What is machine learning (ML)? Machine learning is the use and development of algorithms and statistical models to teach computer systems to analyze and discover patterns in data, without being explicitly programmed. It is part of artificial intelligence (AI). Machine learning is a field devoted to understanding and building methods that let machines use data to improve computer performance on some set of tasks. It’s the art and science of programming computers to learn from data. As machines are exposed to more and more data, they improve their performance.

In order to teach a computer to do a task, it’s provided with examples of how to do that task. Statistics are used to analyze that data and find patterns related to that task.

Back in the 1990’s, the spam filter arrived. It was the first machine learning algorithm that became mainstream to improve the lives of millions of people. The spam filter uses emails flagged by users as spam and also uses unflagged emails (called “ham”) to learn how to spot spam.

Machine learning algorithms build a model based on sample data, known as training data, in order to make predictions or decisions without being explicitly programmed to do so. Machine learning evolved from the study of pattern recognition and computational learning theory in artificial intelligence (AI). Sometimes people use the term “complex model”. In the Google Advanced Data Analytics Professional Certificate program at Coursera, when complex model is used its referring to mathematical or computational models in general, inclusive of everything from regression to deep learning.

Types of Machine Learning

There are three types of machine learning. There is supervised learning, unsupervised learning, and reinforcement learning. Other sources of information I’ve found say that there are two types: supervised and unsupervised. Another source said that there are four types: supervised, deep, reinforcement and unsupervised.

With supervised learning our objective is to predict a target variable given a set of observations. Our two learning tasks are classification and regression. One example would be predicting real estate prices. Generally we have a large set of historical observational data we can use for training, as well as the associated target variable.

With unsupervised learning the machine is not given labelled data. It is asked to find patterns on its own. The objective is different. Here we are organizing data by some inherent structure. Inherent means that it is existing in something as a permanent, essential, or characteristic attribute. We have past data but we don’t have the associated target variables. Unsupervised learning could be used for market segmentation where there are no agreed-upon segments. We want to define segments based on some pattern or common set of characteristics. Another example is fraud detection.

Reinforcement learning a type of machine learning where the computer learns through trial and error. It learns strategies through interactions. A task is given and and a reward system is used that tells the computer how well its doing. It’s programmed to maximize its rewards. Another objective may be to achieve a goal. Here we can teach a computer to play a game of chess or checkers. Another example would be AlphaZero or autonomous vehicles.

Here’s an example of how we might understand the difference between unsupervised and supervised learning. Let’s discuss supervised learning. Our model will recognize apples from pictures of apples. We have a set of pictures of apples and a label that says ‘this is an apple’. We build a model so that the computer can recognize new pictures of apples, even if it hasn’t seen that picture or apple type before. With unsupervised learning we just have a bunch of pictures of fruits and we want to develop a model that will see the differences between the fruits and group them together.

Learn with Books

Hands-On Machine Learning with Scikit-Learn, Keras & TensorFlow, Third Edition by Aurelien Geron, published by O’Reilly in 2023. It rates very well at amazon.com with a 4.6 out of 5. The book’s project code is publicly available at Google Colab. There is one notebook per chapter plus a few extra notebooks and tutorials.