Terry Windeatt

Structural, Syntactic, and Statistical Pattern Recognition

Peer-to-Peer (P2P) lending is an online platform to facilitate borrowing and investment transactions. A central problem for these P2P platforms is how to identify the most influential factors that are closely related to the credit risks. This problem is inherently complex due to the various forms of risks and the numerous influencing factors involved. Moreover, raw data of P2P lending are often high-dimension, highly correlated and unstable, making the problem more untractable by traditional statistical and machine learning approaches. To address these problems, we develop a novel filter-based feature selection method for P2P lending analysis. Unlike most traditional feature selection methods that use vectorial features, the proposed method is based on graphbased features and thus incorporates the relationships between pairwise feature samples into the feature selection process. Since the graph-based features are by nature completed weighted graphs, we use the steady state random walk to encapsulate the main characteristics of the graphbased features. Specifically, we compute a probability distribution of the walk visiting the vertices. Furthermore, we measure the discriminant power of each graph-based feature with respect to the target feature, through the Jensen-Shannon divergence measure between the probability distributions from the random walks. We select an optimal subset of features based on the most relevant graph-based features, through the Jensen-Shannon divergence measure. Unlike most existing state-of-theart feature selection methods, the proposed method can accommodate both continuous and discrete target features. Experiments demonstrate the effectiveness and usefulness of the proposed feature selection algorithm on the problem of P2P lending platforms in China.

Reliable Surface Reconstructiuon from Multiple Range Images

This paper addresses the problem of reconstructing an integrated 3D model from multiple 2.5D range images. A novel integration algorithm is presented based on a continuous implicit surface representation. This is the first reconstruction algorithm to use operations in 3D space only. The algorithm is guaranteed to reconstruct the correct topology of surface features larger than the range image sampling resolution. Reconstruction of triangulated models from multi-image data sets is demonstrated for complex objects. Performance characterization of existing range image integration algorithms is addressed in the second part of this paper. This comparison defines the relative computational complexity and geometric limitations of existing integration algorithms.

Coding and decoding strategies for multi-class learning problems

Abstract It is known that the error correcting output code (ECOC) technique, when applied to multi-class learning problems, can improve generalisation performance. One reason for the improvement is its ability to decompose the original problem into complementary two-class problems. Binary classifiers trained on the sub-problems are diverse and can benefit from combining using a simple distance-based strategy. However there is some discussion about why ECOC performs as well as it does, particularly with respect to the significance of the coding/decoding strategy. In this paper we consider the binary (0,1) code matrix conditions necessary for reduction of error in the ECOC framework, and demonstrate the desirability of equidistant codes. It is shown that equidistant codes can be generated by using properties related to the number of 1’s in each row and between any pair of rows. Experimental results on synthetic data and a few popular benchmark problems show how performance deteriorates as code length is reduced for six decoding strategies.

Diversity measures for multiple classifier system analysis and design

Abstract In the context of Multiple Classifier Systems, diversity among base classifiers is known to be a necessary condition for improvement in ensemble performance. In this paper the ability of several pair-wise diversity measures to predict generalisation error is compared. A new pair-wise measure, which is computed between pairs of patterns rather than pairs of classifiers, is also proposed for two-class problems. It is shown experimentally that the proposed measure is well correlated with base classifier test error as base classifier complexity is systematically varied. However, correlation with unity-weighted sum and vote is shown to be weaker, demonstrating the difficulty in choosing base classifier complexity for optimal fusion. An alternative strategy based on weighted combination is also investigated and shown to be less sensitive to number of training epochs.

Accuracy/Diversity and Ensemble MLP Classifier Design

The difficulties of tuning parameters of multilayer perceptrons (MLP) classifiers are well known. In this paper, a measure is described that is capable of predicting the number of classifier training epochs for achieving optimal performance in an ensemble of MLP classifiers. The measure is computed between pairs of patterns on the training data and is based on a spectral representation of a Boolean function. This representation characterizes the mapping from classifier decisions to target label and allows accuracy and diversity to be incorporated within a single measure. Results on many benchmark problems, including the Olivetti Research Laboratory (ORL) face database demonstrate that the measure is well correlated with base-classifier test error, and may be used to predict the optimal number of training epochs. While correlation with ensemble test error is not quite as strong, it is shown in this paper that the measure may be used to predict number of epochs for optimal ensemble performance. Although the technique is only applicable to two-class problems, it is extended here to multiclass through output coding. For the output-coding technique, a random code matrix is shown to give better performance than one-per-class code, even when the base classifier is well-tuned

Implicit Surface-Based Geometric Fusion

This paper introduces a general purpose algorithm for reliable integration of sets of surface measurements into a single 3D model. The new algorithm constructs a single continuous implicit surface representation which is the zero-set of a scalar field function. An explicit object model is obtained using any implicit surface polygonization algorithm. Object models are reconstructed from both multiple view conventional 2.5D range images and hand-held sensor range data. To our knowledge this is the first geometric fusion algorithm capable of reconstructing 3D object models from noisy hand-held sensor range data.This approach has several important advantages over existing techniques. The implicit surface representation allows reconstruction of unknown objects of arbitrary topology and geometry. A continuous implicit surface representation enables reliable reconstruction of complex geometry. Correct integration of overlapping surface measurements in the presence of noise is achieved using geometric constraints based on measurement uncertainty. The use of measurement uncertainty ensures that the algorithm is robust to significant levels of measurement noise. Previous implicit surface-based approaches use discrete representations resulting in unreliable reconstruction for regions of high curvature or thin surface sections. Direct representation of the implicit surface boundary ensures correct reconstruction of arbitrary topology object surfaces. Fusion of overlapping measurements is performed using operations in 3D space only. This avoids the local 2D projection required for many previous methods which results in limitations on the object surface geometry that is reliably reconstructed. All previous geometric fusion algorithms developed for conventional range sensor data are based on the 2.5D image structure preventing their use for hand-held sensor data. Performance evaluation of the new integration algorithm against existing techniques demonstrates improved reconstruction of complex geometry.

Face verification using error correcting output codes

The error correcting output coding (ECOC) approach to classifier design decomposes a multi-class problem into a set of complementary two-class problems. We show how to apply the ECOC concept to automatic face verification, which is inherently a two-class problem. The output of the binary classifiers defines the ECOC feature space, in which it is easier to separate transformed patterns representing clients and impostors. We propose two different combining strategies as the matching score for face verification. The first uses the first order Minkowski metric, and requires a threshold to be set. The second is a kernel-based method and has no parameters to set. The proposed method exhibits better performance on the well known XM2VTS data set compared with previous reported results.

Binary labelling and decision-level fusion

Abstract Two binary labelling techniques for decision-level fusion are considered for reducing correlation in the context of multiple classifier systems. First, we describe a method based on error correcting coding that uses binary code words to decompose a multi-class problem into a set of complementary two-class problems. We look at the conditions necessary for reduction of error and introduce a modified version that is less sensitive to code word selection. Second, we describe a partitioning method for two-class problems that transforms each training pattern into a vertex of the binary hypercube. A constructive algorithm for binary-to-binary mappings identifies a set of inconsistently classified patterns, random subsets of which are used to perturb base classifier training sets. Experimental results on artificial and real data, using a combination of simple neural network classifiers, demonstrate improvement in performance for these techniques, the first suitable for k-class problems, k>2 and the second for k=2.

Strategies for combining classifiers employing shared and distinct pattern representations

The problem of combining multiple classifiers which employ mixed mode representations consisting of some shared and some distinct features is studied. Two combination strategies are developed and experimentally compared on mammographic data to demonstrate their effectiveness.

Vote counting measures for ensemble classifiers

Various measures, such as Margin and Bias/Variance, have been proposed with the aim of gaining a better understanding of why Multiple Classifier Systems (MCS) perform as well as they do. While these measures provide different perspectives for MCS analysis, it is not clear how to use them for MCS design. In this paper a different measure based on a spectral representation is proposed for two-class problems. It incorporates terms representing positive and negative correlation of pairs of training patterns with respect to class labels. Experiments employing MLP base classifiers, in which parameters are fixed but systematically varied, demonstrate the sensitivity of the proposed measure to base classifier complexity.