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.

Barriers to information transfer for multi-trauma patients upon discharge from the Emergency Department in a tertiary level hospital

Purpose: Communication is integral to effective trauma care provision. This presentation will report on barriers to meaningful information transfer for multi-trauma patients upon discharge from the Emergency Department (ED) to the care areas of Intensive Care Unit, High Dependency Unit, and Perioperative Services. This is an ongoing study at one tertiary level hospital in Queensland. Method: This is a multi-phase, mixed method study. In Phase 1 data were collected about information transfer. This Phase was initially informed by a comprehensive literature review, then via focus groups, chart audit, staff survey and review of national and international trauma forms. Results: The barriers identified related to nursing handover, documented information, time inefficiency, patient complexity and stability and time of transfer. Specifically this included differences in staff expectations and variation in the nursing handover processes, no agreed minimum dataset of information handed over, missing, illegible or difficult to find information in documentation (both medical and nursing), low compliance with some forms used for documentation. Handover of these patients is complex with information coming from many sources, dealing with issues is more difficult for these patients when transferred out of hours. Conclusions and further directions: This study investigated the current communication processes and standards of information transfer to identify barriers and issues. The barriers identified were the structure used for documentation, processes used (e.g. handover), patient acuity and time. This information is informing the development, implementation and evaluation of strategies to ameliorate the issues identified.

A risk simulation framework for information infrastructure protection

Information communication and technology (ICT) systems are almost ubiquitous in the modern world. It is hard to identify any industry, or for that matter any part of society, that is not in some way dependent on these systems and their continued secure operation. Therefore the security of information infrastructures, both on an organisational and societal level, is of critical importance. Information security risk assessment is an essential part of ensuring that these systems are appropriately protected and positioned to deal with a rapidly changing threat environment. The complexity of these systems and their inter-dependencies however, introduces a similar complexity to the information security risk assessment task. This complexity suggests that information security risk assessment cannot, optimally, be undertaken manually. Information security risk assessment for individual components of the information infrastructure can be aided by the use of a software tool, a type of simulation, which concentrates on modelling failure rather than normal operational simulation. Avoiding the modelling of the operational system will once again reduce the level of complexity of the assessment task. The use of such a tool provides the opportunity to reuse information in many different ways by developing a repository of relevant information to aid in both risk assessment and management and governance and compliance activities. Widespread use of such a tool allows the opportunity for the risk models developed for individual information infrastructure components to be connected in order to develop a model of information security exposures across the entire information infrastructure. In this thesis conceptual and practical aspects of risk and its underlying epistemology are analysed to produce a model suitable for application to information security risk assessment. Based on this work prototype software has been developed to explore these concepts for information security risk assessment. Initial work has been carried out to investigate the use of this software for information security compliance and governance activities. Finally, an initial concept for extending the use of this approach across an information infrastructure is presented.

Engaging complexity through collaborative brand design

In this research we used inductive reasoning through design to understand how stakeholders in the Waterfall Way (New South Wales, Australia) perceive the relationships between themselves and the place they live in. This paper describes a collaborative design methodology used to release information about local identities, which guided the regional brand exercise. The methodology is explicit about the uncertainties and complexities of the design process and of its reception system. As such, it aims to engage with local stakeholders and experts in order to help elicit tacit knowledge and identify system patterns and trends that would possibly not be visible if a top-down expert-based process was used. Through collective design, local people were drawn together in search for a symbol to represent the meaning attached to their places/region in relation to sustainable tourism activity.

How difficult are exams? A framework for assessing the complexity of introductory programming exams

Student performance on examinations is influenced by the level of difficulty of the questions. It seems reasonable to propose therefore that assessment of the difficulty of exam questions could be used to gauge the level of skills and knowledge expected at the end of a course. This paper reports the results of a study investigating the difficulty of exam questions using a subjective assessment of difficulty and a purpose-built exam question complexity classification scheme. The scheme, devised for exams in introductory programming courses, assesses the complexity of each question using six measures: external domain references, explicitness, linguistic complexity, conceptual complexity, length of code involved in the question and/or answer, and intellectual complexity (Bloom level). We apply the scheme to 20 introductory programming exam papers from five countries, and find substantial variation across the exams for all measures. Most exams include a mix of questions of low, medium, and high difficulty, although seven of the 20 have no questions of high difficulty. All of the complexity measures correlate with assessment of difficulty, indicating that the difficulty of an exam question relates to each of these more specific measures. We discuss the implications of these findings for the development of measures to assess learning standards in programming courses.

Slice, mine and dice : complexity-aware automated discovery of business process models

Automated process discovery techniques aim at extracting models from information system logs in order to shed light into the business processes supported by these systems. Existing techniques in this space are effective when applied to relatively small or regular logs, but otherwise generate large and spaghetti-like models. In previous work, trace clustering has been applied in an attempt to reduce the size and complexity of automatically discovered process models. The idea is to split the log into clusters and to discover one model per cluster. The result is a collection of process models -- each one representing a variant of the business process -- as opposed to an all-encompassing model. Still, models produced in this way may exhibit unacceptably high complexity. In this setting, this paper presents a two-way divide-and-conquer process discovery technique, wherein the discovered process models are split on the one hand by variants and on the other hand hierarchically by means of subprocess extraction. The proposed technique allows users to set a desired bound for the complexity of the produced models. Experiments on real-life logs show that the technique produces collections of models that are up to 64% smaller than those extracted under the same complexity bounds by applying existing trace clustering techniques.

Complexity of increasing the secure connectivity in wireless ad hoc networks

We consider the problem of maximizing the secure connectivity in wireless ad hoc networks, and analyze complexity of the post-deployment key establishment process constrained by physical layer properties such as connectivity, energy consumption and interference. Two approaches, based on graph augmentation problems with nonlinear edge costs, are formulated. The first one is based on establishing a secret key using only the links that are already secured by shared keys. This problem is in NP-hard and does not accept polynomial time approximation scheme PTAS since minimum cutsets to be augmented do not admit constant costs. The second one extends the first problem by increasing the power level between a pair of nodes that has a secret key to enable them physically connect. This problem can be formulated as the optimal key establishment problem with interference constraints with bi-objectives: (i) maximizing the concurrent key establishment flow, (ii) minimizing the cost. We prove that both problems are NP-hard and MAX-SNP with a reduction to MAX3SAT problem.

The challenges of managing complexity in projects : an Australian perspective

Project Management (PM) as an academic field is relatively new in Australian universities. Moreover, the field is distributed across four main areas: business (management), built environment and construction, engineering and more recently ICT (information systems). At an institutional level, with notable exceptions, there is little engagement between researchers working in those individual areas. Consequently, an initiative was launched in 2009 to create a network of PM researchers to build a disciplinary base for PM in Australia. The initiative took the form of a bi-annual forum. The first forum established the constituency and spread of PM research in Australia (Sense et al., 2011). This special issue of IJPM arose out of the second forum, held in 2012, that explored the notion of an Australian perspective on PM. At the forum, researchers were invited to collaborate to explore issues, methodological approaches, and theoretical positions underpinning their research and to answer the question: is there a distinctly Australian research agenda which responds to the current challenges of large and complex projects in our region? From a research point of view, it was abundantly clear at the forum that many of the issues facing Australian researchers are shared around the world. However, what emerged from the forum as the Australian perspective was a set of themes and research issues that dominate the Australia research agenda.

Experiencing information across the phenomenon of health information literacy

Interest in the concept of information literacy in different contexts is a topic that has steadily gained increased attention in information literacy discourse and research efforts over recent years. In particular the emergence of the term ‘health information literacy’ attests to this interest and has elevated awareness about the importance and relevance of information literacy in a health context. This paper reports on research that has taken a relational perspective to explore how people experience health information literacy. Initially established by Bruce (1997) the relational perspective draws from an experiential framework that emphasizes the relationship between users and information when learning in different contexts. This approach seeks outcomes that are deeply embedded in users’ informational life worlds and complements experiential phenomenological perspectives that have been used in health research. In keeping with the relational approach this research interprets health information literacy as being the different ways in which people experience using information to learn about health. Using interpretive phenomenography, this research explored variation in the lived experience of how people use information to learn about their health, and variation in what is constituted as information when learning about their health. Participants included 23 males and females aged between 45 and 64 years. All participants were residents from the Greater Brisbane area of Queensland, Australia. The research used semi-structured interviews for data collection. The types of questions posed during interviews included ‘Can you describe a time when you used information to learn about your health?’ and ‘What kinds of information have you used to learn about your health?’. This paper will focus on presenting one element of research findings that concerns the differences in ‘what’ participants experienced as information. Analysis of interview data identified significant variation in the experienced nature of information, specifically the different qualities or elements that comprised the ‘object’ of information, or in other words, what was perceived as ‘informing’. Illustrations of this variation include information experienced as traditional information sources, facts and experiences, something exhibiting particular qualities, physical or psychological changes, other people and role models. These findings provide new insights into what people may experience as information, and build upon existing literature regarding information as a theoretical construct. In addition the potential implications of these findings with respect to the design and delivery of health information literacy education will also be discussed. These research findings contribute to the emergence of information literacy investigations in everyday life and community. Although such settings have long been identified as a significant gap for exploration, research to date in this field has predominantly focused on educational and workplace environments. In this way the knowledge gained from this research has further revealed the contextual nature of information literacy, as well as its complexity as a phenomenon and focus of study.

Complexity and the science of implementation in health IT — knowledge gaps and future visions

Objectives The intent of this paper is in the examination of health IT implementation processes – the barriers to and facilitators of successful implementation, identification of a beginning set of implementation best practices, the identification of gaps in the health IT implementation body of knowledge, and recommendations for future study and application. Methods A literature review resulted in the identification of six health IT related implementation best practices which were subsequently debated and clarified by participants attending the NI2012 Research Post Conference held in Montreal in the summer of 2012. Using the framework for implementation research (CFIR) to guide their application, the six best practices were applied to two distinct health IT implementation studies to assess their applicability. Results Assessing the implementation processes from two markedly diverse settings illustrated both the challenges and potentials of using standardized implementation processes. In support of what was discovered in the review of the literature, “one size fits all” in health IT implementation is a fallacy, particularly when global diversity is added into the mix. At the same time, several frameworks show promise for use as “scaffolding” to begin to assess best practices, their distinct dimensions, and their applicability for use. Conclusions Health IT innovations, regardless of the implementation setting, requires a close assessment of many dimensions. While there is no “one size fits all”, there are commonalities and best practices that can be blended, adapted, and utilized to improve the process of implementation. This paper examines health IT implementation processes and identifies a beginning set of implementation best practices, which could begin to address gaps in the health IT implementation body of knowledge.

A conceptualization of complexity in IS-driven organizational transformations

Organizational transformations reliant on successful ICT system developments (continue to) fail to deliver projected benefits even when contemporary governance models are applied rigorously. Modifications to traditional program, project and systems development management methods have produced little material improvement to successful transformation as they are unable to routinely address the complexity and uncertainty of dynamic alignment of IS investments and innovation. Complexity theory provides insight into why this phenomenon occurs and is used to develop a conceptualization of complexity in IS-driven organizational transformations. This research-in-progress aims to identify complexity formulations relevant to organizational transformation. Political/power based influences, interrelated business rules, socio-technical innovation, impacts on stakeholders and emergent behaviors are commonly considered as characterizing complexity while the proposed conceptualization accommodates these as connectivity, irreducibility, entropy and/or information gain in hierarchically approximation and scaling, number of states in a finite automata and/or dimension of attractor, and information and/or variety.

Learning information literacy

Introduction This paper reports on university students' experiences of learning information literacy. Method Phenomenography was selected as the research approach as it describes the experience from the perspective of the study participants, which in this case is a mixture of undergraduate and postgraduate students studying education at an Australian university. Semi-structured, one-on-one interviews were conducted with fifteen students. Analysis The interview transcripts were iteratively reviewed for similarities and differences in students' experiences of learning information literacy. Categories were constructed from an analysis of the distinct features of the experiences that students reported. The categories were grouped into a hierarchical structure that represents students' increasingly sophisticated experiences of learning information literacy. Results The study reveals that students experience learning information literacy in six ways: learning to find information; learning a process to use information; learning to use information to create a product; learning to use information to build a personal knowledge base; learning to use information to advance disciplinary knowledge; and learning to use information to grow as a person and to contribute to others. Conclusions Understanding the complexity of the concept of information literacy, and the collective and diverse range of ways students experience learning information literacy, enables academics and librarians to draw on the range of experiences reported by students to design academic curricula and information literacy education that targets more powerful ways of learning to find and use information.

Active awareness : supporting the intentional disclosure of awareness information in collaborative systems

This thesis opens up the design space for awareness research in CSCW and HCI. By challenging the prevalent understanding of roles in awareness processes and exploring different mechanisms for actively engaging users in the awareness process, this thesis provides a better understanding of the complexity of these processes and suggests practical solutions for designing and implementing systems that support active awareness. Mutual awareness, a prominent research topic in the fields of Computer-Supported Cooperative Work (CSCW) and Human-Computer Interaction (HCI) refers to a fundamental aspect of a person’s work: their ability to gain a better understanding of a situation by perceiving and interpreting their co-workers actions. Technologically-mediated awareness, used to support co-workers across distributed settings, distinguishes between the roles of the actor, whose actions are often limited to being the target of an automated data gathering processes, and the receiver, who wants to be made aware of the actors’ actions. This receiver-centric view of awareness, focusing on helping receivers to deal with complex sets of awareness information, stands in stark contrast to our understanding of awareness as social process involving complex interactions between both actors and receivers. It fails to take into account an actors’ intimate understanding of their own activities and the contribution that this subjective understanding could make in providing richer awareness information. In this thesis I challenge the prevalent receiver-centric notion of awareness, and explore the conceptual foundations, design, implementation and evaluation of an alternative active awareness approach by making the following five contributions. Firstly, I identify the limitations of existing awareness research and solicit further evidence to support the notion of active awareness. I analyse ethnographic workplace studies that demonstrate how actors engage in an intricate interplay involving the monitoring of their co-workers progress and displaying aspects of their activities that may be of relevance to others. The examination of a large body of awareness research reveals that while disclosing information is a common practice in face-to-face collaborative settings it has been neglected in implementations of technically mediated awareness. Based on these considerations, I introduce the notion of intentional disclosure to describe the action of users actively and deliberately contributing awareness information. I consider challenges and potential solutions for the design of active awareness. I compare a range of systems, each allowing users to share information about their activities at various levels of detail. I discuss one of the main challenges to active awareness: that disclosing information about activities requires some degree of effort. I discuss various representations of effort in collaborative work. These considerations reveal that there is a trade-off between the richness of awareness information and the effort required to provide this information. I propose a framework for active awareness, aimed to help designers to understand the scope and limitations of different types of intentional disclosure. I draw on the identified richness/effort trade-off to develop two types of intentional disclosure, both of which aim to facilitate the disclosure of information while reducing the effort required to do so. For both of these approaches, direct and indirect disclosure, I delineate how they differ from related approaches and define a set of design criteria that is intended to guide their implementation. I demonstrate how the framework of active awareness can be practically applied by building two proof-of-concept prototypes that implement direct and indirect disclosure respectively. AnyBiff, implementing direct disclosure, allows users to create, share and use shared representations of activities in order to express their current actions and intentions. SphereX, implementing indirect disclosure, represents shared areas of interests or working context, and links sets of activities to these representations. Lastly, I present the results of the qualitative evaluation of the two prototypes and analyse the results with regard to the extent to which they implemented their respective disclosure mechanisms and supported active awareness. Both systems were deployed and tested in real world environments. The results for AnyBiff showed that users developed a wide range of activity representations, some unanticipated, and actively used the system to disclose information. The results further highlighted a number of design considerations relating to the relationship between awareness and communication, and the role of ambiguity. The evaluation of SphereX validated the feasibility of the indirect disclosure approach. However, the study highlighted the challenges of implementing cross-application awareness support and translating the concept to users. The study resulted in design recommendations aimed to improve the implementation of future systems.