Multilayer Feed Forward Neural Networks samples
Function Approximation [Download]
The sample demonstrates usage of multilayer feed forward neural network on the sample of 1D function approximation.
During learning process a set of specified points are given to network - network is trained to provide desired function's
value for the appropriate input. Once the network is trained, it is used to build functions graph - network is presented
with new inputs it did not see before and its output is used as function's value. The sample uses back propagation learning
algorithm for teaching neural network.
Time Series Prediction [Download]
This sample demonstrates a bit more complex neural network, which is used for time series prediction problem. In this
sample neural network has multiple inputs, which amount is specified by amount of values from the history used to predict
new value, and one output - predicted future value. The sample uses back propagation learning algorithm for teaching
neural network.
Note: this application is just a sample. It will not do any real predictions, so don't need to ask how to feed
Forex data to it or anything else. The application will just reduce training set by the specified amount of points and
then it will compare network's prediction with those value, which were given by user, but not used for training.
|
