Original article was published by PENINAH GATHONI on Artificial Intelligence on Medium
Types of Machine Learning
There are three types of Machine Learning:
- Supervised learning
This is a method of learning in which we train the machine using data which is well labeled. The training and output data is clearly labelled .
The main objective of supervised learning algorithms is to learn an association between input data samples and corresponding outputs after performing multiple training data instances.
An example would be training the machine with labelled images of cats and dogs , in short we present the image of a cat to the machine and tell it that that’s how cats look, same to the dog images. The machine then learns the difference between the two and we can now present it with unlabelled new images of cats and dogs and expect it to correctly label these images.
Supervised learning can be grouped into two categories of algorithms:
2. Unsupervised Learning
This is training the machine with unlabelled data and letting the algorithm classify the data into similar clusters without really knowing what they are. It is unsupervised in the sense that we do not have a supervisor to provide guidance as was the case with data labels .
Unsupervised learning algorithms come in handy where we do not have pre-labelled training data and we want to extract useful patterns from input data.
To use the example given above , we will feed the machine with unlabeled images of cats and dogs . It will identify similar images by their prominent features e.g size of the ears, face structure etc and henceforth classify the images into two clusters. Subsequently , when given input data, it will check for it’s features to determine to which cluster the image belongs.
Unsupervised learning can be categorized into two categories of algorithms:
3. Reinforcement Learning
This is a reward based method of learning in which an agent is placed is placed in an environment , explores the environment and learns by performing actions and observing the rewards that it gets from those actions.
The agent gets a reward for each right action and a penalty for each wrong action. With this feedback, the agent will automatically learn and improve it’s performance over time .
One of the applications of Reinforcement learning is in the self driving cars.
The table below summarizes the above 3 types showing their differences