上QQ阅读APP看书，第一时间看更新

Machine learning process

How do we learn? When we were studying in school or university, we were taught by our teachers. We learned from their teachings (training). At the end of the term, we needed to take a test (testing), which was basically to validate our knowledge. The scores we obtained decided our fate (evaluation). Usually, the evaluation was carried out by considering a threshold to pass (baseline). The scores determined whether we needed to retake the subject or were ready to move to the next level (deployment).

This is exactly how a machine learns as well. The words in the brackets are the terminology used by ML professionals. However, this is just one of the ways through which we, and the machines, learn. This is a typical supervised learning method. People sometimes learn from experience as well, and this is unsupervised learning. Let's study some more details about these learning methods.

Broadly we have two categories of ML algorithms as described earlier—supervised and unsupervised learning. There are a few other types, such as reinforcement learning, transfer learning, and semi-supervised learning, which are less often used and so are not in the scope of this book.