Term of the Moment
digital media
Definition: deep learning
An AI machine learning architecture often employed in "convolutional neural networks." Emerging in the 2010s, deep learning is used in all forms of AI such as computer vision, self-driving cars, natural language processing and chatbots. When the resulting system is used to do work, it has greater accuracy because the data in the learning phase were analyzed in greater depth. See convolutional neural network and AI in a nutshell.
The Layers
A simple neural network has one input, one "hidden" layer in the middle and one output layer. In a deep learning model, neural networks are trained using multiple hidden layers, each connected to the other with various constraints. For example, in image recognition, countless examples of objects, such as a car, truck, horse or human being, are input to the network, and the connections are refined as the example moves from one layer to the next. By the time the image reaches the final layer, its pattern has been recognized more thoroughly. If the training is "supervised," the training data were previously identified as a cat, horse, etc. In an "unsupervised" approach, the system has to figure out which items are similar.
Processing = Inference
The deep learning phase creates the "inference engine," which does the actual decision making such as identifying an object. The greater number of layers in the deep learning phase and the larger the number of examples, the more accurate the inference engine and the better the results. See DLA, machine learning, GAN, neural network and TensorFlow.
Algorithms Make Mistakes
Deep Learning in the Hierarchy
The Layers
A simple neural network has one input, one "hidden" layer in the middle and one output layer. In a deep learning model, neural networks are trained using multiple hidden layers, each connected to the other with various constraints. For example, in image recognition, countless examples of objects, such as a car, truck, horse or human being, are input to the network, and the connections are refined as the example moves from one layer to the next. By the time the image reaches the final layer, its pattern has been recognized more thoroughly. If the training is "supervised," the training data were previously identified as a cat, horse, etc. In an "unsupervised" approach, the system has to figure out which items are similar.
Processing = Inference
The deep learning phase creates the "inference engine," which does the actual decision making such as identifying an object. The greater number of layers in the deep learning phase and the larger the number of examples, the more accurate the inference engine and the better the results. See DLA, machine learning, GAN, neural network and TensorFlow.
Deep learning algorithms can be fooled. A famous example is mistaking a muffin for a chihuahua dog. This verification test is used to ensure the viewer is a human, and images such as these can confuse an algorithm. However, more advanced deep learning systems can, in fact, differentiate between them (see CAPTCHA).
Deep learning is a part of machine learning, which is a major category of artificial intelligence (AI).