What is deep learning?
Reading time: 2 minutes
A few words about Deep Learning
A branch of Machine Learning, Deep Learning is a field of artificial intelligence that is increasingly used on a daily basis: facial recognition, image identification, conversation translation, etc.
Even if it is based on the same foundations as conventional artificial intelligence learning, it is a more advanced and in-depth field. Indeed, Deep Learning, an anglicism for deep learning, is a field of artificial intelligence that consists of programming a machine to teach it to perform specific tasks based on data and examples. The main idea is to create algorithms capable of getting closer to the functioning of the human brain thanks to artificial neural networks. To do this, Deep Learning uses an artificial neural network
While machine learning is already very powerful, deep learning is composed of a more complex mechanism. Thus, this technology allows processing larger quantities of data. This is why it is very relevant in fields such as finance, law and of course, health.
The mechanisms of deep learning.
This consists of using artificial neural networks, inspired by biological neurons. These are dozens or even hundreds of functions linked together, forming artificial neurons. The latter are divided into several layers, connected to each other. Each layer has its own task, its own objective to achieve. Thus, they are asked to interpret the information given by the previous layer of neurons in order to transmit it to the next one.
This is a chain to which each link provides information.
The deeper these neurons are, that is to say the more functions they contain, the more the machine is able to learn to perform complex tasks. (e.g.: identify people present in photos)
Special case of Deep Learning: imaging
In the particular case of medical image analysis, neural networks are convolutional.
Convolution is a mathematical operation. If it is highly compatible with the image, it is because it allows features to be extracted. It manages to translate the pixels of the image.
A convolutional neural network is a special case of artificial neural networks. This network is characterized by its first convolutional layers that apply convolution filtering of the input. The first layers detect the main attributes and the last layers, the most precise.
This is why this architecture is often used in medical image or video recognition.
