Improving the performance of model-order selection criteria by partial-model selection search

The traditional searching method for model-order selection in linear regression is a nested full-parameters-set searching procedure over the desired orders, which we call full-model order selection. On the other hand, a method for model-selection searches for the best sub-model within each order. In this paper, we propose using the model-selection searching method for model-order selection, which we call partial-model order selection. We show by simulations that the proposed searching method gives better accuracies than the traditional one, especially for low signal-to-noise ratios over a wide range of model-order selection criteria (both information theoretic based and bootstrap-based). Also, we show that for some models the performance of the bootstrap-based criterion improves significantly by using the proposed partial-model selection searching method. Index Terms— Model order estimation, model selection, information theoretic criteria, bootstrap 1. INTRODUCTION Several model-order selection criteria can be applied to find the optimal order. Some of the more commonly used information theoretic-based procedures include Akaike’s information criterion (AIC) [1], corrected Akaike (AICc) [2], minimum description length (MDL) [3], normalized maximum likelihood (NML) [4], Hannan-Quinn criterion (HQC) [5], conditional model-order estimation (CME) [6], and the efficient detection criterion (EDC) [7]. From a practical point of view, it is difficult to decide which model order selection criterion to use. Many of them perform reasonably well when the signal-to-noise ratio (SNR) is high. The discrepancies in their performance, however, become more evident when the SNR is low. In those situations, the performance of the given technique is not only determined by the model structure (say a polynomial trend versus a Fourier series) but, more importantly, by the relative values of the parameters within the model. This makes the comparison between the model-order selection algorithms difficult as within the same model with a given order one could find an example for which one of the methods performs favourably well or fails [6, 8]. Our aim is to improve the performance of the model order selection criteria in cases where the SNR is low by considering a model-selection searching procedure that takes into account not only the full-model order search but also a partial model order search within the given model order. Understandably, the improvement in the performance of the model order estimation is at the expense of additional computational complexity.

The evaluation of qualitative methods selection in the field of design and emotion

The aim of this paper is to aid researchers in selecting appropriate qualitative methods in order to develop and improve future studies in the field of emotional design. These include observations, think-aloud protocols, questionnaires, diaries and interviews. Based on the authors’ experiences, it is proposed that the methods under review can be successfully used for collecting data on emotional responses to evaluate user product relationships. This paper reviews the methods; discusses the suitability, advantages and challenges in relation to design and emotion studies. Furthermore, the paper outlines the potential impact of technology on the application of these methods, discusses the implications of these methods for emotion research and concludes with recommendations for future work in this area.

Model-order selection in Zernike polynomial expansion of corneal surfaces using the efficient detection criterion

Corneal-height data are typically measured with videokeratoscopes and modeled using a set of orthogonal Zernike polynomials. We address the estimation of the number of Zernike polynomials, which is formalized as a model-order selection problem in linear regression. Classical information-theoretic criteria tend to overestimate the corneal surface due to the weakness of their penalty functions, while bootstrap-based techniques tend to underestimate the surface or require extensive processing. In this paper, we propose to use the efficient detection criterion (EDC), which has the same general form of information-theoretic-based criteria, as an alternative to estimating the optimal number of Zernike polynomials. We first show, via simulations, that the EDC outperforms a large number of information-theoretic criteria and resampling-based techniques. We then illustrate that using the EDC for real corneas results in models that are in closer agreement with clinical expectations and provides means for distinguishing normal corneal surfaces from astigmatic and keratoconic surfaces.

Psychiatric and substance abuse comorbidity among a large sample of DUI Texas offenders

Research is indicating that individuals who present for DUI treatment may have competing substance abuse and mental health needs. This study aimed to examine the extent of such comorbidity issues among a sample of Texas DUI offenders. Method: Records of 36,372 DUI clients and 308,695 non-DUI clients admitted to Texas treatment programs between 2005 and 2008 were obtained from the State's administrative dataset. The data were analysed to identify the relationship between substance use, psychiatric problems, program completion and recidivism rates. Results: Analysis indicated that while non-DUI clients were more likely to present with more severe illicit substance use problems, DUI clients were more likely to have a primary problem with alcohol. Additionally, a cannabis use problem was also found to be significantly associated with DUI recidivism in the last year. In regards to mental health needs, a major finding was that depression was the most common psychiatric condition reported by DUI clients, including those with more than one DUI offence in the past year. This group were also more at risk of being diagnosed with Bipolar Disorder compared to the general population, and such a diagnosis was also associated with an increased likelihood of not completing treatment. Interestingly, female DUI and non-DUI clients were also more likely to be diagnosed with mental health problems compared to males, as well as more likely to be placed on medications at admission and have problems with methamphetamine, cocaine, and opiates. Conclusion: The findings highlight the complex competing needs of some DUI offenders who enter treatment. The results also suggest that there is a need to utilise mental health and substance abuse screening methods to ensure DUI offenders are directed towards appropriate treatment pathways as well as ensure that such interventions adequately cater for complex substance abuse and psychiatric needs.

Assessing cognitive development in early childhood : a comparison of the Bayley-III and the Stanford-Binet fifth edition

The Bayley Scales of Infant Development, Third Edition (Bayley-III) and Stanford-Binet Intelligence Scale, Fifth Edition (SB5) were administered in a sample of 26 typically developing children (12 males and 14 females) aged 24 – 42 months. Children completed the assessments in two separate sessions, counterbalanced for order of administration. Scores on the two instruments were not significantly related, with the exception of the SB5 Knowledge score, which was moderately correlated with the Language score on the Bayley-III (r = .41, p = 0.04). Despite no other significant correlations, for 22 of the 26 children, scores were very consistent across the two instruments. Implications for test selection are discussed.

A site selection process for environmental offsetting purposes in Queensland

A method of selecting land in any region of Queensland for offsetting purposes is devised, employing uniform standards. The procedure first requires that any core natural asset lands, Crown environmental lands, prime urban and agricultural lands, and highly contentious sites in the region be eliminated from consideration. Other land is then sought that is located between existing large reservations and the centre of greatest potential regional development/disturbance. Using the criteria of rehabilitation (rather than preservation) plus proximity to those officially defined Regional Ecosystems that are most threatened, adjacent sites that are described as ‘Cleared’ are identified in terms of agricultural land capability. Class IV lands – defined as those ‘which may be safely used for occasional cultivation with careful management’,2 ‘where it is favourably located for special usage’,3 and where it is ‘helpful to those who are interested in industry or regional planning or in reconstruction’4 – are examined for their appropriate area, for current tenure and for any conditions such as Mining Leases that may exist. The positive impacts from offsets on adjoining lands can then be designed to be significant; examples are also offered in respect of riparian areas and of Marine Parks. Criteria against which to measure performance for trading purposes include functional lift, with other case studies about this matter reported separately in this issue. The procedure takes no account of demand side economics (financial additionality), which requires commercial rather than environmental analysis.

Incidence, severity and correlates of bicycling injuries in a sample of cyclists in Queensland, Australia

Bicycle injuries, particularly those resulting from single bicycle crashes, are underreported in both police and hospital records. Data on cyclist characteristics and crash circumstances are also often lacking. As a result, the ability to develop comprehensive injury prevention policies is hampered. The aim of this study was to examine the incidence, severity, cyclist characteristics, and crash circumstances associated with cycling injuries in a sample of cyclists in Queensland, Australia. A cross-sectional study of Queensland cyclists was conducted in 2009. Respondents (n=2056) completed an online survey about their cycling experiences, including cycling injuries. Logistic regression modelling was used to examine the associations between demographic and cycling behaviour variables with experiencing cycling injuries in the past year, and, separately, with serious cycling injuries requiring a trip to a hospital. Twenty-seven percent of respondents (n=545) reported injuries, and 6% (n=114) reported serious injuries. In multivariable modelling, reporting an injury was more likely for respondents who had cycled <5 years, compared to ≥10 years (p<0.005); cycled for competition (p=0.01); or experienced harassment from motor vehicle occupants (p<0.001). There were no gender differences in injury incidence, and respondents who cycled for transport did not have an increased risk of injury. Reporting a serious injury was more likely for those whose injury involved other road users (p<0.03). Along with environmental and behavioural approaches for reducing collisions and near-collisions with motor vehicles, interventions that improve the design and maintenance of cycling infrastructure, increase cyclists’ skills, and encourage safe cycling behaviours and bicycle maintenance will also be important for reducing the overall incidence of cycling injuries.

An examination of the relationship between emotional intelligence, leadership style and perceived leadership outcomes in Australian educational institutions

In the field of leadership studies transformational leadership theory (e.g., Bass, 1985; Avolio, Bass, & Jung, 1995) has received much attention from researchers in recent years (Hughes, Ginnet, & Curphy, 2009; Hunt, 1999). Many previous studies have found that transformational leadership is related to positive outcomes such as the satisfaction, motivation and performance of followers in organisations (Judge & Piccolo, 2004; Lowe, Kroeck, & Sivasubramaniam, 1996), including in educational institutions (Chin, 2007; Leithwoood & Jantzi, 2005). Hence, it is important to explore constructs that may predict leadership style in order to identify potential transformational leaders in leadership assessment and selection procedures. Several researchers have proposed that emotional intelligence (EI) is one construct that may account for hitherto unexplained variance in transformational leadership (Mayer, 2001; Watkin, 2000). Different models of EI exist (e.g., Goleman, 1995, 2001; Bar-On, 1997; Mayer & Salovey, 1997) but momentum is growing for the Mayer and Salovey (1997) model to be considered the most useful (Ashkanasy & Daus, 2005; Daus & Ashkanasy, 2005). Studies in non-educational settings claim to have found that EI is a useful predictor of leadership style and leader effectiveness (Harms & Crede, 2010; Mills, 2009) but there is a paucity of studies which have examined the Mayer and Salovey (1997) model of EI in educational settings. Furthermore, other predictor variables have rarely been controlled in previous studies and only self-ratings of leadership behaviours, rather than multiple ratings, have usually been obtained. Therefore, more research is required in educational settings to answer the question: to what extent is the Mayer and Salovey (1997) model of EI a useful predictor of leadership style and leadership outcomes? This project, set in Australian educational institutions, was designed to move research in the field forward by: using valid and reliable instruments, controlling for other predictors, obtaining an adequately sized sample of real leaders as participants and obtaining multiple ratings of leadership behaviours. Other variables commonly used to predict leadership behaviours (personality factors and general mental ability) were assessed and controlled in the project. Additionally, integrity was included as another potential predictor of leadership behaviours as it has previously been found to be related to transformational leadership (Parry & Proctor-Thomson, 2002). Multiple ratings of leadership behaviours were obtained from each leader and their supervisors, peers and followers. The following valid and reliable psychological tests were used to operationalise the variables of interest: leadership styles and perceived leadership outcomes (Multifactor Leadership Questionnaire, Avolio et al., 1995), EI (Mayer–Salovey–Caruso Emotional Intelligence Test, Mayer, Salovey, & Caruso, 2002), personality factors (The Big Five Inventory, John, Donahue, & Kentle, 1991), general mental ability (Wonderlic Personnel Test-Quicktest, Wonderlic, 2003) and integrity (Integrity Express, Vangent, 2002). A Pilot Study (N = 25 leaders and 75 raters) made a preliminary examination of the relationship between the variables included in the project. Total EI, the experiential area, and the managing emotions and perceiving emotions branches of EI, were found to be related to transformational leadership which indicated that further research was warranted. In the Main Study, 144 leaders and 432 raters were recruited as participants to assess the discriminant validity of the instruments and examine the usefulness of EI as a predictor of leadership style and perceived leadership outcomes. Scores for each leadership scale across the four rating levels (leaders, supervisors, peers and followers) were aggregated with the exception of the management-by-exception active scale of transactional leadership which had an inadequate level of interrater agreement. In the descriptive and measurement component of the Main Study, the instruments were found to demonstrate adequate discriminant validity. The impact of role and gender on leadership style and EI were also examined, and females were found to be more transformational as leaders than males. Females also engaged in more contingent reward (transactional leadership) behaviours than males, whilst males engaged in more passive/avoidant leadership behaviours than females. In the inferential component of the Main Study, multiple regression procedures were used to examine the usefulness of EI as a predictor of leadership style and perceived leadership outcomes. None of the EI branches were found to be related to transformational leadership or the perceived leadership outcomes variables included in the study. Openness, emotional stability (the inverse of neuroticism) and general mental ability (inversely) each predicted a small amount of variance in transformational leadership. Passive/avoidant leadership was inversely predicted by the understanding emotions branch of EI. Overall, EI was not found to be a useful predictor of leadership style and leadership outcomes in the Main Study of this project. Implications for researchers and human resource practitioners are discussed.

Resonance : a music director’s theomusicological study of ‘praise’ and ‘glory’ in a selection of four liturgical music excerpts

Some of my most powerful spiritual experiences have come from the splendorous and sublime sounding hymns performed by a choir and church organ at the traditional Anglican church I’ve attended since I was very young. In the later stage of my life, my pursuit of education in the field of engineering caused me to move to Australia where I regularly attended a contemporary evangelical church and subsequently became a music director in the faith community. This environmental and cultural shift altered my perception and musical experiences of Christian music and led me to enquire about the relationship between Christian liturgy and church music. Throughout history church musicians and composers have synthesised the theological, congregational, cultural and musical aspects of church liturgy. Many great composers have taken into account the conditions surrounding the process of sacred composition and arrangement of music to enhance the experience of religious ecstasy – they sought resonances with Christian values and beliefs to draw congregational participation into the light of praising and glorifying God. As a music director in an evangelical church this aspiration has become one I share. I hope to identify and define the qualities of these resonances that have been successful and apply them to my own practice. Introduction and Structure of the Thesis In this study I will examine four purposively selected excerpts of Christian church vocal music combining theomusicological and semiotic analysis to help identify guidelines that might be useful in my practice as a church music director. The four musical excerpts have been selected based upon their sustained musical and theological impact over time, and their ability to affect ecstatic responses from congregations. This thesis documents a personal journey through analysis of music and uses a context that draws upon ethno-musicological, theological and semiotic tools that lead to a preliminary framework and principles which can then be applied to the identified qualities of resonance in church music today. The thesis is comprised of four parts. Part 1 presents a literature study on the relationship between sacred music, the effects of religious ecstasy and the Christian church. Multiple lenses on this phenomenon are drawn from the viewpoints of prominent western church historians, Biblical theologians, and philosophers. The literature study continues in Part 2, where the role of embodiment is examined from the current perspective of cognitive learning environments. This study offers a platform for a critical reflection on two distinctive musical liturgical systems that have treated differently the notion of embodied understanding amidst a shifting church paradigm. This allows an in-depth theological and philosophical understanding of the liturgical conditions around sacred music-making that relates to the monistic and dualistic body/mind. Part 3 involves undertaking a theomusicological methodology that utilises creative case studies of four purposively selected spiritual pieces. A semiotic study focuses on specific sections of sacred vocal works that express the notions of ‘praise’ and ‘glorification’, particularly in relation to these effects,which combine an analysis of theological perspectives around religious ecstasy and particular spiritual themes. Part 4 presents the critiques and findings gathered from the study that incorporate theoretical and technological means to analyse the purposive selected musical artefact, particularly with the sonic narratives expressing notions of ‘Praise' and 'Glory’. The musical findings are further discussed in relation to the notion of resonance, and then a conceptual framework for the role of contemporary musicdirector is proposed. The musical and Christian terminologies used in the thesis are explained in the glossary, and the appendices includes tables illustrating the musical findings, conducted surveys, written musical analyses and audio examples of selected sacred pieces available on the enclosed compact disc.

Selection of solar energy for green building using superiority and inferiority multi-criteria ranking (SIR) method

Green energy is one of the key factors, driving down electricity bill and zero carbon emission generating electricity to green building. However, the climate change and environmental policies are accelerating people to use renewable energy instead of coal-fired (convention type) energy for green building that energy is not environmental friendly. Therefore, solar energy is one of the clean energy solving environmental impact and paying less in electricity fee. The method of solar energy is collecting sun from solar array and saves in battery from which provides necessary electricity to whole house with zero carbon emission. However, in the market a lot of solar arrays suppliers, the aims of this paper attempted to use superiority and inferiority multi-criteria ranking (SIR) method with 13 constraints establishing I-flows and S-flows matrices to evaluate four alternatives solar energies and determining which alternative is the best, providing power to sustainable building. Furthermore, SIR is well-known structured approach of multi-criteria decision support tools and gradually used in construction and building. The outcome of this paper significantly gives an indication to user selecting solar energy.

The sample complexity of pattern classification with neural networks: the size of the weights is more important than the size of the network

Sample complexity results from computational learning theory, when applied to neural network learning for pattern classification problems, suggest that for good generalization performance the number of training examples should grow at least linearly with the number of adjustable parameters in the network. Results in this paper show that if a large neural network is used for a pattern classification problem and the learning algorithm finds a network with small weights that has small squared error on the training patterns, then the generalization performance depends on the size of the weights rather than the number of weights. For example, consider a two-layer feedforward network of sigmoid units, in which the sum of the magnitudes of the weights associated with each unit is bounded by A and the input dimension is n. We show that the misclassification probability is no more than a certain error estimate (that is related to squared error on the training set) plus A3 √((log n)/m) (ignoring log A and log m factors), where m is the number of training patterns. This may explain the generalization performance of neural networks, particularly when the number of training examples is considerably smaller than the number of weights. It also supports heuristics (such as weight decay and early stopping) that attempt to keep the weights small during training. The proof techniques appear to be useful for the analysis of other pattern classifiers: when the input domain is a totally bounded metric space, we use the same approach to give upper bounds on misclassification probability for classifiers with decision boundaries that are far from the training examples.

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.

Shifting : one-inclusion mistake bounds and sample compression

We present new expected risk bounds for binary and multiclass prediction, and resolve several recent conjectures on sample compressibility due to Kuzmin and Warmuth. By exploiting the combinatorial structure of concept class F, Haussler et al. achieved a VC(F)/n bound for the natural one-inclusion prediction strategy. The key step in their proof is a d = VC(F) bound on the graph density of a subgraph of the hypercube—oneinclusion graph. The first main result of this paper is a density bound of n [n−1 <=d-1]/[n <=d] < d, which positively resolves a conjecture of Kuzmin and Warmuth relating to their unlabeled Peeling compression scheme and also leads to an improved one-inclusion mistake bound. The proof uses a new form of VC-invariant shifting and a group-theoretic symmetrization. Our second main result is an algebraic topological property of maximum classes of VC-dimension d as being d contractible simplicial complexes, extending the well-known characterization that d = 1 maximum classes are trees. We negatively resolve a minimum degree conjecture of Kuzmin and Warmuth—the second part to a conjectured proof of correctness for Peeling—that every class has one-inclusion minimum degree at most its VCdimension. Our final main result is a k-class analogue of the d/n mistake bound, replacing the VC-dimension by the Pollard pseudo-dimension and the one-inclusion strategy by its natural hypergraph generalization. This result improves on known PAC-based expected risk bounds by a factor of O(logn) and is shown to be optimal up to an O(logk) factor. The combinatorial technique of shifting takes a central role in understanding the one-inclusion (hyper)graph and is a running theme throughout.