Online Semi-supervised Learning and its Application in Image Classification

Please enable javascript in your browser.

Shaban, Amir Reza | 2011

1576 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 44248 (19)
University: Sharif University of Technology
Department: Computer Engineering
Advisor(s): Rabiee, Hamid Reza
Abstract:
Image classification, i.e. the task of assigning an image to a class chosen from a predefined set of classes, has addressed in this thesis. At first the classifier is divided into two major sub partitions, feature extraction and classifier. Then we show that by using local feature extraction techniques such as BOW the classification accuracy will improve. In addition, using unlabeled data is argued as the fact to deal with high nonlinear structure of features. Recently, many SSL methods have been developed based on the manifold assumption in a batch mode. However, when data arrive sequentially and in large quantities, both computation and storage limitations become a bottleneck. So in large scale classification task developing an online SSL learning machine can be helpful. The learning machine is based on data reduction by coarse graining algorithm that reduces data size while keep manifold structure. The experiments on Caltech101 and CIFAR10 datasets show that by using coarse graining method the accuracy does decrease in minimum compared with alternate batch method. Furthermore, unlike batch methods, our online algorithm classifies data with constant time and memory
Keywords:
Semi-Supervised Learning ; Images Classification ; Online Learning ; Manifold Assumption ; Feature Coding