A triple-random ensemble classification method for mining multi-label data

This paper presents a triple-random ensemble learning method for handling multi-label classification problems. The proposed method integrates and develops the concepts of random subspace, bagging and random k-label sets ensemble learning methods to form an approach to classify multi-label data. It applies the random subspace method to feature space, label space as well as instance space. The devised subsets selection procedure is executed iteratively. Each multi-label classifier is trained using the randomly selected subsets. At the end of the iteration, optimal parameters are selected and the ensemble MLC classifiers are constructed. The proposed method is implemented and its performance compared against that of popular multi-label classification methods. The experimental results reveal that the proposed method outperforms the examined counterparts in most occasions when tested on six small to larger multi-label datasets from different domains. This demonstrates that the developed method possesses general applicability for various multi-label classification problems.

Pharmacy access to the emergency contraceptive pill : a national survey of a random sample of Australian women

Background : The emergency contraceptive pill (ECP) has the potential to assist in reducing unintended pregnancy and abortion rates. Since its rescheduling to pharmacy availability without prescription in Australia in January 2004, there is little information about Australian women's knowledge, attitudes and use of the ECP. The aim of this study was to measure the knowledge about the ECP and sociodemographic patterns of and barriers to use of the ECP.

Study Design : A cross-sectional study, using a computer-assisted telephone interview (CATI) survey conducted with a national random sample of 632 Australian women aged 16–35 years.

Results : Most women had heard of the ECP (95%) and 26% had used it. The majority of women agreed with pharmacy availability of the ECP (72%); however, only 48% were aware that it was available from pharmacies without a prescription. About a third (32%) believed the ECP to be an abortion pill. The most common reason for not using the ECP was that women did not think they were at risk of getting pregnant (57%). Logistic regression showed that women aged 20–29 years (OR 2.58; CI: 1.29–5.19) and 30–35 years (OR 3.16; CI: 1.47–6.80) were more likely to have used the ECP than those aged 16–19 years. Women with poor knowledge of the ECP were significantly less likely to have used it than those with very good knowledge (OR 0.28; CI: 0.09–0.77). Those in a de facto relationship (OR 2.21; CI: 1.27–3.85), in a relationship but not living with the partner (OR 2.46; 95% CI 1.31–4.63) or single women (OR 2.40; CI: 1.33–4.34) were more likely to have used the ECP than married women.

Conclusions : Women in Australia have a high level of awareness of the ECP, but more information and education about how to use it and where to obtain it are still needed.

Conjugated polymer−titania nanoparticle hybrid films : random lasing action and ultrasensitive detection of explosive vapors

We have first demonstrated that a random laser action generated by a hybrid film composed of a semiconducting organic polymer (SOP) and TiO₂ nanoparticles can be used to detect 2,4,6-trinitrotoluene (TNT) vapors. The hybrid film was fabricated by spin-casting SOP solution dispersed with nanosized TiO₂ particles on quartz glass. The SOP in the hybrid film functioned as both the gain medium and the sensory transducer. A random lasing action was observed with a certain pump power when the size (diameter of 50 nm) and concentration (8.9 - 10¹²/cm³) of TiO₂ nanoparticles were optimized. Measurements of fluorescence quenching behavior of the hybrid film in TNT vapor atmosphere (10 ppb) showed that attenuated lasing in optically pumped hybrid film displayed a sensitivity to vapors of explosives more than 20 times higher than was observed from spontaneous emission. This phenomenon has been explained with the four-level laser model. Since the sensory transducer used in the hybrid polymer/nanoparticles system could be replaced by other functional materials, the concept developed could be extended to more general domains of chemical or environment detection.

Malware variant detection using similarity search over sets of control flow graphs

Static detection of polymorphic malware variants plays an important role to improve system security. Control flow has shown to be an effective characteristic that represents polymorphic malware instances. In our research, we propose a similarity search of malware using novel distance metrics of malware signatures. We describe a malware signature by the set of control flow graphs the malware contains. We propose two approaches and use the first to perform pre-filtering. Firstly, we use a distance metric based on the distance between feature vectors. The feature vector is a decomposition of the set of graphs into either fixed size k-sub graphs, or q-gram strings of the high-level source after decompilation. We also propose a more effective but less computationally efficient distance metric based on the minimum matching distance. The minimum matching distance uses the string edit distances between programs' decompiled flow graphs, and the linear sum assignment problem to construct a minimum sum weight matching between two sets of graphs. We implement the distance metrics in a complete malware variant detection system. The evaluation shows that our approach is highly effective in terms of a limited false positive rate and our system detects more malware variants when compared to the detection rates of other algorithms.

Human action categorization using conditional random field

Automatic human action recognition has been a challenging issue in the field of machine vision. Some high-level features such as SIFT, although with promising performance for action recognition, are computationally complex to some extent. To deal with this problem, we construct the features based on the Distance Transform of body contours, which is relatively simple and computationally efficient, to represent human action in the video. After extracting the features from videos, we adopt the Conditional Random Field for modeling the temporal action sequences. The proposed method is tested with an available standard dataset. We also testify the robustness of our method on various realistic conditions, such as body occlusion or intersection.

Assessment of fracture risk : value of random population-based samples - the Geelong Osteoporosis Study

Fracture risk is determined by bone mineral density (BMD). The T-score, a measure of fracture risk, is the position of an individual's BMD in relation to a reference range. The aim of this study was to determine the magnitude of change in the T-score when different sampling techniques were used to produce the reference range. Reference ranges were derived from three samples, drawn from the same region: (1) an age-stratified population-based random sample, (2) unselected volunteers, and (3) a selected healthy subset of the population-based sample with no diseases or drugs known to affect bone. T-scores were calculated using the three reference ranges for a cohort of women who had sustained a fracture and as a group had a low mean BMD (ages 35-72 yr; n = 484). For most comparisons, the T-scores for the fracture cohort were more negative using the population reference range. The difference in T-scores reached 1.0 SD. The proportion of the fracture cohort classified as having osteoporosis at the spine was 26, 14, and 23% when the population, volunteer, and healthy reference ranges were applied, respectively. The use of inappropriate reference ranges results in substantial changes to T-scores and may lead to inappropriate management.

AdaBoost.MRF: boosted Markov random forests and application to multilevel activity recognition

Activity recognition is an important issue in building intelligent monitoring systems. We address the recognition of multilevel activities in this paper via a conditional Markov random field (MRF), known as the dynamic conditional random field (DCRF). Parameter estimation in general MRFs using maximum likelihood is known to be computationally challenging (except for extreme cases), and thus we propose an efficient boosting-based algorithm AdaBoost.MRF for this task. Distinct from most existing work, our algorithm can handle hidden variables (missing labels) and is particularly attractive for smarthouse domains where reliable labels are often sparsely observed. Furthermore, our method works exclusively on trees and thus is guaranteed to converge. We apply the AdaBoost.MRF algorithm to a home video surveillance application and demonstrate its efficacy.

Random Subspace Two-Dimensional PCA for face recognition

The two-dimensional Principal Component Analysis (2DPCA) is a robust method in face recognition. Much recent research shows that the 2DPCA is more reliable than the well-known PCA method in recognising human face. However, in many cases, this method tends to be overfitted to sample data. In this paper, we proposed a novel method named random subspace two-dimensional PCA (RS-2DPCA), which combines the 2DPCA method with the random subspace (RS) technique. The RS-2DPCA inherits the advantages of both the 2DPCA and RS technique, thus it can avoid the overfitting problem and achieve high recognition accuracy. Experimental results in three benchmark face data sets -the ORL database, the Yale face database and the extended Yale face database B - confirm our hypothesis that the RS-2DPCA is superior to the 2DPCA itself.

Hierarchical semi-markov conditional random fields for recursive sequential data

Inspired by the hierarchical hidden Markov models (HHMM), we present the hierarchical semi-Markov conditional random field (HSCRF), a generalisation of embedded undirected Markov chains to model complex hierarchical, nested Markov processes. It is parameterised in a discriminative framework and has polynomial time algorithms for learning and inference. Importantly, we develop efficient algorithms for learning and constrained inference in a partially-supervised setting, which is important issue in practice where labels can only be obtained sparsely. We demonstrate the HSCRF in two applications: (i) recognising human activities of daily living (ADLs) from indoor surveillance cameras, and (ii) noun-phrase chunking. We show that the HSCRF is capable of learning rich hierarchical models with reasonable accuracy in both fully and partially observed data cases.

The L(2,1)-labelling problem for cubic Cayley graphs on dihedral groups

A k-L(2,1)-labelling of a graph G is a mapping f:V(G)→{0,1,2,…,k} such that |f(u)−f(v)|≥2 if uv∈E(G) and f(u)≠f(v) if u,v are distance two apart. The smallest positive integer k such that G admits a k-L(2,1)-labelling is called the λ-number of G. In this paper we study this quantity for cubic Cayley graphs (other than the prism graphs) on dihedral groups, which are called brick product graphs or honeycomb toroidal graphs. We prove that the λ-number of such a graph is between 5 and 7, and moreover we give a characterisation of such graphs with λ-number 5.

Are most samples of animals systematically biased? Consistent individual trait differences bias samples despite random sampling

Sampling animals from the wild for study is something nearly every biologist has done, but despite our best efforts to obtain random samples of animals, ‘hidden’ trait biases may still exist. For example, consistent behavioral traits can affect trappability/catchability, independent of obvious factors such as size and gender, and these traits are often correlated with other repeatable physiological and/or life history traits. If so, systematic sampling bias may exist for any of these traits. The extent to which this is a problem, of course, depends on the magnitude of bias, which is presently unknown because the underlying trait distributions in populations are usually unknown, or unknowable. Indeed, our present knowledge about sampling bias comes from samples (not complete population censuses), which can possess bias to begin with. I had the unique opportunity to create naturalized populations of fish by seeding each of four small fishless lakes with equal densities of slow-, intermediate-, and fast-growing fish. Using sampling methods that are not size-selective, I observed that fast-growing fish were up to two-times more likely to be sampled than slower-growing fish. This indicates substantial and systematic bias with respect to an important life history trait (growth rate). If correlations between behavioral, physiological and life-history traits are as widespread as the literature suggests, then many animal samples may be systematically biased with respect to these traits (e.g., when collecting animals for laboratory use), and affect our inferences about population structure and abundance. I conclude with a discussion on ways to minimize sampling bias for particular physiological/behavioral/life-history types within animal populations.