Parameter estimation of one-class SVM on imbalance text classification

Compared with conventional two-class learning schemes, one-class classification simply uses a single class for training purposes. Applying one-class classification to the minorities in an imbalanced data has been shown to achieve better performance than the two-class one. In this paper, in order to make the best use of all the available information during the learning procedure, we propose a general framework which first uses the minority class for training in the one-class classification stage; and then uses both minority and majority class for estimating the generalization performance of the constructed classifier. Based upon this generalization performance measurement, parameter search algorithm selects the best parameter settings for this classifier. Experiments on UCI and Reuters text data show that one-class SVM embedded in this framework achieves much better performance than the standard one-class SVM alone and other learning schemes, such as one-class Naive Bayes, one-class nearest neighbour and neural network.

Song sharing and repertoire size in the sedge warbler, Acrocephalus schoenobaenus: changes within and between years

The complex song of the male sedge warbler functions mainly in sexual attraction and the evolution of repertoire size is driven primarily by female choice. As male song ceases upon pairing, male–male singing interactions are relatively brief and have not been studied to our knowledge. This study shows that young males in their first breeding season shared significantly more syllables with their nearest neighbour than with their fathers or more distant males. Moreover, daily recordings revealed that rapid learning and modification of syllable repertoires occurred, resulting in a progressive increase in sharing within just a few days. This does not lead to a gradual increase in repertoire size as some syllables are dropped and new ones are acquired. This turnover process allows males to share syllables with their neighbours, whilst repertoire size, known to be important in female choice, remains relatively constant in any one year. Individual males were followed for several years and also showed considerable syllable turnover between years. However, in this case, repertoire size was found to increase between years, the largest increase occurring between the first and second years. We obtained a significant positive correlation between repertoire size and age, suggesting that females choosing males with larger repertoires may gain indirect (genetic) benefits for their offspring, such as good genes for viability. Whilst these results reveal a more flexible picture of repertoire turnover than previously suspected, the relative stability of repertoire size within a season and the increase with age suggests that repertoire size remains a likely target for sexual selection by female choice.

An automated classification system based on the strings of trojan and virus families

Classifying malware correctly is an important research issue for anti-malware software producers. This paper presents an effective and efficient malware classification technique based on string information using several wellknown classification algorithms. In our testing we extracted the printable strings from 1367 samples, including unpacked trojans and viruses and clean files. Information describing the printable strings contained in each sample was input to various classification algorithms, including treebased classifiers, a nearest neighbour algorithm, statistical algorithms and AdaBoost. Using k-fold cross validation on the unpacked malware and clean files, we achieved a classification accuracy of 97%. Our results reveal that strings from library code (rather than malicious code itself) can be utilised to distinguish different malware families.

Dimensionality reduction of protein mass spectrometry data using random projection

Protein mass spectrometry (MS) pattern recognition has recently emerged as a new method for cancer diagnosis. Unfortunately, classification performance may degrade owing to the enormously high dimensionality of the data. This paper investigates the use of Random Projection in protein MS data dimensionality reduction. The effectiveness of Random Projection (RP) is analyzed and compared against Principal Component Analysis (PCA) by using three classification algorithms, namely Support Vector Machine, Feed-forward Neural Networks and K-Nearest Neighbour. Three real-world cancer data sets are employed to evaluate the performances of RP and PCA. Through the investigations, RP method demonstrated better or at least comparable classification performance as PCA if the dimensionality of the projection matrix is sufficiently large. This paper also explores the use of RP as a pre-processing step prior to PCA. The results show that without sacrificing classification accuracy, performing RP prior to PCA significantly improves the computational time.

Use of circle-segments as a data visualization technique for feature selection in pattern classification

One of the issues associated with pattern classification using data based machine learning systems is the “curse of dimensionality”. In this paper, the circle-segments method is proposed as a feature selection method to identify important input features before the entire data set is provided for learning with machine learning systems. Specifically, four machine learning systems are deployed for classification, viz. Multilayer Perceptron (MLP), Support Vector Machine (SVM), Fuzzy ARTMAP (FAM), and k-Nearest Neighbour (kNN). The integration between the circle-segments method and the machine learning systems has been applied to two case studies comprising one benchmark and one real data sets. Overall, the results after feature selection using the circle segments method demonstrate improvements in performance even with more than 50% of the input features eliminated from the original data sets.

AdaM : adaptive-maximum imputation for neighborhood-based collaborative filtering

In the context of collaborative filtering, the well known data sparsity issue makes two like-minded users have little similarity, and consequently renders the k nearest neighbour rule inapplicable. In this paper, we address the data sparsity problem in the neighbourhood-based CF methods by proposing an Adaptive-Maximum imputation method (AdaM). The basic idea is to identify an imputation area that can maximize the imputation benefit for recommendation purposes, while minimizing the imputation error brought in. To achieve the maximum imputation benefit, the imputation area is determined from both the user and the item perspectives; to minimize the imputation error, there is at least one real rating preserved for each item in the identified imputation area. A theoretical analysis is provided to prove that the proposed imputation method outperforms the conventional neighbourhood-based CF methods through more accurate neighbour identification. Experiment results on benchmark datasets show that the proposed method significantly outperforms the other related state-of-the-art imputation-based methods in terms of accuracy.

Emergent effects in massive agent swarms in real-time game environments

 Computational efficiency and hence the scale of agent-based swarm simulations is bound by the nearest neighbour computation for each agent. This article proposes the use of GPU texture memory to implement lookup tables for a spatial partitioning based k-Nearest Neighbours algorithm. These improvements allow simulation of swarms of 220 agents at higher rates than the current best alternative algorithms. This approach is incorporated into an existing framework for simulating steering behaviours allowing for a complete implementation of massive agent swarm simulations, with per agent behaviour preferences, on a Graphics Processing Unit. These simulations have enabled an investigation of the emergent dynamics that occur when massive swarms interact with a choke point in their environment. Various modes of sustained dynamics with temporal and spatial coherence are identified when a critical mass of agents is simulated and some elementary properties are presented. The algorithms presented in this article enable researchers and content designers in games and movies to implement truly massive agent swarms in real time and thus provide a basis for further identification and analysis of the emergent dynamics in these swarms. This will improve not only the scale of swarms used in commercial games and movies but will also improve the reliability of swarm behaviour with respect to content design goals.

Knowing Indonesia : intersections of self, discipline and nation

 This is the first book in almost two decades to bring together scholars of Indonesia from the Australian academy in a single volume to reflect on and engage in a deep critique of their field. This is a timely contribution. The importance of Indonesia to Australia has never been more acute and it is essential that we have the tools for interpreting and understanding our nearest neighbour. Investigation of debates within the field of Indonesian studies will help us interpret better the perceptions and politics informing our study. As is befitting the multi-disciplinary nature of Indonesian studies, the book brings together leading political scientists, historians and anthropologists to give their unique perspectives and analysis of this field in the Australian academy and elsewhere in the West. This approach results in some divergent views on the fundamental questions of how Indonesia should be studied and the uses of Indonesia knowledge for activism, and presents new ideas about how we might pursue our work in the future.

Sparse representation with multi-manifold analysis for texture classification from few training images

Texture classification is one of the most important tasks in computer vision field and it has been extensively investigated in the last several decades. Previous texture classification methods mainly used the template matching based methods such as Support Vector Machine and k-Nearest-Neighbour for classification. Given enough training images the state-of-the-art texture classification methods could achieve very high classification accuracies on some benchmark databases. However, when the number of training images is limited, which usually happens in real-world applications because of the high cost of obtaining labelled data, the classification accuracies of those state-of-the-art methods would deteriorate due to the overfitting effect. In this paper we aim to develop a novel framework that could correctly classify textural images with only a small number of training images. By taking into account the repetition and sparsity property of textures we propose a sparse representation based multi-manifold analysis framework for texture classification from few training images. A set of new training samples are generated from each training image by a scale and spatial pyramid, and then the training samples belonging to each class are modelled by a manifold based on sparse representation. We learn a dictionary of sparse representation and a projection matrix for each class and classify the test images based on the projected reconstruction errors. The framework provides a more compact model than the template matching based texture classification methods, and mitigates the overfitting effect. Experimental results show that the proposed method could achieve reasonably high generalization capability even with as few as 3 training images, and significantly outperforms the state-of-the-art texture classification approaches on three benchmark datasets. © 2014 Elsevier B.V. All rights reserved.

Human rights in Papua New Guinea: is this where we should be settling refugees?

Australia has had a long, and at times tumultuous, relationship with our nearest neighbour, Papua New Guinea. This relationship took a twist in late 2012, with the re-opening of the off-shore processing centre on Manus Island, and again in February 2014, when Iranian asylum seeker Reza Berati was murdered by locals during a violent disturbance at the centre. The latest test of the strength and endurance of the relationship between PNG and Australia came in April 2016, when the PNG Supreme Court ruled that the detention of asylum seekers on Manus Island breached the right to personal liberty in the PNG constitution. This article provides much-needed insight into the human rights situation in PNG, and makes recommendations regarding the prospect of resettling refugees in that country.

Acoustic features involved in the neighbour-stranger vocal recognition process in male Australian fur seals

Many territorial species have the ability to recognise neighbours from stranger individuals. If the neighbouring individual is assumed to pose less of a threat, the territorial individual responds less and avoids unnecessary confrontations with familiar individuals at established boundaries, thus avoiding the costly energy expenditure associated with fighting. Territorial male Australian fur seals respond more to strangers than to neighbouring males. The present study evaluated which acoustic features were important in the neighbour–stranger recognition process in male Australian fur seals. The results reveal that there was an increase in response strength or intensity from males when they heard more bark units, indicating the importance of repetition to detect a caller. However, lengthening and shortening the inter-unit spaces, (i.e. changing the rhythm of the call) did not appear to significantly affect an animal's response. In addition, the whole frequency spectrum was considered important to recognition with results suggesting that they may vary in their importance. A call containing the dominant and surrounding harmonics was considered important to a male's ability to recognise its neighbour. Furthermore, recognition occurs even with a partial bark, but males need to hear between 25 and 75% of each bark unit from neighbouring seals. Our study highlights which acoustic features induce stronger or weaker responses from territorial males, decoding the important features in neighbour–stranger recognition.

Managing data using neighbour replication on a triangular-grid structure

Data in grid research deals with storage, replication, and management of large data sets in a distributed environment. The all-data-to-all-sites replication schemes, like Read-One Write-All (ROWA) and Tree Grid Structure (TGS), are the popular techniques in grid. However, these techniques have a weakness in data storage capacity and data access times. In this paper, we propose the all-data-to-some-sites scheme called the 'Neighbour Replication on Triangular Grid' (NRTG) technique. The proposed scheme minimises the storage capacity as well as data access time with high update availability. It also tolerates failures such as server and site failures.

Love thy neighbour? Associations of social capital and crime with physical activity amongst women

Using a multilevel study design, this study examined the associations between social characteristics of individuals and neighbourhoods and physical activity among women. Women (n = 1405) recruited from 45 Melbourne (Australia) neighbourhoods of varying socioeconomic disadvantage provided data on social factors and leisure-time: physical activity; walking; and walking in one’s own neighbourhood. Individual level social factors were number of neighbours known and social participation. Neighbourhood-level social characteristics (interpersonal trust, norms of reciprocity, social cohesion) were derived by aggregating survey data on these constructs within neighbourhoods. Objective data on crimes within neighbourhoods were obtained from Victoria Police. In bivariable regression models, all social variables at both the individual and neighbourhood level were positively associated with odds of physical activity, walking, and walking in one’s own neighbourhood. Associations with individual social participation (associated with all three physical activity variables) and neighbourhood interpersonal trust (associated with overall physical activity only) remained significant in multivariable models. Neither neighbourhood crime against the person nor incivilities were associated with any form of physical activity. These results demonstrate that women who participated in local groups or events and, less consistently, women living in neighbourhoods where residents trusted one another, were more likely to participate in leisure-time physical activity. While redressing macro-level social and economic policies that contribute to neighbourhood inequalities remains a priority, public health initiatives aimed at promoting physical activity could consider focusing on fostering social interactions targeting both individuals and communities. Further investigation of causal mechanisms underlying these associations is required.

The good neighbour: Australian peace support operations in the Pacific Islands 1980-2006

The Good Neighbour explores the Australian government's efforts to support peace in the Pacific Islands from 1980 to 2006. It tells the story of the deployment of Australian diplomatic, military and policing resources at a time when neighbouring governments were under pressure from political violence and civil unrest. The main focus of this volume is Australian peacemaking and peacekeeping in response to the Bougainville Crisis, a secessionist rebellion that began in late 1988 with the sabotage of a major mining operation. Following a signed peace agreement in 2001, the crisis finally ended in December 2005, under the auspices of the United Nations. During this time Australia's involvement shifted from behind-the-scenes peacemaking, to armed peacekeeping intervention, and finally to a longer-term unarmed regional peacekeeping operation. Granted full access to all relevant government files, Bob Breen recounts the Australian story from decisions made in Canberra to the planning and conduct of operations.