Represents a classifier from observations of type T to labels of type L.
This stupidly named class is a Label-Feature Matrix, which is to say that it's a the weight matrix used by most of the classifier trainers.
This stupidly named class is a Label-Feature Matrix, which is to say that it's a the weight matrix used by most of the classifier trainers. It's basically a matrix with one row per label, and the rows are some Tensor type (TF). TF is a mnemonic for Feature Tensor.
label type
feature tensor type
A LinearClassifier is a multi-class classifier with decision
function:
\hat y_i = \arg\max_y w_y^T x_i + b_y
A LinearClassifier is a multi-class classifier with decision
function:
\hat y_i = \arg\max_y w_y^T x_i + b_y
dialogue 6/19/14
A NeuralNetwork classifier uses a neural network to get unnormalize log probabilities for the scores of the classifier.
A NeuralNetwork classifier uses a neural network to get unnormalize log probabilities for the scores of the classifier. These are used to predict terms.
Implements a Naive-Bayes Classifer over bags of words.
Implements a Naive-Bayes Classifer over bags of words. It automatically trains itself given the collection c of learning examples.
This is the unindexed weights matrix: it acts as a tensor over the label types, rather than their indexed components
This is the unindexed weights matrix: it acts as a tensor over the label types, rather than their indexed components
kNearestNeighbor 6/8/14
nak 7/7/14
A multi-class logistic/softmax/maxent classifier.
This is an example app for creating a logistic classifier from data that is stored as string valued features and string valued labels, e.
This is an example app for creating a logistic classifier from data that is stored as string valued features and string valued labels, e.g.
verb=join,noun=board,prep=as,prep_obj=director,V verb=isIs,noun=chairman,prep=of,prep_obj=N.V.,N verb=named,noun=director,prep=of,prep_obj=conglomerate,N
These are examples from Ratnarparkhi's classic prepositional phrase attachment dataset, discussed in the following homework:
http://ata-s12.utcompling.com/assignments/classification
The homework includes pointers to the data and to Scala code for generating said features.
This example handles creating a feature index and getting the examples into the right data structures for training with the logistic regression classifier, which should serve as a useful example for creating features and classifiers using the API.
Object for creating SupportVectorMachines
Main program that builds a classifier from a breeze.data.SparseFeatureDataset.
Main program that builds a classifier from a breeze.data.SparseFeatureDataset. You can build a logistic classifier or an SVM, at the moment.
Represents a classifier from observations of type T to labels of type L. Implementers should only need to implement score.