SGDText

Package

weka.classifiers.functions

Synopsis

Implements stochastic gradient descent for learning a linear binary class SVM or binary class logistic regression on text data. Operates directly on String attributes. From Weka 3.7.5.

Options

The table below describes the options available for SGDText.

Option	Description
LNorm	The LNorm to use for document length normalization.
debug	If set to true, classifier may output additional info to the console.
epochs	The number of epochs to perform (batch learning). The total number of iterations is epochs * num instances.
lambda	The regularization constant. (default = 0.0001)
learningRate	The learning rate.
lossFunction	The loss function to use. Hinge loss (SVM), log loss (logistic regression) or squared loss (regression).
lowercaseTokens	Whether to convert all tokens to lowercase
minWordFrequency	Ignore any words that don't occur at least min frequency times in the training data. If periodic pruning is turned on, then the dictionary is pruned according to this value
norm	The norm of the instances after normalization.
periodicPruning	How often (number of instances) to prune the dictionary of low frequency terms. 0 means don't prune. Setting a positive integer n means prune after every n instances
seed	The random number seed to be used.
stemmer	The stemming algorithm to use on the words.
stopwords	The file containing the stopwords (if this is a directory then the default ones are used).
tokenizer	The tokenizing algorithm to use on the strings.
useStopList	If true, ignores all words that are on the stoplist.
useWordFrequencies	Use word frequencies rather than binary bag of words representation

Capabilities

The table below describes the capabilities of SGDText.

Capability	Supported
Class	Binary class, Missing class values
Attributes	String attributes, Missing values
Min # of instances	0