There are three types of machine learning based on the nature of the learning feedback available to a learning system.
- Supervised learning:
- Unsupervised learning:
- reinforcement learning:
In supervised learning, the computer is trained on the basis of sample input data and the desired results obtained from the “master”. It is widely used in real applications:
- Face recognition
- Speech recognition
- Sales Forecasting
- Recommendation Apps/tools
Further Classification of Supervised Learning
Unsupervised learning is used to detect anomalies, outliers, such as cheating or faulty equipment, or to group consumers with similar behavior as part of a sales campaign. It is the opposite of supervised learning.
There is no labeled data here. When learning data contains only some indications without any description or labels, it is up to the coder or the algorithm to find the structure of the underlying data, to discover hidden patterns, or to determine how to describe the data.
This kind of learning data is called unlabeled data. Suppose we have several data points and want to divide them into several groups.
We may not know exactly what the classification criteria would be. Thus, the unsupervised learning algorithm attempts to classify a given set of data into a certain number of groups in an optimal way.
Unattended learning algorithms are extremely powerful tools for analyzing data and identifying patterns and trends. They are most often used to group similar inputs into logical groups. Unattended learning algorithms include Kmeans, Random Forests, Hierarchical Grouping and so on.
If some learning samples are marked and some others are not marked, this is semi-supervised learning. Uses a large amount of unlabeled data for training and a small amount of data marked for testing.
Partially supervised learning is used in cases where obtaining a fully labeled data set is costly and labeling a small subset is more practical.
For example, it often requires qualified experts to mark certain images by remote sensing, and many field experiments to locate oil at a specific location while obtaining unlabeled data is relatively easy.
Here, learning data gives feedback, so that the system adapts to dynamic conditions to achieve a specific goal. The system assesses its performance based on feedback and responds accordingly. The most known cases are self-propelled cars and the AlphaGo chess algorithm.