In this paper the performance of bagging in classification problems is theoretically analysed, using a framework developed in works by Tumer and Ghosh and extended by the authors. A bias-variance decomposition is derived, which relates the expected misclassification probability attained by linearly combining classifiers trained on N bootstrap replicates of a fixed training set to that attained ...

In a growing number of domains the data collected has a large number of features. This poses a challenge to classical pattern recognition techniques, since the number of samples often is still limited with respect to the feature size. Classical pattern recognition methods suffer from the small sample size, and robust classification techniques are needed. In order to reduce the dimensionality of...

2. State Key Lab. for Novel Software Technology, Nanjing University, P.R. China Abstract: The small sample size (SSS) and the sensitivity to variations such as illumination, expression and occlusion are two challenging problems in face recognition. In this paper, we propose a novel method, called semi-random subspace (Semi-RS), to simultaneously address the two problems. Different from the trad...

Signal parameter estimation from sensor array data is a problem that is encountered in many engineering applications. Under the assumption of Gaussian distributed emitter signals, the so-called stochastic maximum likelihood (ML) technique is known to be statistically efficient, i.e., the estimation error covariance attains the Cramer-Rao bound (CRB) asymptotically. Herein, it is shown that also...

Ensemble classification involves combining multiple classifiers to obtain more accurate predictions than those obtained using individual models. Ensemble techniques are known to be very useful in improving the generalization ability of a classifier. The random subspace ensemble technique is a simple but effective method of constructing ensemble classifiers, in which some features are randomly d...

Supervised classification tasks like Sentiment Analysis or text classification need labelled training data. These labels can be difficult to obtain, especially for complicated and ambiguous data like texts. Instead of labelling new data, domain adaptation tries to reuse already labelled data from related tasks as training data. We propose a greedy selection strategy to identify a small subset o...

This work investigates the problem of adaptive measurement design for online subspace estimation from compressive linear measurements. We study the previously proposed Grassmannian rank-one online subspace estimation (GROUSE) algorithm with adaptively designed compressive measurements. We propose an adaptive measurement scheme that biases the measurement vectors towards the current subspace est...

We achieve two goals in this paper: (1) to build a novel appearance-based object representation that takes into account variations in contrast often found in training images; (2) to develop a robust appearance-based detection scheme that can handle outliers such as occlusion and structured noise. To build the representation, we decompose the input ensemble into two subspaces: a principal subspa...