Neural Networks

From Verify.Wiki
Jump to: navigation, search

Neural Network or Artificial Neural Network (ANN) is a type of Artificial Intelligence that is inspired by the biological nervous system and its Biological Neural Network (BNN). The process of ANN is the input and output relationship that can be similar to the functions of a human brain. The ANNs are organized with the high number of layers of interconnected neurons where neurons can connect the input layer to the output layer. Functions can be set in the hidden layers between input and output layer, where they can provide a solution function to the given input. ANN can be used in data mining to complete highly complex relationships between inputs and outputs. Also finding patterns in data can be possible due to the non-linear statistical data modeling ability from ANN.

Historical background

Neural network simulations appear to be a recent development. However, this field was established before the advent of computers, and has survived at least one major setback and several eras.

Many importand advances have been boosted by the use of inexpensive computer emulations. Following an initial period of enthusiasm, the field survived a period of frustration and disrepute. During this period when funding and professional support was minimal, important advances were made by relatively few reserchers. These pioneers were able to develop convincing technology which surpassed the limitations identified by Minsky and Papert. Minsky and Papert, published a book (in 1969) in which they summed up a general feeling of frustration (against neural networks) among researchers, and was thus accepted by most without further analysis. Currently, the neural network field enjoys a resurgence of interest and a corresponding increase in funding.


The basic model behind ANN is based on mathematics models from the brain. The first layer is an input layer where the external information is received. The input layer and output layer are separated by one or more intermediate layers called the hidden layers. Each data input layer is transferred into the hidden layer, under which the input layers are combined forming single nonlinear output layer. The last or the highest layer is an output layer where the problem solution is obtained. However, the number of inputs provided for the function method and the number of outputs needed can be changed. The development of ANN can be created according to the users benefit, which depends on the number of inputs the user has and the outputs the user needs.

The diagram below shows the example of simple ANN characterized with three layers (input, hidden, output) of simple processing units.


Arrows from input to hidden layer and hidden layer to output layer indicate the strength of each connection and can be measured by a quantity called as weight. In the input layer and hidden layer, each unit processes it with a transfer function or activation function and lastly distributes the result to the output layer. [1]

Cases of Neural Networks in use

  • Google uses ANN for speech recognition purposes for their android products. [2]
  • Alyuda produces software products for data processing by the algorithm of ANN. [3]
  • Attrasoft provides sound recognition/retrieval, prediction and data mining by the use of ANN. [4]


  • Provides all possible outputs for a prediction variable. [5]
  • Can be used for character recognition. [6]
  • The ability to conduct vast number of inputs can make Image compression possible. [7]


  • 1942 - McCulloch and Pitts designed the first model of an artificial neuron which represents, even today, the basis for most ANNs. It consisted of:
  1. A set of inputs - (dendrites)
  2. A set of variable resistances - (synapses)
  3. A processing element - (neuron)
  4. A single output - (axon) [8]
  • 1943 - Warren McCllouch introduced the first steps of ANNs. [9]
  • 1949 - Donald Hebb developed the first learning rule: "When an axon of cell A is near enough to excite cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A’s efficiency as one of the cells firing B is increased. Cells that fire together wire together!” This has become the common way to calculate changes in connection strengths in a neural network. [10]
  • 1958 - First Neural Computer device was developed by Frank Rosenblatt called perception. [11]
  • 1980 - Multi-layered networks were introduced which include the number of multi inter-connected neurons acting as a function. [12]


  • Due to limitations of processors, ANNs are hard subject to learn which also includes difficulty of using complex non-linear models in general. Template:Reference needed
  • ANNs has the inability to detect data when it is needed for a reliable classification which can cause inaccurate data predictions. [13]


Date Author Tweet
January 18, 2016 @motherboard Busting the Hype behind deep neural networks:
January 18, 2016 @PythonWeekly Recognizing and Localizing Endangered Right Whales with Extremely Deep Neural Networks by @phelixlau #python
January 18, 2016 @readle The power of neural networks...
January 15, 2016 @BigDataBlender #Microsoft #NeuralNetwork Shows #DeepLearning Can Get Way Deeper #BigData #Analytics #AI #ML
January 8, 2016 @BradNeuberg Nice tutorial on attention mechanisms in neural networks:

Top 5 recent News Headlines

Date Title Link
January 18, 2016 Busting the Hype Behind Deep Neural Networks
November 13, 2015 An artificial neural network is learning how to use human language
November 3, 2015 Google's Neural Network Can Now Reply to Gmail Messages For You
October 13, 2015 How Google's neural network will improve YouTube
October 12, 2015 Artificial Neural Networks are changing the world. What are they?
  12. /

Verification history