The Machine Learning Process Applied on Machine Learning Problems
Explore how the five-step machine learning process helps solve complex pattern recognition problems. Understand how data preparation, model training, prediction, and evaluation work together across different domains including image tagging, music recognition, and language translation.
Let’s review machine learning (ML) problems and the ML method of solving them. This first chapter has just these two objectives.
The ML problem
The kind of problems that require some patterns to be identified, such that those patterns do not work for only one instance of the problem, but generally work for the whole class of instances, are known as ML problems. Classification is one such ML problem. For example, we, as humans, do not memorize in a picture-perfect way what one specific spiral galaxy looks like. Instead, we identify some specific patterns which help us distinguish between spiral and other classes of galaxies, after looking at some shots of spiral galaxies.
The ML process
Any complex problem that requires the computer to be able to identify patterns involves an ML process. We saw the following process pipeline for the classification of spiral galaxies:
