Paragraph Number

182

Machine learning can comprise supervised learning, semi-supervised learning, active learning, reinforcement learning or unsupervised learning. With supervised learning, the computer can receive example inputs and desired outputs, which can be provided from a database or using a learning tool; the objective is to learn one or more rules that map the inputs to the outputs. Semi-supervised learning can be different in that the computer can be given an incomplete training example input, optionally with some desired outputs missing. With active learning, the computer can only obtain training inputs for a limited set of examples, and the computer can optimize the choice of inputs to acquire labels for. With reinforcement learning, training data, e.g. inputs and desired outputs, can be given only as feedback to the program's actions in a dynamic environment, such as guiding a surgery. With unsupervised learning, no training input and/or output data are provided, leaving the computer and computer processor on its own to find structure in the inputs.

Added by DJM Jan 2024