What Is Deep Learning, and Why Is It Everywhere?
Deep learning systems copy the human brain's neural networks to take machine learning a step deeper.
A deeper kind of machine learning
Deep learning is a subset of machine learning that relies on artificial neural networks and layers, thanks to which machines can infer more complex meaning from raw data. Let's break that down, shall we? Machine learning relies on algorithms and statistical models to help computers learn from large databases, and make decisions or issue predictions without having been explicitly programmed to do so. Deep learning is also a classification and learning method, but as the name suggests, it goes a tad deeper.
Artificial neurons and layered communication
Most deep learning models are based on artificial neural networks: a series of connected "nodes" that resemble biological neurons, processing information and transmitting signals to each other. These artificial neurons are aggregated into layers, and that's where the "deep" in deep learning comes from: the signal travels through all layers (there can be dozens of them), the machine gaining more understanding of the data each step of the way. Each layer uses the previous layer's output as input: for instance, in face recognition applications, the first layer might identify pixels and edges, the second a nose and eyes, the third hair, the fourth a face, etc.
Deep learning powers today's world
The first works on deep learning models started back in the 1980s, but the technology remained dormant until only a few years ago, when two conditions were finally met: Big Data happened, and computers got powerful enough to process it. In the years 2010, the advances were exponential.
One of deep learning's feats dates back to 2012, when Google Brain, after analysing ten millions of untagged YouTube screengrabs, learned by itself the concept of "cat": it taught itself to recognise cats without having been told that cats even existed.
Today, deep learning powers many applications, from image and speech recognition (used by personal voice assistants like Siri or Alexa) to medical image analysis and diagnosis, self-driving cars, fraud detection or customer relationship management. In short, it is at the core of many of today's most promising innovations. And now you know why.
Stay in the loop
Subscribe to our email newsletter. No spam, just occasional insight from our experts.