Train Neural Net classifier
Train a classifier of the 'Neural Net' type.
Analyzer
-
Activation function
Choice: LINEAR LOGISTIC_SIGMOID SOFTMAX Required -
Classification
InputColumn<Object> Required -
Cross validation sample rate
Determine how much (if any) of the records should be used for cross-validation.
Percentage Required -
Epochs
int Required -
Error function
Choice: LEAST_MEAN_SQUARES CROSS_ENTROPY Required -
Feature modifier types
List of Choice: Scaled (Min-Max) Direct (0.0 to 1.0) Direct (1 or 0) Vector (One Hot Encoding) Vector (2-gram) Vector (3-gram) Vector (4-gram) Vector (5-gram) Mapped with Features Required -
Features
List of InputColumn<Object> Required -
Hidden layers
List of int Required -
Include unique value features
Include generated features that are only triggered once in the training data set.
boolean Required -
Learning rate
double Required -
Max features generated per column
Defines the maximum number of features to generate per column. Applies to feature vectors such as 'One-Hot Encoding' or n-grams.
Integer Required -
Momentum
double Required -
Save model to file
File Optional