Evaluates the output of svm_rank_classify or other reranking classifier This currently computes P@1 This works for the QA system, but it should work for every generic svm_rank_classify like output User: mihais Date: 9/11/13
Dataset containing only BVFDatums Important note: to encode feature values > 1, simply store the same feature multiple times (equal to feature value)!
Datum that contains only binary- (or Int) valued features
Classifier that implements bagging over another Classifier Created by dfried, mihais Date: 4/25/14
Trait for iid classification For reranking problems, see RankingClassifier User: mihais Date: 11/17/13
Dataset that represents datums as explicit counters This is more efficient for the training of various algorithms such as random forests
Parent class for classification datasets User: mihais Date: 4/23/13
Stores one fold, containing testing and training partitions Each tuple stores start and end offsets, starting from 0 User: mihais Date: 5/1/13
Operations on datasets User: mihais Date: 5/1/13
Trait for ML datums.
Trait for ML datums. L indicates the type of the label; F indicates the type of the feature User: mihais Date: 4/23/13
Analogous to iterable -- defines a method for geting the feature traverser and updater Created by dfried on 5/27/14.
Allows traversal of a dataset (or ranking dataset)'s features and values, and also destructive updates of the values.
Allows traversal of a dataset (or ranking dataset)'s features and values, and also destructive updates of the values. Useful for finding the range of values and then rescaling them. Analogous to iterator with destructive updates Created by dfried on 5/27/14.
Linear SVM with L1 regularization
LR with L1 regularization
Modified from mihais's Liblinear wrapper by dfried on 5/2/14
Wrapper for liblinear classifiers, which includes LR and linear SVM Note: this only supports classification; it does not support regression by design User: mihais Date: 11/16/13
Linear SVM with L2 regularization
Vanilla logistic regression with L2 regularization
Multiclass perceptron classifier, in primal mode Includes averaging, hard margin, burn-in iterations User: mihais Date: 12/15/13
Perceptron classifier for ranking, in primal mode Includes averaging, hard margin, burn-in iterations User: mihais Date: 12/10/13
An in-house implementation of random forests User: mihais Date: 11/23/15
Dataset containing only RVFDatums
Dataset containing only RVFDatums
Type of labels
Type of features
Datum that contains real-valued features
Datum that contains real-valued features and kernelized representation
Generic trait for ranking classifiers; for iid classification see Classifier User: mihais Date: 4/23/13
Parent class for all datasets used for ranking problems User: mihais Date: 4/23/13
Wrapper for SVMrank: trains using svm_rank_learn but predicts using native Scala code Only the linear kernel is supported User: mihais Date: 4/23/13
Simple example of how to use RVF data and classifiers User: mihais Date: 11/18/15
Dataset containing only BVFDatums Important note: to encode feature values > 1, simply store the same feature multiple times (equal to feature value)!
Type of labels
Type of features