Query by Transduction

There has been recently a growing interest in the use of transductive inference for learning. We expand here the scope of transductive inference to active learning in a stream-based setting. Towards that end this paper proposes Query-by-Transduction (QBT) as a novel active learning algorithm. QBT qu...

Celý popis

Uloženo v:

Podrobná bibliografie
Vydáno v:	IEEE transactions on pattern analysis and machine intelligence Ročník 30; číslo 9; s. 1557 - 1571
Hlavní autoři:	Shen-Shyang Ho, Wechsler, H.
Médium:	Journal Article
Jazyk:	angličtina
Vydáno:	Los Alamitos, CA IEEE 01.09.2008 IEEE Computer Society The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Témata:	Algorithms Applied sciences Artificial Intelligence Bayesian methods Classification Computer science; control theory; systems Data processing. List processing. Character string processing Exact sciences and technology Feasibility Frequency Inference Inference algorithms Information Storage and Retrieval - methods Information systems. Data bases Labeling Learning Machine learning Mathematics Memory organisation. Data processing Pattern analysis Pattern Recognition, Automated - methods Probability and statistics Query processing Sampling methods Sampling theory, sample surveys Sciences and techniques of general use Sensitivity and Specificity Software Statistical Statistical analysis Statistics Support vector machine classification Support vector machines Tasks Testing Utilities Statistical Machine learning Bayes estimation Statistical analysis Divergence Random sampling Database query hypothesis testing Active learning Inference Shannon theory transductive inference Hypothesis test Statistical test support vector machine Active system Kolmogorov complexity Vector support machine Feasibility Learning algorithm Pattern analysis Kolmogorov equation Artificial intelligence Multiclass
ISSN:	0162-8828, 1939-3539
On-line přístup:	Získat plný text
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!

Popis
Shrnutí:	There has been recently a growing interest in the use of transductive inference for learning. We expand here the scope of transductive inference to active learning in a stream-based setting. Towards that end this paper proposes Query-by-Transduction (QBT) as a novel active learning algorithm. QBT queries the label of an example based on the p-values obtained using transduction. We show that QBT is closely related to Query-by-Committee (QBC) using relations between transduction, Bayesian statistical testing, Kullback-Leibler divergence, and Shannon information. The feasibility and utility of QBT is shown on both binary and multi-class classification tasks using SVM as the choice classifier. Our experimental results show that QBT compares favorably, in terms of mean generalization, against random sampling, committee-based active learning, margin-based active learning, and QBC in the stream-based setting.
Bibliografie:	ObjectType-Article-2 SourceType-Scholarly Journals-1 ObjectType-Feature-1 content type line 14 content type line 23 ObjectType-Article-1 ObjectType-Feature-2
ISSN:	0162-8828 1939-3539
DOI:	10.1109/TPAMI.2007.70811