Extensible Markup Language
For Artificial Neural Network
Glossary
Accretion

Approximation formed by combining information from several training patterns ( as in counter propagation). as opposed to interpolation between training patterns.

 
Activation

A node's level of activity : the result of applying the activation function to the net input to the node, Typically this is also the value the node transmits.

 
Activations function

A function that transforms the net input to a neuron into its activation . Also known as a transfer, or output, function.

 
ADALNE( Adaptive Linear Neuron)

Developed by Bernard Widrow, an Adaline''s output is +1 if the weighted sum of its input is greater than a threshold, -1 otherwise. The weights are calculated by the delta rule, which is also known as the Widrow Hoff rule [ Widrow Y Hoff. 1960].

 
Adaptive resonance theory ( ART)

Adaptive resonance theory is a quantitative explanation of learning and memory developed by Gail Carpenter, Stephen Grossberg and others. ART1 and ART2 are neural net architectures based on adaptive resonance theory. each of thess neural nets self-organizes the input data into categories with the variation allowed within a category depending on a user selected vigilance parameter . ART1 is used for binary input, ART2 for continuous input [Carpenter & Grossberg, 1987a, 1987b].

 
Algorithm

A computational procedure; a neural net training algorithm is a step by step procedure for setting the weights of the net .Training algorithms are also known as learning rules.

 
Annealing schedule

Plan for systematic reduction of temperature parameter in a neural network that uses simulated annealing.

 
Architecture

Arrangement of nodes and pattern of connection links between them in a neural network .

 
Associative memory

A neural net in which stored information ( patterns, or pattern pairs ) can be accessed by presenting an input pattern that is similar to a stored pattern. The input pattern may b an inexact or incomplete version of a stored pattern.

 
Autoassociative memory

An associative memory in which the desired response is the stored pattern.

 
Autoassociator

A neural net used to store patterns for future retrieval [Mc-Clelland & Rumehhart, 1988] . the net consists of a single slab of completely interconnected units , trained using the Hebb rule. The activation in this net may become very large . very quickly because unit''s connection to itself acts as a self-reinforcing feedback. See also Associative Memory. Brain-State-in-a-Box, and Hopfield Net.

 
Axon

fiber over which a biological neuron transmits its output signal to other neurons.

 
Backpropagation

A learning algorithm for multilayer neural nets based on minimizing the mean . or total , squared error.

 
Bias ( j )

the weight on the connection between node j and a mythical unit whose output is always 1 I,e, a term which is included in the net input for node j along with the weighted inputs from all nodes connected to node j.

 
Bidirectional associative memory (BAM )

A recurrent heteroassociative neural net developed by bart Kosko [ Kosko, 1988, 1992].

 
Boltzmann machine ( with learning )

A net that adjusts its weights so that the equilibrium configuration of the net will solve a given problem .such as an encoder problem [ Ackley, Hinton, Y sejnowski, 1985].

 
Boltzmann machine ( without learning )

A class of neural networks used for solving constrained optimization problems. In a typical Boltzmann machine , the weights are fixed to represent the constraints of the problem and the function to be optimized. the net seeks the solution by changing the activations ( either 1 or 0 ) of the units based on a probability distribution and the effect that the change would have on the energy function or consensus function for the net [Aarts 7 Korst, 1989] See also simulated annealing.

 
Bottom-up weights

Weights from the F1 layer to the F2 layer in an Adaptive resonance theory neural net.

 
Boundary contour system ( BCS)

Neural network developed by Stephen Grossberg and Ennio Miggolla for image segmentation problems [ Grossberg & Mingolla , 1985a 1985b].See also discussion by Maren (1990).

 
Brain-state-in-a-Box ( BSB )

Neural net developed by James Anderson to overcome the difficulty encountered when an auto-associator neural net iterates, namely the activations of the units may grow without bound . In the BSB neural net the activations are constrained to stay between fixed upper and lower bounds ( usually -1 and +1 ) [ Anderson. 1972] See also autoassociator.

 
<< Start < Prev 1 2 3 Next > End >>

Page 1 of 3
Copyright © 2010 xmlann.org is owned by Smart Neural Tech LTD Company registration Number 7026088, a limited company registered in England and Wales.