Machine Learning (ML) has become popular these days. But what is machine learning? Machine Learning is a subset of Artificial Intelligence. (Read the difference between ML and AI). Machine Learning is set of computer algorithms that helps to make predictions without being explicitly programmed. What does that mean?
So, What is Machine Learning?
In traditional programming, we enter rules and data to the computer and get answers. Programmers should figure out the rules to get the expected answers. In ML, we feed data and answers to the computer and we let computer to figure out the rules. This process is called training. After the training, the ML algorithms build a mathematical model based on the training data we provided. This model can be used to predict or make decisions.
There are four main types of ML methods;
- Supervised Learning
- Unsupervised Learning
- Semi-Supervised Learning
- Reinforcement Learning
Supervised Learning
In Supervised Learning, we provide training data with their desired output (i.e. labels). So we train the model using that data until it predict the answers accurately. This is the most common method used. This is used for image recognition, object detection, face detection, voice recognition etc. Most common example for this is to teach a machine to recognize and categorize emails as spam or non-spam.
The results of this type of algorithms are usually marked as classification or regression. Classification algorithms finds a way of mapping equal elements in specific categories, as in the example above, “spam” and “non-spam”. Regression models focus on
identifying a trend for data that even allows predicting the future based on historical data.
Unsupervised Learning
In unsupervised learning, we provide training data only. The ML algorithms try to find clusters of data based on their similar features. In supervised learning we use structured data, but in unsupervised learning we don’t use structured data. It is up to the ML algorithm to find hidden structures in the data.
Unsupervised learning can be used in recommendation systems. It is widely used in e-commerce stores, to recommend products that may be interesting to the customer who visits the site. Similarly, there are streaming services that suggest music or videos to the user based on user preferences. YouTube, Netflix and Amazon use this approach to create their recommendation algorithms.
Semi-Supervised Learning
Semi-supervised learning is a hybrid approach that combines elements of both supervised and unsupervised learning. In this technique, the model is trained on a dataset that consists of a small amount of labeled data and a large amount of unlabeled data. The labeled data is used to guide the learning process, while the unlabeled data is used to improve the model’s generalization ability.
By leveraging both labeled and unlabeled data, semi-supervised learning can often achieve better performance than supervised learning alone, especially when labeled data is scarce or expensive to obtain.
Reinforcement Learning
In reinforcement learning, we let the computer to interact with a specific environment and perform actions. When the computer perform correct action, we reward it, if not we punish it. Well, the goal of the ML algorithm is to collect maximum rewards. This is like trial and error training. This approach is commonly used in game development. AI agents like AlphaGo which defeated world best Go players are trained using reinforcement learning.
Reinforcement learning was inspired by behavioral psychologists, who believed in the effectiveness of rewards and punishments in the education of human beings. It also recalls the procedure for training domestic animals. It is, therefore, a method based on building experience. From this, the algorithm knows which paths are better than others and which processes are more agile.
This is a very brief introduction to machine learning. We’ll learn how to use these methods in the future. Meanwhile, you can download the machine learning algorithms cheat sheet I have created.
