You would start of by creati g a C# project using your IDE, or the dot et command line. Next you would implement the algorithm.
If you already have C++ code then you could try porting it. I advice caution though. You have to understand the code that you are porting. If you don't understand the code, then you will likely be better off starting from scratch and learn the algorithm, and then implement it straight into C#. Porting code that you don't understand will not actually save you time when it is time to debug the code. Even worse, if you are doing this in an academic environment, you will have to deal with any plagiarism issues if you do not cite your source for the original C++ code.