Deep learning allows computational models that are composed of multiple processing layers to learn representations of data with multiple levels of abstraction. These methods have dramatically improved the state-of-the-art in speech recognition, visual object recognition, object detection and many other domains such as drug discovery and genomics. Deep learning discovers intricate structure in large data sets by using the backpropagation algorithm to indicate how a machine should change its internal parameters that are used to compute the representation in each layer from the representation in the previous layer. Deep convolutional nets have brought about breakthroughs in processing images, video, speech and audio, whereas recurrent nets have shone light on sequential data such as text and speech.

– Y. LeCun, Y. Bengio, and G. Hinton [2015]

The previous chapter assumed that the input were features; you might wonder where the features come from. The inputs to real-world agents are diverse, including pixels from cameras, sound waves from microphones, or character sequences from web requests. Using these directly as inputs to the methods from the previous chapter often does not work well; useful features need to be created from the raw inputs. This could be done by designing features from the raw inputs using feature engineering. Learned features, however, are now state-of-the-art for many applications, and can beat engineered features for cases that have abundant data.

This chapter is about how to learn features. The methods here learn features that are useful for the tasks trained on, even though they may not have an interpretation that can be easily explained. Often features learned for some tasks are useful for other tasks.

Learning the features that are useful for prediction is called representation learning. The most common form of representation reasoning is in terms of multilayer neural networks. These networks are inspired by the neurons in the brain but do not actually simulate neurons. Artificial neurons are called units. Each unit has many real-valued parameters. Large artificial neural networks (in 2022) contain on the order of one hundred billion (${10}^{11}$) trained parameters, which is approximately the number of neurons in the human brain. Neurons are much more complicated than the units in artificial neural networks. For example, the roundworm Caenorhabditis elegans, which is about 1 mm long, has 302 neurons and exhibits complex behavior, which simple models of neurons cannot account for.

As pointed out by LeCun et al., above, artificial neural networks (ANNs) have had considerable success in unstructured and perception tasks for which there is abundant training data, such as for image interpretation, speech recognition, machine translation, and game playing. The models used in state-of-the-art applications are trained on huge datasets, including more cats than any one person has ever seen, more sentences than any one person has ever read, and more games than any one person has played. They can take advantage of the data because they are very flexible, with the capability of inventing low-level features that are useful for the higher-level task.

Artificial neural networks are interesting to study for a number of reasons:

• •

  As part of neuroscience, to understand real neural systems, researchers are simulating the neural systems of simple animals such as worms, which promises to lead to an understanding of which aspects of neural systems are necessary to explain the behavior of these animals.

• •

  Some researchers seek to automate not only the functionality of intelligence (which is what the field of artificial intelligence is about) but also the mechanism of the brain, suitably abstracted. One hypothesis is that the best way to build the functionality of the brain is to use the mechanism of the brain. This hypothesis can be tested by attempting to build intelligence using the mechanism of the brain, as well as attempting it without using the mechanism of the brain.

• •

  The brain inspires a new way to think about computation that contrasts with traditional computers. Unlike conventional computers, which have a few processors and a large but essentially inert memory, the brain consists of a huge number of asynchronous distributed processes, all running concurrently with no master controller. Conventional computers are not the only architecture available for computation. Current neural network systems are often implemented on parallel architectures, including GPUs and specialized tensor processing units.

• •

  As far as learning is concerned, neural networks provide a different measure of simplicity as a learning bias than, for example, boosted decision trees. Multilayer neural networks, like decision trees, can represent any function of a set of discrete features. However, the bias is different; functions that correspond to simple neural networks do not necessarily correspond to simple ensembles of decision trees. In neural networks, low-level features that are useful for multiple higher-level features are learned.