How Does Machine Learning Work?

by Carson

How do programs make decisions that are too complicated to be explained by their programmers? The programs are using machine learning, which is the process in which programs improve themselves based on data that they have collected, and is a type of artificial intelligence. Find out how it works in this article.

The Three Types of Machine Learning

Firstly, we will introduce the three types of machine learning, which help algorithms improve their accuracy.

  • Supervised learning
  • Unsupervised learning
  • Reinforcement learning

We’ll explain them one by one in the following parts of the article.

Supervised Learning

Supervised learning means that data labeled with associated outcomes or decisions are fed into the program. After that, the program will identify certain features or specialties of the data points so that they get classified into one group. Or they will try to discover relationships between the variables to identify the factors causing one data point to be associated with the output, making the algorithm more accurate.

Unsupervised Learning

In contrast to supervised learning, unsupervised learning means feeding unlabeled data to the algorithm to categorize the data points without human intervention. That way, the systems can detect the hidden patterns in the data points classified into the same group so that the algorithm can make decisions more correctly.

Reinforcement Learning

Reinforcement learning lets algorithms evaluate the results of each action it takes. If the results are positive after taking action, it may use the same method the second time. However, if the results are not desirable, the algorithm will try to learn about what has gone wrong and improve itself in related aspects. This is the method of trial-and-error and is often used in programming.

Neural Networks

But how about neural networks? A neural network consists of one input layer, one or more hidden layers, and one output layer. The input layer accepts data, the hidden layers process the information using various algorithms, and the output layer is responsible for the output. Deep learning is a type of machine learning whose neural network consists of more than one hidden layer.

How about neurons in machine learning? A neuron is a unit that accepts input, processes it, and produces an output. The neurons will then pass the data it produces to the next level of the neural network. The output will reach another neuron and process it further, or be compiled as the output of the neural network in the output layer.

An illustration about the input layer, hidden layers, and the output layer of a neural network
Image created using Canva

Methods for Algorithms to Learn

After knowing the types of machine learning and a fundamental concept known as neural networks, we will have to introduce some methods algorithms can use to improve their accuracy in classifying data and making decisions.

The first method is known as clustering. Clustering algorithms automatically identify features of certain data points in a database and classify them into groups so that the database is ready for further processing. For instance, if an algorithm is comparing photos between Jupiter and Saturn, the magnificent ring of Saturn is one of the main features to notice. Clustering is predominantly used in unsupervised learning, where information fed to the algorithms is not labeled.

Another method is known as regression. Regression algorithms use mathematical functions to produce a continuous output from a few inputs. Regression algorithms are mainly classified into linear regression and nonlinear regression. Each category contains a few more main types of regression algorithms, which we won’t explain because it could make the article too long. Regression algorithms are primarily used in supervised learning, where data points are labeled. These algorithms can be installed in one of the neurons in a neural network since it accepts input, do some calculations on them, and produces an output.


In this article, we’ve mentioned the three main types of machine learning, a model known as neural networks, and two methods for algorithms to understand what’s going on with these data points to produce more accurate results next time. If you’d like to learn more, please visit the pages mentioned in the references below.

References and Credits

  1. (2020, August 19). What is Supervised Learning? Retrieved December 1, 2021, from
  2. TED-Ed. (2021, March 12). How does artificial intelligence learn? – Briana Brownell. Retrieved December 1, 2021, from
  3. (n.d.). Unsupervised Learning and Data Clustering. Retrieved December 1, 2021, from
  4. (2020, September 21). Unsupervised Learning. Retrieved December 1, 2021, from
  5. Amal Joby. (2021, March 19). Unsupervised Learning: How Machines Learn on Their Own. Retrieved December 3, 2021, from
  6. Shweta Bhatt. (2018, March 19). Reinforcement Learning 101. Retrieved December 3, 2021, from
  7. Gavril Ognjanovski. (2019, January 14). Everything you need to know about Neural Networks and Backpropagation — Machine Learning Easy and Fun. Retrieved December 3, 2021, from
  8. Victor Zhou. (2019, March 3). Machine Learning for Beginners: An Introduction to Neural Networks. Retrieved December 3, 2021, from
  9. Nick McCullum. (2020, June 28). Deep Learning Neural Networks Explained in Plain English. Retrieved December 3, 2021, from
  10. (2020, February 10). What is Clustering? Retrieved December 3, 2021, from

Related Posts

Leave a Comment

* By using this form you agree with the storage and handling of your data by this website.