Machine Learning

• Statistics requires assumptions, and there are a range of other techniques that go under the guise of machine learning.
• What these systems do is take input data and learn things from it. They usually don't align to statistics, but are powerful.
• There are connectionist systems (neural nets) including multi-layer perceptrons learning via backpropagation, Hopfield nets, and self organising maps, genetic algorithms, decision tree learning and many others.
• One proof shows that an MLP of sufficient size learning via backpropagaion presented with enough data can learn any function.
• There is also a subfield of Data Mining called text mining, for dealing with textual data.