Difference between deep learning and machine learning?

You hear it all over the tech world. We have entered the “era of big data”. You hear tearms like “AI”, “Machine Learning”, “Deep Learning”, and “Neural Networks” getting thrown around in forums and on tech news sites.

What does it all really mean? There’s too much to say on the subject of AI and machine learning for a single article, so let’s start small. Let’s look at the difference between “machine learning” and “deep learning”. What are they? Do they overlap?

Let’s start with machine learning. Machine learning is actually a branch of computer science: it is essentially science (some would say art) of getting computers to perform a given task well without giving explicit instructions on how to do the task. In short, it’s about developing techniques that allow the computer to find patterns in input data and make inferences about those patterns.

Most work in machine learning is focused on finding efficient algorithms which can take a set of input data and produce from them a set of outputs (“right answers”) based on patterns identified in the input data. For instance, you might give your algorithm input data about sale prices for homes in Melbourne last year, accompanied by data about each house sold, such as its ZIP code, size in square feet, number of bedrooms, and so on. Once you had “trained” your machine learning model on this data, say by providing it with all information *except* house prices and having it learn to predict them, then you could use this model to accurately judge the sale prices of homes that the system had not seem before.

Where does “deep learning” fit into all this? Deep learning is just one of the many algorithms that has come from the work done in machine learning. You can think of “deep learning” as being one of many tools in machine learning’s toolbox.

Specifically, deep learning is a neural network based learning technique which uses neural networks composed of many layers of artificial neurons. The “deep” in deep learning refers to the fact that deep learning’s neural networks have many layers. The layers are what makes deep learning, “deep”. The moniker “deep” is often misinterpreted to mean that deep learning is somehow more insightful or powerful than other machine learning techniques. Although it’s a very powerful and flexible technique, it is not necessarily better than the other tools machine learning gives us: it’s one option among many.