Machine learning cookbook / reference card / cheatsheet?
Each classifier has it's own advantages and disadvantages.
E.g. train/test speed, classification/regression (and how many classes can be handled), how many degrees of freedom, suitable/not suitable for abstract kinds of datasets (of possible to say), how interpretable is the resulting model (like for max-ent: importance of single features), and so on and so on
Do you know a good overview? If not, would you be in to create one?