ABSTRACT
This chapter looks at learning algorithms that gradually increase in complexity and sophistication. The most basic algorithms, such as the various parameter modification techniques in the next section, are often not thought of as learning at all. The simplest learning algorithms are those that calculate the value of one or more parameters. A common way of understanding parameter learning is the “fitness landscape” or “energy landscape.” When an N-Gram algorithm is used for online learning, there is a balance between the maximum predictive power and the performance of the algorithm during the initial stages of learning. In online learning, however, it is common for the artificial intelligence to make decisions based on very sketchy information, so the confidence threshold can be small. The learning algorithm learns to choose the correct behavior given the observation inputs. Reinforcement learning is the name given to a range of techniques for learning based on experience.
