#Inputs#Hiddens (n0,n1,n2,...)#OutputsUse Bias

Net Evo
size:
POPULATION
rows: cols:
L(x):
Training Data



epochs:
learning rate:
Net Evo
weight mutation:%
epochs: node mutation:%
kill rate:% layer mutation:%
Borrower I/O:
Lender I/O:
Lender ID:
Selected Network:
Classification error:
Classification rate:
Highlighted Node:







INSTRUCTIONS:

Overview
This page allows you to create, train, and visualize neural networks.

Input/Output Data
The space under "Training Data" allows you to define new data to add to the training set, shown in the box below the "add data" and "clear data buttons". There must be some data in this box in order for training to proceed. To view how a particular data point feeds forward into the network, press "F" on the right side of that line of data. To run one back-propagation on that single data point, press "B" on the right. To delete a data point from the training set, press "X" on the right.

Back-propagation
You can press "create network" to create a new random network with the desired architecture. The "backpropagation" button will train the displayed network on the selected training set with the desired parameters.

Evolution
Press "create population" after loading training data to create the specified number of networks. They will be displayed under "population" in boxes showing their fitness figure. The "evolution" button will use a genetic algorithm with the desired parameters to train the population, and will display the fittest memeber of the population.

Other
It is strongly suggested to check the "use bias" box, as networks with bias units are a great deal easier to train.