for an awesome feature for Libsvm (and open-source implementation of the popular support vector machine technique for machine learning). Will implementing it help my application ?
Admission to graduate school is based primarily on your potential for research. Independent, creative, intellectual work of any kind strongly correlates with potential for research, especially if that work is directly related to the interests described in your research statement. Contributing to open-source software projects is independent, creative, intellectual work.
Of course, your contribution would matter more if it were merged into the main branch of the project, but something is better than nothing.
There are a bunch of questions in one here. I'll answer the one that seems central:
Yes, I would consider that contributing to a well-known machine learning (ML) toolkit will significantly strengthen your case for an application in ML. It shows that you know the fundamentals of support vector machines already, and that you are genuinely interested in the field. Plus, it is presumably something that distinguishes you from all other candidates.