Machine Learning Roadmap

Hi,

I am searching for a nice structured overview (graphical or not) over Data Mining and ML procedures...Apart from the rough distinction of unsupervised,semi- or supervised problems in these classes it would be good to have an overview over how different properties of the given data make on algorithm or the other more effective.. (for example if you have a huge dataset don't use ridge regression because you will have to perform an extremely costly matrix inversion...better use a sparse method..)..Also in the unsupervised case when to prefer a manifold learning technique over kernelized linear techniques for DimRed... If you know something like this, even in a book or so, please let me know..(I would be happy not to design it myself but also don't know where to search for such an overview).

Thank you very much ! :>