Rosetta Stone for Machine Learning

I am looking to help build an archive of straightforward examples for making Machine Learning work via different platforms. I am looking for very terse documentation on how to make a given platform absorb a CSV containing mixed categorical and numeric data, attempt to build a model, report on whether the model is more predictive than chance, and then apply the model to new data. I want this for as many platforms as possible, including: MLpack, mxnet, [login to view URL], torch, keras, theano/tensorflow, dlib, vowpal wabbit, caffe, xgboost, cntk, scikit-learn (with all its various modes), aerosolve. smile, h2o, weka, spark mllib, deeplearning4j, [login to view URL], [login to view URL], anything unmentioned in python or r. Assume no more than 256 columns, assume columns may be either floating point or text fields that may be hashed into a unique identifier. You don't have to document installation. Assume prereqs are there, I just want to know how to apply to data. You're welcome to document knobs but they're not required. This can't just be a word document, code has to work, but you can do things like hardcode filenames and configurations. BE TERSE. The more platforms you can document the more interested I am.