Many many thanks in advance to anyone who answers: I've futzed around with ML docs (dataquest, random articles, first 3 weeks of andrew ng course on coursera)

Many many thanks in advance to anyone who answers: I've futzed around with ML docs (dataquest, random articles, first 3 weeks of andrew ng course on coursera) & what I don't get about this (http://tensorflow.org/tutorials/mnist/beginners/index.md http://tensorflow.org/tutorials/mnist/beginners/index.md) is how to actually make a prediction / classification on something that is not "test" data. Can anyone point me in the right direction of how to "use" a model once I have it validated, like we do at the end of that tutorial?