Papers by Theodore Kalamboukis
Proceedings of the 31st annual …, Jan 1, 2008
Lecture Notes in Computer Science, 2013
ABSTRACT This article presents an experimental evaluation on image representation using Latent Se... more ABSTRACT This article presents an experimental evaluation on image representation using Latent Semantic Analysis (LSA) for searching very large image databases. Our aim is twofold: First, we experimentally investigate the structure and size of the feature space in order for LSA to bring efficient results. Second, we replace the Singular Value Decomposition (SVD) analysis on the feature matrix, by solving the eigenproblem of the term correlation matrix, a much less memory demanding task which significantly improved the performance in both accuracy and computational time (preprocessing and query response time) on three large image collections. Finally the new approach overcomes the high cost of updating the database after new insertions.
Computerized medical imaging and graphics : the official journal of the Computerized Medical Imaging Society, 2015
Latent Semantic Analysis (LSA) although has been used successfully in text retrieval when applied... more Latent Semantic Analysis (LSA) although has been used successfully in text retrieval when applied to CBIR induces scalability issues with large image collections. The method so far has been used with small collections due to the high cost of storage and computational time for solving the SVD problem for a large and dense feature matrix. Here we present an effective and efficient approach of applying LSA skipping the SVD solution of the feature matrix and overcoming in this way the deficiencies of the method with large scale datasets. Early and late fusion techniques are tested and their performance is calculated. The study demonstrates that early fusion of several composite descriptors with visual words increase retrieval effectiveness. It also combines well in a late fusion for mixed (textual and visual) ad hoc and modality classification. The results reported are comparable to state of the art algorithms without including additional knowledge from the medical domain.
Proceedings of the 17th Panhellenic Conference on Informatics - PCI '13, 2013
ABSTRACT This paper addresses the problem of learning to classify texts by exploiting information... more ABSTRACT This paper addresses the problem of learning to classify texts by exploiting information derived from clustering both training and testing sets. The incorporation of knowledge resulting from clustering into the feature space representation of the texts is expected to boost the performance of a classifier. Two different approaches to clustering are described, an unsupervised and a semi-supervised one. We present an empirical study of the proposed algorithms on a variety of datasets. The results are encouraging, revealing that information resulting from clustering can create text classifiers of high-accuracy.
International Journal on Artificial Intelligence Tools, 2011
ABSTRACT This paper addresses the problem of learning to classify texts by exploiting information... more ABSTRACT This paper addresses the problem of learning to classify texts by exploiting information derived from clustering both training and testing sets. The incorporation of knowledge resulting from clustering into the feature space representation of the texts is expected to boost the performance of a classifier. We present an empirical study of the proposed algorithm on a variety of datasets. We attempt to answer various questions such as the effect of independence or relevance amongst features, the effect of the size of the training set and the effect of noise. The results are encouraging, revealing that information resulting from clustering can create text classifiers of high-accuracy.
The Discovery Challenge Workshop
... Train a SVM/TSVM classifier based on the expanded training examples. Classify the expanded te... more ... Train a SVM/TSVM classifier based on the expanded training examples. Classify the expanded testing examples. ... On all the collections the clustering approach combined witha SVM/TSVM classifier outperformed the standard SVM/TSVM classifier. ...
ECML PKDD Discovery Challenge 2009 (DC09), 2009
Abstract. A modified and fast to converge Perceptron learning rule algorithm is proposed as a gen... more Abstract. A modified and fast to converge Perceptron learning rule algorithm is proposed as a general classification algorithm for linearly separable data. The strategy of the algorithm takes advantage of training errors to successively refine an initial Perceptron Classifier. Original Perceptron learning rule uses training errors along with a parameter α (learning rate parameter that has to be determined) to define a better classifier. The proposed modification does not need such a parameter (in fact it is automatically determined during execution of ...
A fast and accurate linear supervised algorithm is presented which compares favorably to other st... more A fast and accurate linear supervised algorithm is presented which compares favorably to other state of the art algorithms over several real data collections on the problem of text categorization. Although it has been already presented in [6], no proof of its convergence is given. From the geometric intuition of the algorithm it is evident that it is not a Perceptron or a gradient descent algorithm thus an algebraic proof of its convergence is provided in the case of linearly separable classes. Additionally we present experimental results on many ...
ECML PKDD Discovery …, Jan 1, 2008
Uploads
Papers by Theodore Kalamboukis