Category: 

What is a Neural Net?

Article Details
  • Written By: John Lister
  • Edited By: Bronwyn Harris
  • Last Modified Date: 22 October 2017
  • Copyright Protected:
    2003-2017
    Conjecture Corporation
  • Print this Article

A neural net is an artificial replication of a biological network of neurons. These are the nerve cells which are the basis of the nervous system in animals. A neural net attempts to simulate this network for purposes such as finding out more about biology, or for working on artificial intelligence.

To distinguish between the two networks, those which exist in real animals are often known as biological neural networks. These consist of neurons, which are the basis of the way the brain, spinal cord, and nerve endings work together. These include sensory neurons which detect inputs from the five senses, and motor neurons which cause muscles to contract and thus create movement in the body.

The phrase neural net is usually reserved for artificial simulations of the neural networks. Generally they are designed to simulate the human system. As difficult as it may be to imagine, even the most advanced computers can only carry out a considerably simplified simulation of the human brain, which has around 100 billion neurons. One of the most complicated simulations ever attempted involved two million artificial neurons in an attempt to simulate a cat’s intelligence. Even this modest task proved too ambitious in practice.

Ad

A neural net is effectively a computer program which uses a network of artificial neurons. These are individual elements which each carry out a particular task at a particular moment. As in the brain, the artificial neurons do not have a permanent dedicated task. Instead, the network automatically divides a task into its smallest possible parts and has the artificial neurons each work on a part simultaneously. This means the network must adapt the way it divides the work among the neurons to the particular demands of a task.

There are two main uses of neural nets. The first, commonly known as cognitive modelling, looks at how the network works rather than what it produces. By creating increasingly complicated neural nets and setting them to a variety of tasks, researchers aim to see how they operate and adapt to new requirements. The hope is that this will help explain more about how biological neural networks operate.

The second use is known as artificial intelligence and is concerned more with using a neural net to carry out a particular task. This usually involves trying to make computers which can carry out tasks which humans can do, but computers usually struggle with. These are situations where the sheer processing power of a computer is not sufficient in itself because the problem requires a different approach to the way computers normally work, that being the approach used by the brain. Examples of such problems include recognizing images or voices.

Ad

Recommended

Discuss this Article

Post your comments

Post Anonymously

Login

username
password
forgot password?

Register

username
password
confirm
email