A biologically inspired object spectral-texture descriptor and its application to vegetation classification in power-line corridors

The use of appropriate features to represent an output class or object is critical for all classification problems. In this paper, we propose a biologically inspired object descriptor to represent the spectral-texture patterns of image-objects. The proposed feature descriptor is generated from the pulse spectral frequencies (PSF) of a pulse coupled neural network (PCNN), which is invariant to rotation, translation and small scale changes. The proposed method is first evaluated in a rotation and scale invariant texture classification using USC-SIPI texture database. It is further evaluated in an application of vegetation species classification in power line corridor monitoring using airborne multi-spectral aerial imagery. The results from the two experiments demonstrate that the PSF feature is effective to represent spectral-texture patterns of objects and it shows better results than classic color histogram and texture features.

Trust and reputation management in web-based social network

know personally. They also communicate with other members of the network who are the friends of their friends and may be friends of their friend’s network. They share their experiences and opinions within the social network about an item which may be a product or service. The user faces the problem of evaluating trust in a service or service provider before making a choice. Opinions, reputations and ecommendations will influence users' choice and usage of online resources. Recommendations may be received through a chain of friends of friends, so the problem for the user is to be able to evaluate various types of trust recommendations and reputations. This opinion or ecommendation has a great influence to choose to use or enjoy the item by the other user of the community. Users share information on the level of trust they explicitly assign to other users. This trust can be used to determine while taking decision based on any recommendation. In case of the absence of direct connection of the recommender user, propagated trust could be useful.

Ontology-based specific and exhaustive user profiles for constraint information fusion for multi-agents

Intelligent agents are an advanced technology utilized in Web Intelligence. When searching information from a distributed Web environment, information is retrieved by multi-agents on the client site and fused on the broker site. The current information fusion techniques rely on cooperation of agents to provide statistics. Such techniques are computationally expensive and unrealistic in the real world. In this paper, we introduce a model that uses a world ontology constructed from the Dewey Decimal Classification to acquire user profiles. By search using specific and exhaustive user profiles, information fusion techniques no longer rely on the statistics provided by agents. The model has been successfully evaluated using the large INEX data set simulating the distributed Web environment.

Multi-step ahead direct prediction for the machine condition prognosis using regression trees and neuro-fuzzy systems

This paper presents an approach to predict the operating conditions of machine based on classification and regression trees (CART) and adaptive neuro-fuzzy inference system (ANFIS) in association with direct prediction strategy for multi-step ahead prediction of time series techniques. In this study, the number of available observations and the number of predicted steps are initially determined by using false nearest neighbor method and auto mutual information technique, respectively. These values are subsequently utilized as inputs for prediction models to forecast the future values of the machines’ operating conditions. The performance of the proposed approach is then evaluated by using real trending data of low methane compressor. A comparative study of the predicted results obtained from CART and ANFIS models is also carried out to appraise the prediction capability of these models. The results show that the ANFIS prediction model can track the change in machine conditions and has the potential for using as a tool to machine fault prognosis.

Fault diagnosis of induction motor based on decision trees and adaptive neuro-fuzzy inference

This paper presents a fault diagnosis method based on adaptive neuro-fuzzy inference system (ANFIS) in combination with decision trees. Classification and regression tree (CART) which is one of the decision tree methods is used as a feature selection procedure to select pertinent features from data set. The crisp rules obtained from the decision tree are then converted to fuzzy if-then rules that are employed to identify the structure of ANFIS classifier. The hybrid of back-propagation and least squares algorithm are utilized to tune the parameters of the membership functions. In order to evaluate the proposed algorithm, the data sets obtained from vibration signals and current signals of the induction motors are used. The results indicate that the CART–ANFIS model has potential for fault diagnosis of induction motors.

Children with special health care needs : social and learning competence in the early years

The number of children with special health care needs surviving infancy and attending school has been increasing. Due to their health status, these children may be at risk of low social-emotional and learning competencies (e.g., Lightfoot, Mukherjee, & Sloper, 2000; Zehnder, Landolt, Prchal, & Vollrath, 2006). Early social problems have been linked to low levels of academic achievement (Ladd, 2005), inappropriate behaviours at school (Shiu, 2001) and strained teacher-child relationships (Blumberg, Carle, O‘Connor, Moore, & Lippmann, 2008). Early learning difficulties have been associated with mental health problems (Maughan, Rowe, Loeber, & Stouthamer-Loeber, 2003), increased behaviour issues (Arnold, 1997), delinquency (Loeber & Dishion, 1983) and later academic failure (Epstein, 2008). Considering the importance of these areas, the limited research on special health care needs in social-emotional and learning domains is a factor driving this research. The purpose of the current research is to investigate social-emotional and learning competence in the early years for Australian children who have special health care needs. The data which informed this thesis was from Growing up in Australia: The Longitudinal Study of Australian Children. This is a national, longitudinal study being conducted by the Commonwealth Department of Families, Housing, Community Services and Indigenous Affairs. The study has a national representative sample, with data collection occurring biennially, in 2004 (Wave 1), 2006 (Wave 2) and 2008 (Wave 3). Growing up in Australia uses a cross-sequential research design involving two cohorts, an Infant Cohort (0-1 at recruitment) and a Kindergarten Cohort (4-5 at recruitment). This study uses the Kindergarten Cohort, for which there were 4,983 children at recruitment. Three studies were conducted to address the objectives of this thesis. Study 1 used Wave 1 data to identify and describe Australian children with special health care needs. Children who identified as having special health care needs through the special health care needs screener were selected. From this, descriptive analyses were run. The results indicate that boys, children with low birth weight and children from families with low levels of maternal education are likely to be in the population of children with special health care needs. Further, these children are likely to be using prescription medications, have poor general health and are likely to have specific condition diagnoses. Study 2 used Wave 1 data to examine differences between children with special health care needs and their peers in social-emotional competence and learning competence prior to school. Children identified by the special health care needs screener were chosen for the case group (n = 650). A matched case control group of peers (n = 650), matched on sex, cultural and linguistic diversity, family socioeconomic position and age, were the comparison group. Social-emotional competence was measured through Social/Emotional Domain scores taken from the Growing up in Australia Outcome Index, with learning competence measured through Learning Domain scores. Results suggest statistically significant differences in scores between the two groups. Children with special health care needs have lower levels of social-emotional and learning competence prior to school compared to their peers. Study 3 used Wave 1 and Wave 2 data to examine the relationship between special health care needs at Wave 1 and social-emotional competence and learning competence at Wave 2, as children started school. The sample for this study consisted of children in the Kindergarten Cohort who had teacher data at Wave 2. Results from multiple regression models indicate that special health care needs prior to school (Wave 1) significantly predicts social-emotional competence and learning competence in the early years of school (Wave 2). These results indicate that having special health care needs prior to school is a risk factor for the social-emotional and learning domains in the early years of school. The results from these studies give valuable insight into Australian children with special health care needs and their social-emotional and learning competence in the early years. The Australia population of children with special health care needs were primarily male children, from families with low maternal education, were likely to be of poor health and taking prescription medications. It was found that children with special health care needs were likely to have lower social-emotional competence and learning competence prior to school compared to their peers. Results indicate that special health care needs prior to school were predictive of lower social-emotional and learning competencies in the early years of school. More research is required into this unique population and their competencies over time. However, the current research provides valuable insight into an under researched 'at risk' population.

Illegal street racing and associated (hooning) behaviours

In an Australian context, the term hooning refers to risky driving behaviours such as illegal street racing and speed trials, as well as behaviours that involve unnecessary noise and smoke, which include burn outs, donuts, fish tails, drifting and other skids. Hooning receives considerable negative media attention in Australia, and since the 1990s all Australian jurisdictions have implemented vehicle impoundment programs to deal with the problem. However, there is limited objective evidence of the road safety risk associated with hooning behaviours. Attempts to estimate the risk associated with hooning are limited by official data collection and storage practices, and the willingness of drivers to admit to their illegal behaviour in the event of a crash. International evidence suggests that illegal street racing is associated with only a small proportion of fatal crashes; however, hooning in an Australian context encompasses a broader group of driving behaviours than illegal street racing alone, and it is possible that the road safety risks will differ with these behaviours. There is evidence from North American jurisdictions that vehicle impoundment programs are effective for managing drink driving offenders, and drivers who continue to drive while disqualified or suspended both during and post-impoundment. However, these programs used impoundment periods of 30 – 180 days (depending on the number of previous offences). In Queensland the penalty for a first hooning offence is 48 hours, while the vehicle can be impounded for up to 3 months for a second offence, or permanently for a third or subsequent offence within three years. Thus, it remains unclear whether similar effects will be seen for hooning offenders in Australia, as no evaluations of vehicle impoundment programs for hooning have been published. To address these research needs, this program of research consisted of three complementary studies designed to: (1) investigate the road safety implications of hooning behaviours in terms of the risks associated with the specific behaviours, and the drivers who engage in these behaviours; and (2) assess the effectiveness of current approaches to dealing with the problem; in order to (3) inform policy and practice in the area of hooning behaviour. Study 1 involved qualitative (N = 22) and quantitative (N = 290) research with drivers who admitted engaging in hooning behaviours on Queensland roads. Study 2 involved a systematic profile of a large sample of drivers (N = 834) detected and punished for a hooning offence in Queensland, and a comparison of their driving and crash histories with a randomly sampled group of Queensland drivers with the same gender and age distribution. Study 3 examined the post-impoundment driving behaviour of hooning offenders (N = 610) to examine the effects of vehicle impoundment on driving behaviour. The theoretical framework used to guide the research incorporated expanded deterrence theory, social learning theory, and driver thrill-seeking perspectives. This framework was used to explore factors contributing to hooning behaviours, and interpret the results of the aspects of the research designed to explore the effectiveness of vehicle impoundment as a countermeasure for hooning. Variables from each of the perspectives were related to hooning measures, highlighting the complexity of the behaviour. This research found that the road safety risk of hooning behaviours appears low, as only a small proportion of the hooning offences in Study 2 resulted in a crash. However, Study 1 found that hooning-related crashes are less likely to be reported than general crashes, particularly when they do not involve an injury, and that higher frequencies of hooning behaviours are associated with hooning-related crash involvement. Further, approximately one fifth of drivers in Study 1 reported being involved in a hooning-related crash in the previous three years, which is comparable to general crash involvement among the general population of drivers in Queensland. Given that hooning-related crashes represented only a sub-set of crash involvement for this sample, this suggests that there are risks associated with hooning behaviour that are not apparent in official data sources. Further, the main evidence of risk associated with the behaviour appears to relate to the hooning driver, as Study 2 found that these drivers are likely to engage in other risky driving behaviours (particularly speeding and driving vehicles with defects or illegal modifications), and have significantly more traffic infringements, licence sanctions and crashes than drivers of a similar (i.e., young) age. Self-report data from the Study 1 samples indicated that Queensland’s vehicle impoundment and forfeiture laws are perceived as severe, and that many drivers have reduced their hooning behaviour to avoid detection. However, it appears that it is more common for drivers to have simply changed the location of their hooning behaviour to avoid detection. When the post-impoundment driving behaviour of the sample of hooning offenders was compared to their pre-impoundment behaviour to examine the effectiveness of vehicle impoundment in Study 3, it was found that there was a small but significant reduction in hooning offences, and also for other traffic infringements generally. As Study 3 was observational, it was not possible to control for extraneous variables, and is, therefore, possible that some of this reduction was due to other factors, such as a reduction in driving exposure, the effects of changes to Queensland’s Graduated Driver Licensing scheme that were implemented during the study period and affected many drivers in the offender sample due to their age, or the extension of vehicle impoundment to other types of offences in Queensland during the post-impoundment period. However, there was a protective effect observed, in that hooning offenders did not show the increase in traffic infringements in the post period that occurred within the comparison sample. This suggests that there may be some effect of vehicle impoundment on the driving behaviour of hooning offenders, and that this effect is not limited to their hooning driving behaviour. To be more confident in these results, it is necessary to measure driving exposure during the post periods to control for issues such as offenders being denied access to vehicles. While it was not the primary aim of this program of research to compare the utility of different theoretical perspectives, the findings of the research have a number of theoretical implications. For example, it was found that only some of the deterrence variables were related to hooning behaviours, and sometimes in the opposite direction to predictions. Further, social learning theory variables had stronger associations with hooning. These results suggest that a purely legal approach to understanding hooning behaviours, and designing and implementing countermeasures designed to reduce these behaviours, are unlikely to be successful. This research also had implications for policy and practice, and a number of recommendations were made throughout the thesis to improve the quality of relevant data collection practices. Some of these changes have already occurred since the expansion of the application of vehicle impoundment programs to other offences in Queensland. It was also recommended that the operational and resource costs of these laws should be compared to the road safety benefits in ongoing evaluations of effectiveness to ensure that finite traffic policing resources are allocated in a way that produces maximum road safety benefits. However, as the evidence of risk associated with the hooning driver is more compelling than that associated with hooning behaviour, it was argued that the hooning driver may represent the better target for intervention. Suggestions for future research include ongoing evaluations of the effectiveness of vehicle impoundment programs for hooning and other high-risk driving behaviours, and the exploration of additional potential targets for intervention to reduce hooning behaviour. As the body of knowledge regarding the factors contributing to hooning increases, along with the identification of potential barriers to the effectiveness of current countermeasures, recommendations for changes in policy and practice for hooning behaviours can be made.

Multi-modal summarization of key events and top players in sports tournament videos

To detect and annotate the key events of live sports videos, we need to tackle the semantic gaps of audio-visual information. Previous work has successfully extracted semantic from the time-stamped web match reports, which are synchronized with the video contents. However, web and social media articles with no time-stamps have not been fully leveraged, despite they are increasingly used to complement the coverage of major sporting tournaments. This paper aims to address this limitation using a novel multimodal summarization framework that is based on sentiment analysis and players' popularity. It uses audiovisual contents, web articles, blogs, and commentators' speech to automatically annotate and visualize the key events and key players in a sports tournament coverage. The experimental results demonstrate that the automatically generated video summaries are aligned with the events identified from the official website match reports.

Model selection and error estimation

We study model selection strategies based on penalized empirical loss minimization. We point out a tight relationship between error estimation and data-based complexity penalization: any good error estimate may be converted into a data-based penalty function and the performance of the estimate is governed by the quality of the error estimate. We consider several penalty functions, involving error estimates on independent test data, empirical VC dimension, empirical VC entropy, and margin-based quantities. We also consider the maximal difference between the error on the first half of the training data and the second half, and the expected maximal discrepancy, a closely related capacity estimate that can be calculated by Monte Carlo integration. Maximal discrepancy penalty functions are appealing for pattern classification problems, since their computation is equivalent to empirical risk minimization over the training data with some labels flipped.

Rademacher and Gaussian complexities: Risk bounds and structural results

We investigate the use of certain data-dependent estimates of the complexity of a function class, called Rademacher and Gaussian complexities. In a decision theoretic setting, we prove general risk bounds in terms of these complexities. We consider function classes that can be expressed as combinations of functions from basis classes and show how the Rademacher and Gaussian complexities of such a function class can be bounded in terms of the complexity of the basis classes. We give examples of the application of these techniques in finding data-dependent risk bounds for decision trees, neural networks and support vector machines.

Exponentiated gradient algorithms for conditional random fields and max-margin Markov Networks

Log-linear and maximum-margin models are two commonly-used methods in supervised machine learning, and are frequently used in structured prediction problems. Efficient learning of parameters in these models is therefore an important problem, and becomes a key factor when learning from very large data sets. This paper describes exponentiated gradient (EG) algorithms for training such models, where EG updates are applied to the convex dual of either the log-linear or max-margin objective function; the dual in both the log-linear and max-margin cases corresponds to minimizing a convex function with simplex constraints. We study both batch and online variants of the algorithm, and provide rates of convergence for both cases. In the max-margin case, O(1/ε) EG updates are required to reach a given accuracy ε in the dual; in contrast, for log-linear models only O(log(1/ε)) updates are required. For both the max-margin and log-linear cases, our bounds suggest that the online EG algorithm requires a factor of n less computation to reach a desired accuracy than the batch EG algorithm, where n is the number of training examples. Our experiments confirm that the online algorithms are much faster than the batch algorithms in practice. We describe how the EG updates factor in a convenient way for structured prediction problems, allowing the algorithms to be efficiently applied to problems such as sequence learning or natural language parsing. We perform extensive evaluation of the algorithms, comparing them to L-BFGS and stochastic gradient descent for log-linear models, and to SVM-Struct for max-margin models. The algorithms are applied to a multi-class problem as well as to a more complex large-scale parsing task. In all these settings, the EG algorithms presented here outperform the other methods.

Geometry vs appearance for discriminating between posed and spontaneous emotions

Spontaneous facial expressions differ from posed ones in appearance, timing and accompanying head movements. Still images cannot provide timing or head movement information directly. However, indirectly the distances between key points on a face extracted from a still image using active shape models can capture some movement and pose changes. This information is superposed on information about non-rigid facial movement that is also part of the expression. Does geometric information improve the discrimination between spontaneous and posed facial expressions arising from discrete emotions? We investigate the performance of a machine vision system for discrimination between posed and spontaneous versions of six basic emotions that uses SIFT appearance based features and FAP geometric features. Experimental results on the NVIE database demonstrate that fusion of geometric information leads only to marginal improvement over appearance features. Using fusion features, surprise is the easiest emotion (83.4% accuracy) to be distinguished, while disgust is the most difficult (76.1%). Our results find different important facial regions between discriminating posed versus spontaneous version of one emotion and classifying the same emotion versus other emotions. The distribution of the selected SIFT features shows that mouth is more important for sadness, while nose is more important for surprise, however, both the nose and mouth are important for disgust, fear, and happiness. Eyebrows, eyes, nose and mouth are important for anger.

Evaluation of texture and geometry for dimensional facial expression recognition

Facial expression recognition (FER) algorithms mainly focus on classification into a small discrete set of emotions or representation of emotions using facial action units (AUs). Dimensional representation of emotions as continuous values in an arousal-valence space is relatively less investigated. It is not fully known whether fusion of geometric and texture features will result in better dimensional representation of spontaneous emotions. Moreover, the performance of many previously proposed approaches to dimensional representation has not been evaluated thoroughly on publicly available databases. To address these limitations, this paper presents an evaluation framework for dimensional representation of spontaneous facial expressions using texture and geometric features. SIFT, Gabor and LBP features are extracted around facial fiducial points and fused with FAP distance features. The CFS algorithm is adopted for discriminative texture feature selection. Experimental results evaluated on the publicly accessible NVIE database demonstrate that fusion of texture and geometry does not lead to a much better performance than using texture alone, but does result in a significant performance improvement over geometry alone. LBP features perform the best when fused with geometric features. Distributions of arousal and valence for different emotions obtained via the feature extraction process are compared with those obtained from subjective ground truth values assigned by viewers. Predicted valence is found to have a more similar distribution to ground truth than arousal in terms of covariance or Bhattacharya distance, but it shows a greater distance between the means.