RProp MLP Learner

Implementation of the RProp algorithm for multilayer feedforward networks. RPROP performs a local adaptation of the weight-updates according to the behavior of the error function. For further details see: Riedmiller, M. Braun, H. : "A direct adaptive method for faster backpropagation learning: theRPROP algorithm",Proceedings of the IEEE International Conference on Neural Networks (ICNN) (Vol. 16, pp. 586-591). Piscataway, NJ: IEEE. This node provides a view of the error plot.
If the optional PMML inport is connected and contains preprocessing operations in the TransformationDictionary those are added to the learned model.

Options

Maximum number of iterations

The number of learning iterations.

Number of hidden layers

Specifies the number of hidden layers in the architecture of the neural network.

Number of hidden neurons per layer

Specifies the number of neurons contained in each hidden layer.

Class column

Choose the column that contains the target variable: it can either be nominal or numerical. All nominal class values are extracted and assigned to output neurons. If you use a numerical target variable (regression), please make sure it is normalized!

Ignore missing values

If this checkbox is set, rows with missing values will not be used for training.

Use seed for random initialization

If this checkbox is set, a seed (see next field) can be set for initializing the weights and thresholds can be set.

Random seed

Seed for the random number generator.

Input Ports

Datatable with training data

Optional PMML port object containing preprocessing operations.

Output Ports

RProp trained Neural Network

Views

Error Plot

Displays the error for each iteration.

Workflows

• 03_​ZIP_​Local_​vs_​remoteHTTP
• 01_​Analytics
• 02_​Techniques_​for_​Dimensionality_​Reduction

Installation

To use this node in KNIME, install KNIME Base Nodes from the following update site: