As new technologies are created, introduced and then developed further, new and sometimes hard to follow buzzwords come as part of the package. Artificial intelligence was no different. It brought about the term machine learning, followed shortly after by deep learning and reinforcement learning, which we have to wrap our heads around.
All of the aforementioned fall under the same umbrella of sci-fi inspired uses of the computer, known as AI. In this article, we focus on the two subsets of machine learning, which are deep learning and reinforcement learning.
What Is Deep Learning?
Deep learning is a subset of machine learning that dates as far back as the 1950s. Also sometimes referred to as Deep Structured Learning, is a self-teaching computer system that makes use of synthetic neural networks to sift through information, thus enabling an AI to recognise and identify sound or objects in an image.
This is achieved through a lengthy process of feeding sounds or images into a deep learning algorithm.
How Does It Work?
Deep learning uses artificial neural networks that are designed to mimic the way the neural networks of the human mind work, this results in a learning capability that is significantly greater than that of standard machine learning.
What Is Reinforcement Learning?
In much the same way as Deep learning, reinforcement learning is another branch of the artificial intelligence tree. This form of machine learning can be thought of as a program that self-educates through a process of trial and error. In fact it is very similar to how we learn.
The program will perform tasks and analyse the feedback data in order to learn what works and what doesn’t. It will then reinforce (tweaking and modifying) the actions that returned the best results. This is how Google’s DeepMind AI was taught to play classic Atari games.
How Does It Work?
Reinforcement learning algorithms are trained to complete tasks, like playing a video game, through a process of acting and then observing the reaction, with the purpose of achieving a goal. This objective is usually referred to as the reward and the aim of the AI is to maximise the reward.
In essence, it is a process of figuring out the most effective means of reaching a goal in a specific environment.
Differences Between Deep Learning & Reinforcement Learning
Though deep learning and reinforcement learning fall under the same umbrella and the latter does borrow a few concepts from the former, the AI disciplines are in fact distinct from one another.
The distinction between the two lays in the learning process and the application of that learned data. Deep learning algorithms are trained on one data set and then taking what the program has learned and applied it to another set of data. This makes deep learning agents better suited to tasks like identifying images and audio.
Reinforcement learning is a dynamic process of learning through adjustments to actions taken, and then reviewing the feedback to increase the reward.