Deep learning refers to an advanced type of neural network that is used in computer vision, self-driving cars and natural language processing. For example, deep learning enables facial recognition to be more accurate, and it allows medical scans to be interpreted without human analysis.
In the training phase of a deep learning model, the network is fed thousands of images of similar objects such as a car, truck, horse or human. The image is divided into blocks of pixels, and the blocks are transformed into patterns by an algorithm known as a "convolutional filter." The result of each transformation is handed to another filter and so on, each layer synthesizing the previous and adding a finer grain interpretation of the original input. These are the so called "hidden layers" between the input and output of a deep learning model.
The final result is a newly generated algorithm that is used to do the actual processing such as identifying an object or making a decision. The greater number of layers in the training phase, the more accurate the results.
Rather than programmed to solve a problem, the software is "trained" by experience. Because the algorithms in deep learning applications use intricate mathematics, pundits claim they may cause serious damage if used for life and death decision making in the future. See GAN
, neural network