deep learning (deep neural networking)

This definition is part of our Essential Guide: Special Report: Artificial intelligence apps come of age

Deep learning is an aspect of artificial intelligence (AI) that is concerned with emulating the learning approach that human beings use to gain certain types of knowledge. At its simplest, deep learning can be thought of as a way to automate predictive analytics

While traditional machine learning algorithms are linear, deep learning algorithms are stacked in a hierarchy of increasing complexity and abstraction. To understand deep learning, imagine a toddler whose first word is “dog.” The toddler learns what is (and what is not) a dog by pointing to objects and saying the word “dog.” The parent says “Yes, that is a dog” or “No, that is not a dog.” As the toddler continues to point to objects, he becomes more aware of the features that all dogs possess. What the toddler does, without knowing it, is to clarify a complex abstraction (the concept of dog) by building a hierarchy in which each level of abstraction is created with knowledge that was gained from the preceding layer of the hierarchy. 

Computer programs that use deep learning go through much the same process. Each algorithm in the hierarchy applies a non-linear transformation on its input and uses what it learns to create a statistical model as output. Iterations continue until the output has reached an acceptable level of accuracy. The number of processing layers through which data must pass is what inspired the label “deep.” 

In traditional machine learning, the learning process is supervised and the programmer has to be very, very specific when telling the computer what types of things it should be looking for when deciding if an image contains a dog or does not contain a dog. This is a laborious process called feature extraction and the computer’s success rate depends entirely upon the programmer’s ability to accurately define a feature set for "dog." The advantage of deep learning is that the program builds the feature set by itself without supervision. This is not only faster, it is usually more accurate.

machine learning vs deep learning

Initially, the computer program might be provided with training data, a set of images for which a human has labeled each image “dog” or “not dog” with meta tags. The program uses the information it receives from the training data to create a feature set for dog and build a predictive model. In this case, the model the computer first creates might predict that anything in an image that has four legs and a tail should be labeled "dog." Of course, the program is not aware of the labels “four legs” or "tail," it will simply look for patterns of pixels in the digital data. With each iteration, the predictive model the computer creates becomes more complex and more accurate.

Because this process mimics human thought, deep learning is sometimes referred to as deep neural learning or deep neural networking. Unlike the toddler, who will take weeks or even months to understand the concept of “dog,” a computer program that uses deep learning algorithms can be shown a training set and sort through millions of images, accurately identifying which images have dogs in them within a few minutes. 

In order to achieve an acceptable level of accuracy, deep learning programs require access to immense amounts of training data and processing power, neither of which were easily available to programmers until the era of big data and cloud computing. Because deep learning programming is able to create complex statistical models directly from its own iterative output, it is able to create accurate predictive models from large quantities of unlabeled, unstructured data. This is important as the Internet of Things(IoT) continues to become more pervasive, because most of the data humans and machines create is unstructured and is not labeled. Use cases today for deep learning include all types of big data analytics applications, especially those focused on natural language processing (NLP), language translation, medical diagnosis, stock market trading signals, network security and image identification.

Here is a very simple illustration of how a deep learning program works. This video by the LuLu Art Group shows the output of a deep learning program after its initial training with raw motion capture data. This is what the program predicts the abstract concept of "dance" looks like.

With each iteration, the program's predictive model became more complex and more accurate.

This was last updated in June 2016

Continue Reading About deep learning (deep neural networking)



Find more PRO+ content and other member only offers, here.

Join the conversation


Send me notifications when other members comment.

By submitting you agree to receive email from TechTarget and its partners. If you reside outside of the United States, you consent to having your personal data transferred to and processed in the United States. Privacy

Please create a username to comment.

What other uses cases for deep learning do you predict?
If the program requires a man-made training set the use is still limited. It's the ability to analyze broad spectrum of information and extract patterns that opens broad use.
The difference is, people can do this job without knowing what they looking for.
I signed up for Amazon Mechanical Turk and picked a HIT about image recognition and was still shown an example of what I was to look for.  And yes, I had to use my own brain to create a feature extraction for the object I was tasked with finding in images. That's why deep learning is also referred to as neural networking -- the computer program is doing what I did, only much faster and probably more accurately.  (I missed a few in my HIT!)
This can be very useful in politics. On the long run people should have enough faith in machine learning outputs that most of us would be willing to follow a decision taken by a machine instead of a politician.
While the human input on political and national matters will always be important, it can be effective in the second stage of the decision making process, but the fact remains that a machine based system that produces an output based on accurate data and algorithms will be more sound than a human with an ideological background.
It's been interesting watching the pollsters who used deep learning algorithms to predict election results in the US try to figure out where they went wrong. Bad data or bad programming? 
In all, a very good article.

I'd like to pick on one particular comparison that was made.

"Unlike the toddler, who will takes weeks or even months to understand
the concept of “dog,” a computer program that uses deep learning
algorithms can be shown a training set and sort through millions of
images, accurately identifying which images have dogs in them within a
few minutes. "

This is a linear comparison. Toddler's brain develops understanding and abilities of countless things in parallel. Human in general, and children in particular are also capable of developing meta-knowlegde, draw conclusions, form abstractions - and ask questions. So while the program ends up with image recognition, toddlers will learn about cats, horses, and bees. They will also ask "why" millions of times. 
You're right. To replicate real intelligence, deep learning will need to be used in tandem with many other approaches to replicating human thinking.

This might be a little out of context, just wanted to share my views on deep learning though.

We as humans without an exception know what is the right thing to do and we still wont do it because we have our own apprehensions. Now a machine eventually will figure out the right way to do certain things , but my question is "Will humans limit its ability to just learn and not to act?"


File Extensions and File Formats

Powered by: