top of page

What Is Deep Learning (DL)

Deep Learning (DL) is an approach to Machine Learning (ML) that represents a vast group of techniques. Using different types of architectures and algorithms, DL is able to make sense of data on a relatively higher level than most other ML approaches. Although it has been in existence since the 1940s, Deep Learning has recently become a very popular area of AI R&D.


Deep Learning is a machine learning approach that learns by example from data. In the introduction to Ian Goodfellow, Yoshua Bengio, and Aaron Courville’s “Deep Learning” textbook, it is quoted as:

“To allow computers to learn from experience and understand the world in terms of a hierarchy of concepts, with each concept defined through its relation to simpler concepts.”
"The hierarchy of concepts allows computers to learn complicated concepts by building models out of simple ones. Deep Learning is thus able to solve intuitive problems with many, many layers." Deep Learning- Ian Goodfellow, Yoshua Bengio, & Aaron Courville

Deep Learning is a kind of feature learning, or Representation Learning. Representation Learning algorithms are able to learn features from raw data. Representation Learning uses machine learning to discover not only the mapping from representation to output but also the representation itself. Simply put, representation algorithms “learn how to learn”. RL algorithms discover a quality set of features from unstructured data (such as images, video, audio, and digital text) and use such features to complete a task.

This makes Deep Learning quite valuable, for it is a form of ML that can process nonlinear relationships through many layers. Many ML algorithms are limited in mapping linear relationships, making DL a more advanced approach to accomplishing advanced AI.


Deep learning is mostly associated with Deep Artificial Neural Networks (ANNs). Deep ANNs have more than one hidden layer, and thus their designated name for “depth”. ANNs are inspired by the architecture of biological neural networks (AKA the human brain). Artificial Neural Networks are the essential models used to construct DL algorithms. An Artificial Neural Network (or Multilayer Perceptron) is a model that maps a set of input values to output values. Below is diagram illustrating the architecture of a deep neural network:

Each purple node represents an artificial “neuron”, just like the biological neurons of the human brain. Neurons send and receive input signals from each other. In the process of forward propagation, signals from the input layer (x) flow forwards to the hidden layers, where the output (y) is produced in the output layer. Each neuron also has an assigned weight. In order to send input signals to get the output (y), a neuron adds up all the values of its input signals and multiplies each by its associated weight (w).

Once this is done, the ANN outputs a prediction, which is compared to the target output. If the prediction is not equal to the expected output, the information flows backwards to figure out how to minimize the cost function. This process is called back-propagation.


Deep Learning has contributed many interesting applications to a wide span of industries.

  • Autonomous Driving- autonomous driving is one of the most exciting areas of AI at this current time. DL technology has contributed an important part to self-driving progress. DL methods are being used for detecting and recognizing objects in images and from video cameras. Neural networks such as CNNs are being used for path planning and to detect traffic, pedestrians, and other vehicles to improve the safety standards of autonomous driving. For in depth research on DL applications to Autonomous Driving, refer to:


  • Aerospace/Defense- According to businesswire, the U.S. Aerospace and Defense industry has invested about $245.8 million in AI and is planning to increase spending to $2,027.6 million by the year 2025. Deep Learning AI technology has contributed to numerous areas of the Aerospace industry, including industrial robotics, crisis-management, auto-piloting, predictive maintenance and machine control, data security, weather forecasting, air traffic movement optimization, threat detection, etc.


  • Medicine/Healthcare- AI has been a generous contributor to aiding the industries of medicine and healthcare. Certain DL algorithms have been applied to help medicinal research and explore new areas of health. For Example: Using image recognition, DL algorithms can detect atrial fibrillation earlier, or spot early signs of cancer. It can accurately measure the doses of medicine, or help health professionals in the general management of patients.

-"What Has AI In Medicine Ever Done For Us"

- "Top 5 Use Cases Of Artificial Intelligence In Medical Imaging"


These are only a few examples of where Deep Learning AI technology is being applied in the real world. There are numerous ways that DL is currently being used in the sectors of security, finance, international travel, retail, etc, but such listings would take up the entirety of this blog post and more! In many areas, Deep Learning algorithms are making it easier, faster, and cheaper to perform tasks that originally were too challenging or tedious for humans to perform by themselves. With this in mind, there is reason for caution. In this subset of AI, it is vital to consider the fact that Deep Learning could take up most of the roles that humans currently occupy in the near future.


*Other sources listed below real world application examples*

bottom of page