chiteri/shdnbi2013_machinelearning

A project based on ideas of Machine Learning (M.L) : We need to train computers to accurately recognize sentences in either Swahili or English out of reading entries from Wikipedia in both languages. We will then use similar rules and techniques to detect the presence of particular sub-atomic particles inside Physics data collected from the Compact Muon Solenoid (C.M.S) experiment at CERN. SCIENCE HACK DAY, Nairobi - 2013: MACHINE LEARNING FOR PHYSICS =============================================================== A project based on ideas of Machine Learning _(M.L)_ : We need to train computers to accurately recognize sentences in either Swahili or English out of reading entries from Wikipedia from both languages. We will then use similar rules and techniques to detect the presence of particular sub-atomic particles inside Physics data collected by collaborators at the Compact Muon Solenoid (C.M.S) experiment at CERN. scikit-learn example and our experiements at Siena. ======================================================= ``Hi Morris and Martin, It's great to see you've already started looking at the scikit-learn tools! That will probably really help things this weekend. Here at Siena, I've started to look into this stuff as well with a student of mine, Jess Muenkel (cc'ed on this email). I've wanted to look at these python tools for some time, so this is a great excuse for me to dedicate this time! :) I've attached the very first example that Jess and I wrote. All of this is based on the scikit examples. For now, we've chosen to focus on the Decision Tree (DT) algorithm, as I think it's supposed to be relatively easy to use. I should point out, we've not spent any time trying to actually learn what the algorithm does (yet) we've only got it to work. scikit-learn.org scikit-learn.org The example (reasonably commented) walks through the standard procedure for ML (Machine Learning): * Choose an algorithm. * Give it two (or more) datasets from different populations. These are called TRAINING samples. * Train the algorithm. * Give it another sample of *different* data where you know the answers and see how well the algorithm predicts the dataset. This is called the TESTING sample. If you have scikits and matplotlib installed, you should be able to run this. In fact, I urge you to test it out …