scikit-learn : Machine Learning Simplified

Chapter 3. Unsupervised Learning

Nowadays, it is a common assertion that huge amounts of data are available from the Internet for learning. If you read the previous chapters, you will see that even though supervised learning methods are very powerful in predicting future values based on the existing data, they have an obvious drawback: data must be curated; a human being should have annotated the target class for a certain number of instances. This labor is typically done by an expert (if you want to assign the correct species to iris flowers, you need somebody who knows about these flowers at least); it will probably take some time and money to complete, and it will typically not produce significant amounts of data (at least not compared with ...