Prediciton of emergency events: a multi-task multi-label learning approach

Prediction of patient outcomes is critical to plan resources in an hospital emergency department. We present a method to exploit longitudinal data from Electronic Medical Records (EMR), whilst exploiting multiple patient outcomes. We divide the EMR data into segments where each segment is a task, and all tasks are associated with multiple patient outcomes over a 3, 6 and 12 month period. We propose a model that learns a prediction function for each task-label pair, interacting through two subspaces: the first subspace is used to impose sharing across all tasks for a given label. The second subspace captures the task-specific variations and is shared across all the labels for a given task. The proposed model is formulated as an iterative optimization problems and solved using a scalable and efficient Block co-ordinate descent (BCD) method. We apply the proposed model on two hospital cohorts - Cancer and Acute Myocardial Infarction (AMI) patients collected over a two year period from a large hospital emergency department. We show that the predictive performance of our proposed models is significantly better than those of several state-of-the-art multi-task and multi-label learning methods.

A new transfer learning framework with application to model-agnostic multi-task learning

Learning from small number of examples is a challenging problem in machine learning. An effective way to improve the performance is through exploiting knowledge from other related tasks. Multi-task learning (MTL) is one such useful paradigm that aims to improve the performance through jointly modeling multiple related tasks. Although there exist numerous classification or regression models in machine learning literature, most of the MTL models are built around ridge or logistic regression. There exist some limited works, which propose multi-task extension of techniques such as support vector machine, Gaussian processes. However, all these MTL models are tied to specific classification or regression algorithms and there is no single MTL algorithm that can be used at a meta level for any given learning algorithm. Addressing this problem, we propose a generic, model-agnostic joint modeling framework that can take any classification or regression algorithm of a practitioner’s choice (standard or custom-built) and build its MTL variant. The key observation that drives our framework is that due to small number of examples, the estimates of task parameters are usually poor, and we show that this leads to an under-estimation of task relatedness between any two tasks with high probability. We derive an algorithm that brings the tasks closer to their true relatedness by improving the estimates of task parameters. This is achieved by appropriate sharing of data across tasks. We provide the detail theoretical underpinning of the algorithm. Through our experiments with both synthetic and real datasets, we demonstrate that the multi-task variants of several classifiers/regressors (logistic regression, support vector machine, K-nearest neighbor, Random Forest, ridge regression, support vector regression) convincingly outperform their single-task counterparts. We also show that the proposed model performs comparable or better than many state-of-the-art MTL and transfer learning baselines.

Toxicity prediction in cancer using multiple instance learning in a multi-task framework

Treatments of cancer cause severe side effects called toxicities. Reduction of such effects is crucial in cancer care. To impact care, we need to predict toxicities at fortnightly intervals. This toxicity data differs from traditional time series data as toxicities can be caused by one treatment on a given day alone, and thus it is necessary to consider the effect of the singular data vector causing toxicity. We model the data before prediction points using the multiple instance learning, where each bag is composed of multiple instances associated with daily treatments and patient-specific attributes, such as chemotherapy, radiotherapy, age and cancer types. We then formulate a Bayesian multi-task framework to enhance toxicity prediction at each prediction point. The use of the prior allows factors to be shared across task predictors. Our proposed method simultaneously captures the heterogeneity of daily treatments and performs toxicity prediction at different prediction points. Our method was evaluated on a real-word dataset of more than 2000 cancer patients and had achieved a better prediction accuracy in terms of AUC than the state-of-art baselines.

Teachers' aides training and professional learning in mainstream secondary schools

 Dianne Gibson investigated the training and professional learning needs of teachers’ aides supporting students with disabilities in mainstream secondary schools in Victoria, Australia. Using survey and focus group methods, analysis showed that teachers’ aides perceived they had participated in training and professional learning to effectively perform work-related items and student-related tasks.

Personality types and languages learning strategies: chameleons changing colours

This exploration of associations between the reported Language Learning Strategy (LLS) preferences of learners of English as a Second Language (ESL) and their personality types is positioned within the contention that the two are generally related. Our findings unequivocally support the existence of this relationship. Moreover, they also provide a platform from which to understand the contribution to learning a second language of two very commonly cited personality traits, introversion/extroversion and neuroticism. However, they also provide the basis for the important caution that the association between personality types and LLS is quite volatile. We have found that it is variation rather than unwavering stability that features in how personality traits apply as predictive of ESL learners' specific LLS preferences. Such prediction is specified even further by the particular contexts of ESL learning where the LLS are applied, for example for listening or speaking and whether this occurs inside or outside a classroom. The implications of these findings for ESL teaching and learning are discussed as is the explanatory power of the chameleon metaphor.

Developement and evidencing achievement of graduate learning outcomes in Deakin University's enhanced mooc

There is now a plethora of Massive Open On-line courses (MOOCs) offered worldwide. Whilst many MOOCs focus on discipline-specific content, little attention has been paid to how MOOCs can explicitly help participants develop generic employability skills such as communication, digital literacy, global citizenship and the like. Similarly little attention been paid to explicitly assuring the quality of MOOCs with respect to alignment with regulatory body standards. Deakin University's first MOOC, DeakinPrimer, is an introduction to humanitarian responses to 21st century disasters. It has been designed to assist participants to explicitly evidence generic or employability skills, some of Deakin's eight Graduate Learning Outcomes (GLOs) including communication, digital literacy, critical thinking and global citizenship. Other key features of DeakinPrimer include opportunities for networking with fellow participants and experts within the humanitarian field, and the opportunity to apply for credit towards the Graduate Certificate in International Community Development (level 8 in the Australian Qualifications Framework [AQF]) and for those with a prior Bachelor degree, the Masters in Humanitarian Assistance or the Masters of International Community Development (level 9 in the AQF). DeakinPrimer is designed as a test bed for a learning innovation, particularly micro-credentialing GLOs using digital badges to enable self and peer endorsement of evidence of learning. Badging is integrated in two ways. Firstly, DeakinPrimer participants build portfolios of learning artefacts associated with learning activities, then assess their work against a set of holistic, generic learning outcomes standards rubrics. If they judge their evidence as meeting the required standard, they can claim a badge (self endorsement) associated with particular GLOs. Secondly, participants can request and provide peer feedback and endorsement (using peer badges). The integration of self and peer review in the assessment tasks helps participants develop important employability skills, the ability to critically self-reflect on their own work and critically analyse the work of others and provide evidence-based feedback. DeakinPrimer is scheduled to commence in July 2013. This paper explains the way in which the course curricula has been designed to use technologies to enable participants to curate evidence of learning, and self and peer endorse such learning against defined standards.

The pedagogy of using video to develop reflective practice in learning to teach science

Purpose The purpose of this chapter is to identify the pedagogical approaches that foster critical reflection using video among the preservice teachers during tutorials. Methodology/approach The research is situated in a school-based teaching programme in which pairs of pre-service teachers taught small groups of primary aged children over a period of seven weeks. Volunteer pre-service teachers videotaped their lessons and selected video excerpts to share with their peers in the tutorial. The educator guided the preservice teachers’ reflection using the video. A case study drawing on interviews with pre-service teachers and audio recordings of tutorials, charted the development of pedagogical decisions made by the educators to promote reflection.Findings The pre-service teachers had difficulties undertaking deep reflection of their own and peers’ teaching practice. The response by educators was to promote collaboration among pre-service teachers by discussing specific aspects of the teaching in small groups and to use a jigsaw approach. This enabled a deeper analysis of particular elements of the lesson that were then integrated to produce a more holistic understanding of the teaching. The video data is most suitable for reflection and provides valuable evidence for pre-service teachers to develop their practice. Practical implications For pre-service teachers to develop effective skills to analyse their own practice they need to experience teaching in a safe but challenging environment, over a sustained period; have opportunities to develop a shared understanding of what constitutes quality teaching; have opportunities to critically analyse their teaching in discussion with peers and educators and be able to be guided by a framework of reflective strategies.

Deep ecology and Australian suburbia: learning from Aboriginal Australia philosophy

Naess’ Deep Ecology [50] represents a fundamental philosophical and conceptual shift from the dominant Western thinking that can be traced back to the Greek and Roman Empires. Like all philosophy, Naess’ Deep Ecology was born of and is most relevant to a specific time and place being northern Europe. Although the fundamentals of the Deep Ecology philosophy were new to modern Western thinking, it is not new to traditional Indigenous cultures, including the world’s oldest culture, that of Aboriginal Australia. While the past four decades has seen an increasing recognition of Aboriginal philosophical approaches, there is very little understanding of what this philosophical approach is and means for the management of the Australian environment in which humans are a central part. Since European arrival, Australia has been one of the world’s most urban societies. Unlike northern Europe, urban Australia is low density and suburban, a legacy of British and North American influences. Nearly 90% of Australians live in detached houses surrounded by gardens. Managed by individual residents, this land use accounts for about 70% of the total area of cities like Melbourne. Deeply culturally embedded, the Australian desire for living in low-density suburbs is unlikely to change soon. Contemporary cities are widely recognized as causing severe environmental degradation and are not sustainable. Yet in Australia introduced philosophical and design approaches are still used to address the unsustainable impacts of urban forms introduced from another time and place. While impractical to remove the existing suburban form in Australian cities, there is a significant opportunity to retrofit them using Australian Aboriginal philosophical and land management understandings developed and tested over tens of thousands of years. This paper establishes a contemporary Australian Deep Ecology philosophical approach to sustainably living in the suburbs that recognizes and works with the legacies of Australian Aboriginal, English, North American and contemporary Australian influences.

A machine learning approach to measure and monitor physical activity in children to help fight overweight and obesity

Physical Activity is important for maintaining healthy lifestyles. Recommendations for physical activity levels are issued by most governments as part of public health measures. As such, reliable measurement of physical activity for regulatory purposes is vital. This has lead research to explore standards for achieving this using wearable technology and artificial neural networks that produce classifications for specific physical activity events. Applied from a very early age, the ubiquitous capture of physical activity data using mobile and wearable technology may help us to understand how we can combat childhood obesity and the impact that this has in later life. A supervised machine learning approach is adopted in this paper that utilizes data obtained from accelerometer sensors worn by children in free-living environments. The paper presents a set of activities and features suitable for measuring physical activity and evaluates the use of a Multilayer Perceptron neural network to classify physical activities by activity type. A rigorous reproducible data science methodology is presented for subsequent use in physical activity research. Our results show that it was possible to obtain an overall accuracy of 96 % with 95 % for sensitivity, 99 % for specificity and a kappa value of 94 % when three and four feature combinations were used.

The relation between receptive grammar and procedural, declarative, and working memory in specific language impairment

What memory systems underlie grammar in children, and do these differ between typically developing (TD) children and children with specific language impairment (SLI)? Whilst there is substantial evidence linking certain memory deficits to the language problems in children with SLI, few studies have investigated multiple memory systems simultaneously, examining not only possible memory deficits but also memory abilities that may play a compensatory role. This study examined the extent to which procedural, declarative, and working memory abilities predict receptive grammar in 45 primary school aged children with SLI (30 males, 15 females) and 46 TD children (30 males, 16 females), both on average 9;10 years of age. Regression analyses probed measures of all three memory systems simultaneously as potential predictors of receptive grammar. The model was significant, explaining 51.6% of the variance. There was a significant main effect of learning in procedural memory and a significant group × procedural learning interaction. Further investigation of the interaction revealed that procedural learning predicted grammar in TD but not in children with SLI. Indeed, procedural learning was the only predictor of grammar in TD. In contrast, only learning in declarative memory significantly predicted grammar in SLI. Thus, different memory systems are associated with receptive grammar abilities in children with SLI and their TD peers. This study is, to our knowledge, the first to demonstrate a significant group by memory system interaction in predicting grammar in children with SLI and their TD peers. In line with Ullman's Declarative/Procedural model of language and procedural deficit hypothesis of SLI, variability in understanding sentences of varying grammatical complexity appears to be associated with variability in procedural memory abilities in TD children, but with declarative memory, as an apparent compensatory mechanism, in children with SLI.

Cocaine-specific neuroplasticity in the ventral striatum network is linked to delay discounting and drug relapse

AIMS: To contrast functional connectivity on ventral and dorsal striatum networks in cocaine dependence relative to pathological gambling, via a resting-state functional connectivity approach; and to determine the association between cocaine dependence-related neuroadaptations indexed by functional connectivity and impulsivity, compulsivity and drug relapse. DESIGN: Cross-sectional study of 20 individuals with cocaine dependence (CD), 19 individuals with pathological gambling (PG) and 21 healthy controls (HC), and a prospective cohort study of 20 CD followed-up for 12 weeks to measure drug relapse. SETTING AND PARTICIPANTS: CD and PG were recruited through consecutive admissions to a public clinic specialized in substance addiction treatment (Centro Provincial de Drogodependencias) and a public clinic specialized in gambling treatment (AGRAJER), respectively; HC were recruited through community advertisement in the same area in Granada (Spain). MEASUREMENTS: Seed-based functional connectivity in the ventral striatum (ventral caudate and ventral putamen) and dorsal striatum (dorsal caudate and dorsal putamen), the Kirby delay-discounting questionnaire, the reversal-learning task and a dichotomous measure of cocaine relapse indicated with self-report and urine tests. FINDINGS: CD relative to PG exhibit enhanced connectivity between the ventral caudate seed and subgenual anterior cingulate cortex, the ventral putamen seed and dorsomedial pre-frontal cortex and the dorsal putamen seed and insula (P≤0.001, kE=108). Connectivity between the ventral caudate seed and subgenual anterior cingulate cortex is associated with steeper delay discounting (P≤0.001, kE=108) and cocaine relapse (P≤0.005, kE=34). CONCLUSIONS: Cocaine dependence-related neuroadaptations in the ventral striatum of the brain network are associated with increased impulsivity and higher rate of cocaine relapse.

Responding to the widening participation agenda through improved access to and within 3D virtual learning environments

The Australian Government's widening participation agenda - also referred to as the social inclusion agenda - considers equity through the triple focus of access, participation and outcomes. These foci are catalysts for re-examining teaching and learning approaches in formal education. This article considers this national refocus and the possibilities for addressing access and equity issues through and within threedimensional virtual learning environments (3DVLEs). The findings of an Australian Learning and Teaching Council (ALTC)-funded project that investigated the potential of an accessible 3DVLE for increasing access and participation of students with disabilities are reported, and strategies for improving outcomes (i.e. retention, success and completion) proposed. The article also highlights some of the remaining challenges with regard to the goal of improving outcomes for under-represented learner groups. The final section of the article identifies areas for further research.

Mealtime behaviour among children with disabilities: family interactions and stress

 The impact of feeding problems on parental stress, quality of mealtime interactions and opportunity for socialisation and naturalistic learning during mealtimes was examined in children with developmental disabilities and typically developing children through survey and qualitative analysis of video recordings of family mealtimes.