In this post I’d like to cover some of the different training approaches in Machine Learning, Deep Learning, and Artificial Intelligence to give you a practical guide when selecting the best approach depending on your goals and desired outcomes.
Here’s a short list with a brief overview followed by a couple of concrete examples.
- Supervised learning is when you are training an algorithm and you have access to both the input and output data to test with.
- Unsupervised learning is when you have only the input data available and you desire the algorithm to find meaning in the data.
- Semi-supervised learning is similar to Supervised learning in the sense that you have both the input and output data to test the algorithm however, it is often the case where you have much more input than output data available for testing.
- Reinforcement learning is related to psychological training with people and animals where the subject is given a reward or punishment for successfully or unsuccessfully completing a specific task.
- Supervised learning — learning the names of the US capitals, names of colors, or games like selecting the odd word out in a list.
- Unsupervised learning — image recognition like identifying handwritten characters, facial recognition, and relationships in sales data.
- Semi-supervised learning — identifying and grouping each subject’s dialog in a recording, removing background noise, identifying the name of a song in a noisy room or selecting specific information like the name of a person or place in a newspaper article.
- Reinforcement learning — games such as Checkers, Chess, and Go as well as in robotics like self-flying drones, robotic arm movement and navigation. The outcome of a game might result in gaining or loosing a point where the negative feedback in robotics might be from bumping into an obstacle. As you can see there isn’t a single approach that is better or worse than others but rather selecting and testing multiple approaches until finding the preferred approach depending on your specific goals and use cases.
This article just touches the very surface of the approaches or combinations of approaches available. Additional details can be found at the following references below.
References & Additional Reading:
- A Tour of Machine Learning Algorithms by Jason Brownlee
- Supervised and Unsupervised Machine Learning Algorithms by Jason Brownlee
- What is the difference between supervised and unsupervised learning algorithms?
- Is reinforcement learning the combination of unsupervised learning and supervised learning?
- Lecture 4 — Supervised Learning by Andrew Ng
- Lecture 5 — Unsupervised Learning by Andrew Ng
- Reinforcement Learning by Alex J. Champandard
As always, if you have questions please leave them in the comments or reach out over Twitter.