Machine Learning: Machine is Learning Like a Human Being

Yihui Xie 2007-10-29

During the class of this evening, Dr MA mentioned his experiences about the Arcing algorithm, among which he said that a few years ago he realized a fact that what the computer does in machine learning is just simulating the process in the brain of the human being.

I strongly agree with him in this point, because I always make computer programs by myself to finish some computations or algorithms; I know the difference between the humankind and the computers. Sometimes we don’t need iterations in our brain to classify sample points by some straight lines or curves (or hyperplanes), but computers do need!

Computers seem to be “stupid”, but they “don’t care”; we just need to “teach” them. For machine learning, what we do is to teach machines to learn from their “mistakes” (say, misclassification). Both Ada-Boost and Arcing are such kind of algorithms.