This project is read-only.

Project Description
Marvin (or short Marvy) is a machine learning library for the .NET framework. The library is designed for high flexibility. For example, you can substitute the standard minimization algorithm (currently gradient descent in the first release; then BFGS or l-BFGS) by your own choice to optimize for speed in high-demand production environments.

Current Stage of Development
This project is still in the very early stages of develpoment. Currently, I'm following a hybrid approach of partly prototyping and test driven development. First I design interfaces, then the implement a prototype with test-driven development. The prototype will test the viability of the interface design.

The prototype I'm currently developing is a multivariate linear regression algorithm and a gradient descent. Then I'll implement prototypes of a broad set of algorithms to figure out whether the interfaces work in a broad set of different ML situations. Once this Pre-Alpha stage is over, the interfaces will be declared stable. Then I'll focus development on the implementation. 

Last edited Apr 16, 2013 at 11:05 PM by daebwae, version 5