Factors influencing robustness and effectiveness of conditional random fields in active learning frameworks

Active learning approaches reduce the annotation cost required by traditional supervised approaches to reach the same effectiveness by actively selecting informative instances during the learning phase. However, effectiveness and robustness of the learnt models are influenced by a number of factors. In this paper we investigate the factors that affect the effectiveness, more specifically in terms of stability and robustness, of active learning models built using conditional random fields (CRFs) for information extraction applications. Stability, defined as a small variation of performance when small variation of the training data or a small variation of the parameters occur, is a major issue for machine learning models, but even more so in the active learning framework which aims to minimise the amount of training data required. The factors we investigate are a) the choice of incremental vs. standard active learning, b) the feature set used as a representation of the text (i.e., morphological features, syntactic features, or semantic features) and c) Gaussian prior variance as one of the important CRFs parameters. Our empirical findings show that incremental learning and the Gaussian prior variance lead to more stable and robust models across iterations. Our study also demonstrates that orthographical, morphological and contextual features as a group of basic features play an important role in learning effective models across all iterations.

Moving forwards with the aim of going backwards fast : high performance rowing as a learning environment

Background and Purpose: - This paper focuses on the learning culture within the high performance levels of rowing. In doing so, we explore the case of an individual’s learning as he moves across athletic, coaching and administrative functions. This exploration draws on a cultural learning framework and complementary theorisings related to reflexivity. Method - This study makes use of an intellectually, morally and collaboratively challenging approach whereby one member of the research team was also the sole participant of this study. The participant’s careers as a high performance athlete, coach and administrator, coupled with his experience in conducting empirical research presented a rare opportunity to engage in collaborative research (involving degrees of insider and outsider status for each of the research team). We acknowledge that others have looked to combine roles of coach / athlete / administrator with that of researcher however few (if any) have attempted to combine them all in one project. Moreover, coupled with the approach to reflexivity adopted in this study and the authorship contributions we consider this scholarly direction uncommon. Data were comprised of recorded research conversations, a subsequently constructed learning narrative, reflections on the narrative, a stimulated reflective piece from the participant, and a final (re)construction of the participant’s story. Accordingly, data were integrated through an iterative process of thematic analysis. Results - The cultural (i.e., the ways things get done) and structural (e.g., the rules and regulations) properties of high performance rowing were found to shape both the opportunities to be present (e.g., secure a place in the crew) and to learn (e.g., learn the skills required to perform at an Olympic level). However, the individual’s personal properties were brought to bear on re-shaping the constraints such that many limitations could be overcome. In keeping with the theory of learning cultures, the culture of rowing was found to position individuals (a coxswain in this case) differentially. In a similar manner, a range of structural features was found to be important in shaping the cultural and personal elements in performance contexts. For example, the ‘field of play’ was found to be important as a structural feature (i.e., inability of coach to communicate with athletes) in shaping the cultural and personal elements of learning in competition (e.g., positioning the coxswain as an in-boat coach and trusted crewmate). Finally, the cultural and structural elements in rowing appeared to be activated by the participant’s personal elements, most notably his orientation towards quality performance. Conclusion - The participant in this study was found to be driven by the project that he cares about most and at each turn he has bent his understanding of his sport back on itself to see if he can find opportunities to learn and subsequently explore ways to improve performance. The story here emphasises the importance of learner agency, and this is an aspect that has often been missing in recent theorising about learning. In this study, we find an agent using his ‘personal emergent powers to activate the resources in the culture and structure of his sport in an attempt to improve performance. We conclude from this account that this particular high performance rowing culture is one that provided support but nonetheless encouraged those involved, to ‘figure things out’ for themselves – be it as athletes, coaches and/or administrators.

Automatic UAV forced landing site detection using machine learning

The commercialization of aerial image processing is highly dependent on the platforms such as UAVs (Unmanned Aerial Vehicles). However, the lack of an automated UAV forced landing site detection system has been identified as one of the main impediments to allow UAV flight over populated areas in civilian airspace. This article proposes a UAV forced landing site detection system that is based on machine learning approaches including the Gaussian Mixture Model and the Support Vector Machine. A range of learning parameters are analysed including the number of Guassian mixtures, support vector kernels including linear, radial basis function Kernel (RBF) and polynormial kernel (poly), and the order of RBF kernel and polynormial kernel. Moreover, a modified footprint operator is employed during feature extraction to better describe the geometric characteristics of the local area surrounding a pixel. The performance of the presented system is compared to a baseline UAV forced landing site detection system which uses edge features and an Artificial Neural Network (ANN) region type classifier. Experiments conducted on aerial image datasets captured over typical urban environments reveal improved landing site detection can be achieved with an SVM classifier with an RBF kernel using a combination of colour and texture features. Compared to the baseline system, the proposed system provides significant improvement in term of the chance to detect a safe landing area, and the performance is more stable than the baseline in the presence of changes to the UAV altitude.

Machine learning approaches to analysing textual injury surveillance data: A systematic review

Objective To synthesise recent research on the use of machine learning approaches to mining textual injury surveillance data. Design Systematic review. Data sources The electronic databases which were searched included PubMed, Cinahl, Medline, Google Scholar, and Proquest. The bibliography of all relevant articles was examined and associated articles were identified using a snowballing technique. Selection criteria For inclusion, articles were required to meet the following criteria: (a) used a health-related database, (b) focused on injury-related cases, AND used machine learning approaches to analyse textual data. Methods The papers identified through the search were screened resulting in 16 papers selected for review. Articles were reviewed to describe the databases and methodology used, the strength and limitations of different techniques, and quality assurance approaches used. Due to heterogeneity between studies meta-analysis was not performed. Results Occupational injuries were the focus of half of the machine learning studies and the most common methods described were Bayesian probability or Bayesian network based methods to either predict injury categories or extract common injury scenarios. Models were evaluated through either comparison with gold standard data or content expert evaluation or statistical measures of quality. Machine learning was found to provide high precision and accuracy when predicting a small number of categories, was valuable for visualisation of injury patterns and prediction of future outcomes. However, difficulties related to generalizability, source data quality, complexity of models and integration of content and technical knowledge were discussed. Conclusions The use of narrative text for injury surveillance has grown in popularity, complexity and quality over recent years. With advances in data mining techniques, increased capacity for analysis of large databases, and involvement of computer scientists in the injury prevention field, along with more comprehensive use and description of quality assurance methods in text mining approaches, it is likely that we will see a continued growth and advancement in knowledge of text mining in the injury field.

Do-Rank: DCG optimization for learning-to-rank in tag-based item recommendation systems

Discounted Cumulative Gain (DCG) is a well-known ranking evaluation measure for models built with multiple relevance graded data. By handling tagging data used in recommendation systems as an ordinal relevance set of {negative,null,positive}, we propose to build a DCG based recommendation model. We present an efficient and novel learning-to-rank method by optimizing DCG for a recommendation model using the tagging data interpretation scheme. Evaluating the proposed method on real-world datasets, we demonstrate that the method is scalable and outperforms the benchmarking methods by generating a quality top-N item recommendation list.

Exploring with Grammar in the Primary Years: Learning about Language in the Australian Curriculum: English

This new volume, Exploring with Grammar in the Primary Years (Exley, Kevin & Mantei, 2014), follows on from Playing with Grammar in the Early Years (Exley & Kervin, 2013). We extend our thanks to the ALEA membership for their take up of the first volume and the vibrant conversations around our first attempt at developing a pedagogy for the teaching of grammar in the early years. Your engagement at locally held ALEA events has motivated us to complete this second volume and reassert our interest in the pursuit of socially-just outcomes in the primary years. As noted in Exley and Kervin (2013), we believe that mastering a range of literacy competences includes not only the technical skills for learning, but also the resources for viewing and constructing the world (Freire and Macdeo, 1987). Rather than seeing knowledge about language as the accumulation of technical skills alone, the viewpoint to which we subscribe treats knowledge about language as a dialectic that evolves from, is situated in, and contributes to active participation within a social arena (Halliday, 1978). We acknowledge that to explore is to engage in processes of discovery as we look closely and examine the opportunities before us. As such, we draw on Janks’ (2000; 2014) critical literacy theory to underpin many of the learning experiences in this text. Janks (2000) argues that effective participation in society requires knowledge about how the power of language promotes views, beliefs and values of certain groups to the exclusion of others. Powerful language users can identify not only how readers are positioned by these views, but also the ways these views are conveyed through the design of the text, that is, the combination of vocabulary, syntax, image, movement and sound. Similarly, powerful designers of texts can make careful modal choices in written and visual design to promote certain perspectives that position readers and viewers in new ways to consider more diverse points of view. As the title of our text suggests, our activities are designed to support learners in exploring the design of texts to achieve certain purposes and to consider the potential for the sharing of their own views through text production. In Exploring with Grammar in the Primary Years, we focus on the Year 3 to Year 6 grouping in line with the Australian Curriculum, Assessment and Reporting Authority’s (hereafter ACARA) advice on the ‘nature of learners’ (ACARA, 2014). Our goal in this publication is to provide a range of highly practical strategies for scaffolding students’ learning through some of the Content Descriptions from the Australian Curriculum: English Version 7.2, hereafter AC:E (ACARA, 2014). We continue to express our belief in the power of using whole texts from a range of authentic sources including high quality children’s literature, the internet, and examples of community-based texts to expose students to the richness of language. Taking time to look at language patterns within actual texts is a pathway to ‘…capture interest, stir the imagination and absorb the [child]’ into the world of language and literacy (Saxby, 1993, p. 55). It is our intention to be more overt this time and send a stronger message that our learning experiences are simply ‘sample’ activities rather than a teachers’ workbook or a program of study to be followed. We’re hoping that teachers and students will continue to explore their bookshelves, the internet and their community for texts that provide powerful opportunities to engage with language-based learning experiences. In the following three sections, we have tried to remain faithful to our interpretation of the AC:E Content Descriptions without giving an exhaustive explanation of the grammatical terms. This recently released curriculum offers a new theoretical approach to building students’ knowledge about language. The AC:E uses selected traditional terms through an approach developed in systemic functional linguistics (see Halliday and Matthiessen, 2004) to highlight the dynamic forms and functions of multimodal language in texts. For example, the following statement, taken from the ‘Language: Knowing about the English language’ strand states: English uses standard grammatical terminology within a contextual framework, in which language choices are seen to vary according to the topics at hand, the nature and proximity of the relationships between the language users, and the modalities or channels of communication available (ACARA, 2014). Put simply, traditional grammar terms are used within a functional framework made up of field, tenor, and mode. An understanding of genre is noted with the reference to a ‘contextual framework’. The ‘topics at hand’ concern the field or subject matter of the text. The ‘relationships between the language users’ is a description of tenor. There is reference to ‘modalities’, such as spoken, written or visual text. We posit that this innovative approach is necessary for working with contemporary multimodal and cross-cultural texts (see Exley & Mills, 2012). Other excellent tomes, such as Derewianka (2011), Humphrey, Droga and Feez (2012), and Rossbridge and Rushton (2011) provide more comprehensive explanations of this unique metalanguage, as does the AC:E Glossary. We’ve reproduced some of the AC:E Glossary at the end of this publication. We’ve also kept the same layout for our learning experiences, ensuring that our teacher notes are not only succinct but also prudent in their placement. Each learning experience is connected to a Content Description from the AC:E and contains an experience with an identified purpose, suggested resource text and a possible sequence for the experience that always commences with an orientation to text followed by an examination of a particular grammatical resource. Our plans allow for focused discussion, shared exploration and opportunities to revisit the same text for the purpose of enhancing meaning making. Some learning experiences finish with deconstruction of a stimulus text while others invite students to engage in the design of new texts. We encourage you to look for opportunities in your own classrooms to move from text deconstruction to text design. In this way, students can express not only their emerging grammatical understandings, but also the ways they might position readers or viewers through the creation of their own texts. We expect that each of these learning experiences will vary in the time taken. Some may indeed take a couple if not a few teaching episodes to work through, especially if students are meeting a concept or a pedagogical strategy for the first time. We hope you use as much, or as little, of each experience as is needed for your students. We do not want the teaching of grammar to slip into a crisis of irrelevance or to be seen as a series of worksheet drills with finite answers. We firmly believe that strategies for effective deconstruction and design practice, however, have much portability. We three are very keen to hear from teachers who are adopting and adapting these learning experiences in their classrooms. Please email us on b.exley@qut.edu.au, lkervin@uow.edu.au or jessicam@ouw.edu.au. We’d love to continue the conversation with you over time. Beryl Exley, Lisa Kervin & Jessica Mantei

Not learning from the past: Adaptive governance challenges for Australian natural resource management

Australia’s governance arrangements for NRM have evolved considerably over the last thirty years. The impact of changes in governance on NRM planning and delivery requires assessment. We undertake a multi-method program evaluation using adaptive governance principles as an analytical frame and apply this to Queensland to assess the impacts of governance change on NRM planning and governance outcomes. Data to inform our analysis includes: 1) a systematic review of sixteen audits/evaluations of Australian NRM over a fifteen-year period; 2) a review of Queensland’s first generation NRM Plans; and 3) outputs from a Queensland workshop on NRM planning. NRM has progressed from a bottom-up grassroots movement into a collaborative regional NRM model that has been centralised by the Australian Government. We found that while some adaptive governance challenges have been addressed, others remained unresolved. Results show that collaboration and elements of multi-level governance under the regional model were positive moves, but also that NRM arrangements contained structural deficiencies across multiple governance levels in relation to public involvement in decision-making and knowledge production for problem responsiveness. These problems for adaptive governance have been exacerbated since 2008. We conclude that the adaptive governance framework for NRM needs urgent attention so that important environmental management problems can be addressed.

Early maternal deprivation reduces prepulse inhibition and impairs spatial learning ability in adulthood: No further effect of post-pubertal chronic corticosterone treatment

Prolonged maternal deprivation leads to long-term alterations in hypothalamic–pituitary–adrenal (HPA) axis activity, disturbances of auditory information processing and neurochemical changes in the adult brain, some of which are similar to that observed in schizophrenia. Here we report the adult behavioural effects of maternal deprivation (12 h on postnatal days 9 and 11) in Wistar rats on paradigms of auditory information processing (prepulse inhibition), sensitivity to dopamimetics (amphetamine-induced hyper-locomotion) and cognition (T-maze delayed alternation and Morris water-maze). In addition, we examined the long-lasting effect of chronic 21-day corticosterone treatment during the post-pubertal period (i.e., postnatal days 56–76) on each of these behavioural paradigms in maternally deprived and control rats. Behavioural testing commenced 2 weeks after the termination of corticosterone treatment. Maternal deprivation led to a significant reduction in PPI and impaired spatial learning ability in adulthood, but did not affect the behavioural response to amphetamine. Post-pubertal chronic corticosterone treatment did not have any major long-lasting effects on any of the behavioural measures in either maternally deprived or control rats. Our findings further support maternal deprivation as an animal model of specific aspects of schizophrenia.

Improving deep convultional neural networks with unsupervised feature learning

The latest generation of Deep Convolutional Neural Networks (DCNN) have dramatically advanced challenging computer vision tasks, especially in object detection and object classification, achieving state-of-the-art performance in several computer vision tasks including text recognition, sign recognition, face recognition and scene understanding. The depth of these supervised networks has enabled learning deeper and hierarchical representation of features. In parallel, unsupervised deep learning such as Convolutional Deep Belief Network (CDBN) has also achieved state-of-the-art in many computer vision tasks. However, there is very limited research on jointly exploiting the strength of these two approaches. In this paper, we investigate the learning capability of both methods. We compare the output of individual layers and show that many learnt filters and outputs of the corresponding level layer are almost similar for both approaches. Stacking the DCNN on top of unsupervised layers or replacing layers in the DCNN with the corresponding learnt layers in the CDBN can improve the recognition/classification accuracy and training computational expense. We demonstrate the validity of the proposal on ImageNet dataset.

Teacher librarians’ understandings of inquiry learning

This paper presents a study investigating teacher librarians’ understandings of inquiry learning. Teacher librarians have traditionally been involved in information literacy education. For some teacher librarians, this has involved collaborating with the classroom teacher on inquiry learning units of work. For others, it has involved offering a parallel library curriculum. The findings of this study are based on semi-structured interviews with nine teacher librarians in Queensland schools. The study revealed that teacher librarians saw inquiry learning in two ways as (a) student-centred investigation and (b) teaching a process.

A Fuzzy Approximation Scheme for Sequential Learning in Pattern Recognition

An adaptive learning scheme, based on a fuzzy approximation to the gradient descent method for training a pattern classifier using unlabeled samples, is described. The objective function defined for the fuzzy ISODATA clustering procedure is used as the loss function for computing the gradient. Learning is based on simultaneous fuzzy decisionmaking and estimation. It uses conditional fuzzy measures on unlabeled samples. An exponential membership function is assumed for each class, and the parameters constituting these membership functions are estimated, using the gradient, in a recursive fashion. The induced possibility of occurrence of each class is useful for estimation and is computed using 1) the membership of the new sample in that class and 2) the previously computed average possibility of occurrence of the same class. An inductive entropy measure is defined in terms of induced possibility distribution to measure the extent of learning. The method is illustrated with relevant examples.

Learning Temporal Alignment Uncertainty for Efficient Event Detection

In this paper we tackle the problem of efficient video event detection. We argue that linear detection functions should be preferred in this regard due to their scalability and efficiency during estimation and evaluation. A popular approach in this regard is to represent a sequence using a bag of words (BOW) representation due to its: (i) fixed dimensionality irrespective of the sequence length, and (ii) its ability to compactly model the statistics in the sequence. A drawback to the BOW representation, however, is the intrinsic destruction of the temporal ordering information. In this paper we propose a new representation that leverages the uncertainty in relative temporal alignments between pairs of sequences while not destroying temporal ordering. Our representation, like BOW, is of a fixed dimensionality making it easily integrated with a linear detection function. Extensive experiments on CK+, 6DMG, and UvA-NEMO databases show significant performance improvements across both isolated and continuous event detection tasks.

Action learning in partnership with Landcare and catchment management groups to support increased pasture sowings in southern inland Queensland

The incorporation of sown pastures as short-term rotations into the cropping systems of northern Australia has been slow. The inherent chemical fertility and physical stability of the predominant vertisol soils across the region enabled farmers to grow crops for decades without nitrogen fertiliser, and precluded the evolution of a crop–pasture rotation culture. However, as less fertile and less physically stable soils were cropped for extended periods, farmers began to use contemporary farming and sown pasture technologies to rebuild and maintain their soils. This has typically involved sowing long-term grass and grass–legume pastures on the more marginal cropping soils of the region. In partnership with the catchment management authority, the Queensland Murray–Darling Committee (QMDC) and Landcare, a pasture extension process using the LeyGrain™ package was implemented in 2006 within two Grain & Graze projects in the Maranoa-Balonne and Border Rivers catchments in southern inland Queensland. The specific objectives were to increase the area sown to high quality pasture and to gain production and environmental benefits (particularly groundcover) through improving the skills of producers in pasture species selection, their understanding and management of risk during pasture establishment, and in managing pastures and the feed base better. The catalyst for increasing pasture sowings was a QMDC subsidy scheme for increasing groundcover on old cropping land. In recognising a need to enhance pasture knowledge and skills to implement this scheme, the QMDC and Landcare producer groups sought the involvement of, and set specific targets for, the LeyGrain workshop process. This is a highly interactive action learning process that built on the existing knowledge and skills of the producers. Thirty-four workshops were held with more than 200 producers in 26 existing groups and with private agronomists. An evaluation process assessed the impact of the workshops on the learning and skill development by participants, their commitment to practice change, and their future intent to sow pastures. The results across both project catchments were highly correlated. There was strong agreement by producers (>90%) that the workshops had improved knowledge and skills regarding the adaptation of pasture species to soils and climates, enabling a better selection at the paddock level. Additional strong impacts were in changing the attitudes of producers to all aspects of pasture establishment, and the relative species composition of mixtures. Producers made a strong commitment to practice change, particularly in managing pasture as a specialist crop at establishment to minimise risk, and in the better selection and management of improved pasture species (particularly legumes and the use of fertiliser). Producers have made a commitment to increase pasture sowings by 80% in the next 5 years, with fourteen producers in one group alone having committed to sow an additional 4893 ha of pasture in 2007–08 under the QMDC subsidy scheme. The success of the project was attributed to the partnership between QMDC and Landcare groups who set individual workshop targets with LeyGrain presenters, the interactive engagement processes within the workshops themselves, and the follow-up provided by the LeyGrain team for on-farm activities.

A cooperative game of a pair of learning automata

A cooperative game played in a sequential manner by a pair of learning automata is investigated in this paper. The automata operate in an unknown random environment which gives a common pay-off to the automata. Necessary and sufficient conditions on the functions in the reinforcement scheme are given for absolute monotonicity which enables the expected pay-off to be monotonically increasing in any arbitrary environment. As each participating automaton operates with no information regarding the other partner, the results of the paper are relevant to decentralized control.