Inline videos. See also:Category: Articles with embedded Videos..

Neural network

From Biocrawler, the free encyclopedia.

Simplified view of an artificial neural network

A neural network is an interconnected group of artificial or biological neurons. It is possible to differentiate between two major groups of neural networks:

Biological neural networks, for example the human brain or parts thereof.
Artificial neural networks originally referred to electrical, mechanical or computational simulations or models of biological neural networks. The field has expanded so that some applications do not clearly resemble any existing biological counterpart.

In modern usage the term most often refers to artificial neural networks, especially in computer science and related fields. There exist hybrids, incorporating biological neurons as part of electronic circuits, so there is not always a clear delineation.

Contents

1 Characterization

2 Comparison of biological and artificial neural networks

3 Types of artificial neural networks

4 History of the neural network analogy

5 See also

6 References

7 External links

[edit]

Characterization

In general, a neural network is composed of a group of connected neurons. A single neuron can be connected to many other neurons, so that the overall structure of the network can be very complex.

Artificial intelligence and cognitive modeling try to simulate some properties of neural networks. Approaching human learning and memory is the main interest in these models. These artificial neural networks are advantageous, especially in pattern recognition and classification tasks. They have found an application in the control of processes in the chemical industry, speech recognition, optical character recognition and adaptive software such as software agents (e.g. in computer and video games) and autonomous robots.

[edit]

Comparison of biological and artificial neural networks

Perhaps the most fundamental difference between brain and computer is that today's computers operate primarily sequentially, or with a small amount of parallelism (for details, see hyper-threading, SIMD, MMX and SSE2), while human brains are massively parallel. Given Turing's model of computation, the Turing machine (which shows that any computation that can be performed by a parallel computer can be done by a sequential computer), this is likely to be a functional, not fundamental, distinction. Furthermore, while a computer is centralized with a processor at its core, the question of whether the brain is centralized or decentralized (distributed) is unresolved.

Some other basic differences between neural networks in the brain and artificial neural networks follow. The brain is made up of a great number of components (about $1011$ ), each of which is connected to many other components (about $104$ ), each of which performs some relatively simple computation, whose nature is unclear, in slow fashion (less than a kHz), and based mainly on the information it receives from its local connections. In connectionism, neurons are connected randomly or uniformly, and all neurons perform the same computation. Each connection has associated with it a numerical weight. Each neuron's output is a single numerical activity which is computed as a monotonic function of the sum of the products of the activity of the input neurons with their corresponding connection weights. (Agre, 1997)

In some comparisons between the brain and computers, the following calculation is made: There are billions of neurons in the human brain; estimates differ and there are individual differences, some suggest about 2× $1012$ neurons. Since the relaxation time of these neurons is about 10 ms, this could amount to a processing speed of 100 Hz. The whole brain could therefore have a processing power of roughly 2× $1014$ logical operations per second. To compare, a 64-bit PowerPC 970 processor at a frequency of 3 GHz corresponds to 2× $1011$ logical operations per second, making the brain roughly one thousand times as powerful as a current high-end consumer PC. However, this comparison is extremely speculative. The working of biological neural networks is not well understood; it is not clear that anything like the "logical operations" performed by a computer actually occur in biological neural networks.

[edit]

Types of artificial neural networks

Feedforward neural networks

Recurrent neural networks

Hopfield network

Stochastic neural networks

Boltzmann machine

Modular neural networks

Dynamic neural networks

Alternatively, neural networks can be divided in two classes, supervised and unsupervised neural networks. Supervised neural networks, such as the perceptron, use a supervised learning algorithm, which means that input and output data is required during the training phase. The most common training algorithm is the backpropagation algorithm. On the other hand, unsupervised neural networks, such as the Kohonen Network, require only input data to be trained. They organise the input data themselves, according to a similarity metric.

[edit]

History of the neural network analogy

(main article: Connectionism)

This section is a stub. You can help by adding to it (http://www.biocrawler.com/w/index.php?title=Neural_network&action=edit).

The parallel distributed processing of the mid-1980s became popular under the name connectionism. In early 1950s Friedrich Hayek was one of the first to posit the idea of spontaneous order in the brain arising out of decentralized networks of simple units (neurons). A design issue in cognitive modeling, also relating to neural networks, is additionally a decision between holistic and atomism, or (more concrete) modular in structure.

[edit]

References

Agre, Philip E., et al. (1997). Comparative Cognitive Robotics: Computation and Human Experience. Cambridge University Press. ISBN 0521386039., p. 80

[edit]