All datasests are taken from the UCI machine learning database
- Abalone Dataset - Classification
- Concrete Slump Dataset - Regression
- Yacht Hydrodynamics Data Set - Regression
- Wine - Classification
- Letter Recognition - Classification
The user is presented with a GUI consisting of three columns.
1st Column) Select a local data file to load, if a local file is desired. Select type of problem (classification, regression) from second column
2nd Column) Enter required fields to load data directly from the UCI repository
1. -UCI source URL, a URL like the ones shown with the datasets above.
2. -Class labels, add one label at a time, must match labels in dataset exactly
3. How many features? The number of features associated with the instances
4. How many instances? The number of instances in the dataset
5. Problem Type: Drop down menu providing a selection of classification or regression problem
6. Label Index: Is the label the first or last value in a dataset instance
7. Load!: Load the dataset
3rd Column) Network and training parameters
1) Maximum iterations: Maximum number of iterations or generations to run if training does not reach
convergence criteria.
2) Hidden Layer Nodes: The number of nodes to use in the hidden layers. Specified as comma separated values.
For example: 3, 7 would create a network with 3 nodes in the 1st hidden layer and 7 nodes in the 2nd
hidden layer
3) Activation Function: Drop down menu that provides choice of sigmoid or hyperbolic tangent function
4) Learning Rate: The learning rate used in weight updates during backpropagation training
5) Momentum: If box is checked momentum will be incorporated in the weight updates during backpropagation
6) Beta: The parameter used to set the influence of momentum in the weight updates.
Typically between 0.5 and 1.
7) Population Size: Population size to use in the evolutionary algorithms
8) Algorithm Selection: Drop down menu to select the desired training method.
9) Write ouput: Check box if the user wants to write the test output to a file
10) Build and run: Start training and testing