Unsupervised learning and reinforcement learning
Richard E. Neapolitan, Xia Jiang · 2018
This chapter discusses two specialized types of learning: namely, unsupervised learning and reinforcement learning. Unsupervised learning concerns trying to find hidden structure in data. Researchers in automatic discovery endeavors to develop algorithms that discover properties or laws such as those in mathematics and science from data. The chapter presents an application of dynamic Bayesian networks to mobile target localization, which was developed by Basye et al. Dynamic Bayesian networks model relationships among random variables that change over time. So the latter can be used for reinforcement learning. The chapter illustrates how updating can be done effectively in such networks. It introduces dynamic Bayesian networks, which do model the temporal aspects of a problem. In the dynamic Bayesian network model, future actions are simply performed in some preprogrammed probabilistic way, which is not related to the sensor data.