Asked By
piyushh
300 points
N/A
Posted on - 04/20/2011
What is Multilayer Perceptron? How we can train Multilayer Perceptrons?
Questions on Multilayer Perceptron or MLP
"Multi layer Perceptron" (MLP) is a concept of Neural networks.
Its basically an input/output model that uses mapping methods for setting inputs data onto a specific set of outputs. Its basically a 'Feedforward artificial neural network'. Here, Feedforward basically means that data is flowing in unidirectional i.e in one direction from the input towards output layer
A multi layer perceptron uses multiple nodes on layers mapped in a directed graph as such each layer is fully connected to the later one. Each node (except the input nodes) is a processing element known as neuron.
Your next question is how to train Multi layer perceptions.
This can be done by various training algorithms. The main
goal for training multi layer perceptron is to find the lowest point in the multi dimensional surface. The concept of error surface also helps in training MLPs, but its very complex. There are many characteristics such as local minima,saddle-points ,flat spots which are to be calculated. While performing it, the algorithm stops at the lowest point, which results in local minima.
Answered By
piyushh
300 points
N/A
#85704
Questions on Multilayer Perceptron or MLP
Hello,
Thanks for the Solution. Let me clear about the concept of Local Minima Please.
Questions on Multilayer Perceptron or MLP
Well basically in mathematics, Maxima and minima when taken collectively is known as Extrema. Local minima is an area which is above the 'Global minimum' and is lower than the 'Surrounding Terrain'.
Answered By
piyushh
300 points
N/A
#85707
Questions on Multilayer Perceptron or MLP
Can you explain me the concept of Error Surface?
Questions on Multilayer Perceptron or MLP
Yeah of course! We can say an 'Error surface' is a (n+1) dimensional surface area that represents the error terms of a network model based on n parameters.
Answered By
piyushh
300 points
N/A
#85710
Questions on Multilayer Perceptron or MLP
Okay. Thanks for the help.
