There are various approaches for training algorithms from computer science, statistics, cognitive science, and philosophy. Version space, neural networks, decision trees, genetic algorithm, graphical  models, instance-based learning, learning logical rules, and reinforcement are typical methods to learn patterns being in training examples.

 

Graphical models can be divided into two-fold in general. The first method is directed graphical model(DAG). It is a way that enables to represent causality. Markov models, HMM, and Bayesian networks are representative algorithms in DAG. The second method is undirected graphical model. Conditional random field(CRF) is a well known algorithm.

 

Machine Learning can be applied for either explanatory or predictive use. A predictive use of machine learning means that it tries to fill in unknows in data. An explanatory use is to identify a model which can give a human user relationships hidden in the data.