What is Machine Learning?

Machine learning is a subset of artificial intelligence in the field of computer science that often uses statistical techniques to give computers the ability to "learn" information directly from data, without being explicitly programmed.

Machine learning can be categorized into three main types:
  • Supervised Learning (with label)
    • Regression
    • Classification
  • Unsupervised Learning (without label)
    • Clustering
    • Dimensionality Reduction
  • Reinforcement Learning


 Types of Machine Learning and Examples of Applications

Supervised learning is like the usual way that human learns when they are in kindergarten. It is a spoon-feeding learning mode. Everything is instructed by teachers with model answers and strict rules. Things taught are also simple and direct like the order of number and grouping of things. Similarly, in supervised learning, everything is stated clearly. For example, in a classification problem, the number of groups and what in which group are predefined. If the rule is not clearly stated, the kid will learn the things "wrong" and so as the machines. Thus, the quality of the input data will strongly affect the output.

Unsupervised learning involves more thinking. As no answers are given at the beginning. The computer needs to find out the hidden pattern and do the judgment based on the pattern. When doing clustering, if the training set has more distinct characteristics, it would be more easily for a machine to find out the correct algorithm to distinguish between different objects.

The first two types of machine learning are more commonly known and more easily understood, but reinforcement learning is now getting prevalent. On YouTube, there are more and more videos of training machines by using reinforcement learning.  This type of learning can be used to accomplished many purposes, playing games is an example that can be explained more easily in a human way. Imagined training machine-like training your dog, the reward will be given when the dog can perform the task, punishment or no reward when it fails to do. Through the response to the instructor, the machine can determine the dos and don'ts and only perform the dos. That is the main logic of reinforcement training. For instance, in training a machine to play Super Mario Bros, at first, the machine can only run, it will die when it faces obstacles, but after rounds of training, it will find out the time to jump, swim or crawl.

Comments