Advanced search
Start date
Betweenand

Active, semi-supervised learning for classification and clustering

Grant number: 14/01352-0
Support type:Scholarships in Brazil - Master
Effective date (Start): May 01, 2014
Effective date (End): February 29, 2016
Field of knowledge:Physical Sciences and Mathematics - Computer Science
Principal Investigator:Ricardo José Gabrielli Barreto Campello
Grantee:Antônio José de Lima Batista
Home Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil

Abstract

There are many situations in machine learning in which it is easy to obtain vast amounts of unlabeled data, but it is costly to obtain labels for these data. This leads to typical situations in which only a small subset of labeled data is available, which is not representative enough for learning a model in a supervised way. In such situations, techniques that are able to learn from both labeled and unlabeled data simultaneosuly, the so-called semi-supervised learning techniques, become essential. Specifically, in scenarios where it is possible to arbitrarily label any observation for a cost you want to minimize, it becomes particularly important the use of semi-supervised learning techniques which go beyond the conventional techniques and are also able to select the observations to be labeled in a proactive way. The goal is to maximize the quality of the learned models from the smallest possible amount of labels required. Techniques with this capability fit the paradigm of active learning. In active semi-supervised learning, an algorithm can determine at any time the labels still unknown that are deemed more important to proceed with the learning process. Generally speaking, this research project aims to investigate possible improvements of existing methods of active learning for the tasks of semi-supervised pattern classification and semi-supervised data clustering, as well as to evaluate the behavior of the methods investigated in experiments using real and simulated data.

Academic Publications
(References retrieved automatically from State of São Paulo Research Institutions)
BATISTA, Antônio José de Lima. Active semi-supervised classification based on multiple clustering hierarchies. 2016. Master's Dissertation - Universidade de São Paulo (USP). Instituto de Ciências Matemáticas e de Computação São Carlos.

Please report errors in scientific publications list by writing to: cdi@fapesp.br.