| Abstract: |
There is a variety of patterns we desire to learn from radiological sciences (diagnostics and therapeutic) data. The previous chapters described how these various learning objectives can commonly be formulated in theoretical nomenclatures. This chapter introduces different conventional machine learning algorithms that could cater to readers’ specific learning goals. We intend to provide conceptual outlines of some of the widely used algorithms with minimal mathematical conundrum and examples drawn from the radiological sciences literature. In this chapter, we classify the algorithms into three types, based on the availability of information: unsupervised, supervised, and reinforcement learning. The methods illustrated in this chapter include principal component analysis and clustering (unsupervised), logistic regression, neural network, support vector machine, decision tree, Bayesian networks, and naive Bayes (supervised) in addition to reinforcement learning. © Springer Nature Switzerland AG 2022. |
