Open problems in the security of learning

Machine learning has become a valuable tool for detecting and preventing malicious activity. However, as more applications employ machine learning techniques in adversarial decision-making situations, increasingly powerful attacks become possible against machine learning systems. In this paper, we present three broad research directions towards the end of developing truly secure learning. First, we suggest that finding bounds on adversarial influence is important to understand the limits of what an attacker can and cannot do to a learning system. Second, we investigate the value of adversarial capabilities-the success of an attack depends largely on what types of information and influence the attacker has. Finally, we propose directions in technologies for secure learning and suggest lines of investigation into secure techniques for learning in adversarial environments. We intend this paper to foster discussion about the security of machine learning, and we believe that the research directions we propose represent the most important directions to pursue in the quest for secure learning.

Learning to act in uncertain environments : technical perspective

The problem of decision making in an uncertain environment arises in many diverse contexts: deciding whether to keep a hard drive spinning in a net-book; choosing which advertisement to post to a Web site visitor; choosing how many newspapers to order so as to maximize profits; or choosing a route to recommend to a driver given limited and possibly out-of-date information about traffic conditions. All are sequential decision problems, since earlier decisions affect subsequent performance; all require adaptive approaches, since they involve significant uncertainty. The key issue in effectively solving problems like these is known as the exploration/exploitation trade-off: If I am at a cross-roads, when should I go in the most advantageous direction among those that I have already explored, and when should I strike out in a new direction, in the hopes I will discover something better?

Implicit online learning

Online learning algorithms have recently risen to prominence due to their strong theoretical guarantees and an increasing number of practical applications for large-scale data analysis problems. In this paper, we analyze a class of online learning algorithms based on fixed potentials and nonlinearized losses, which yields algorithms with implicit update rules. We show how to efficiently compute these updates, and we prove regret bounds for the algorithms. We apply our formulation to several special cases where our approach has benefits over existing online learning methods. In particular, we provide improved algorithms and bounds for the online metric learning problem, and show improved robustness for online linear prediction problems. Results over a variety of data sets demonstrate the advantages of our framework.

A unifying view of multiple kernel learning

Recent research on multiple kernel learning has lead to a number of approaches for combining kernels in regularized risk minimization. The proposed approaches include different formulations of objectives and varying regularization strategies. In this paper we present a unifying general optimization criterion for multiple kernel learning and show how existing formulations are subsumed as special cases. We also derive the criterion's dual representation, which is suitable for general smooth optimization algorithms. Finally, we evaluate multiple kernel learning in this framework analytically using a Rademacher complexity bound on the generalization error and empirically in a set of experiments.

Matrix regularization techniques for online multitask learning

In this paper we examine the problem of prediction with expert advice in a setup where the learner is presented with a sequence of examples coming from different tasks. In order for the learner to be able to benefit from performing multiple tasks simultaneously, we make assumptions of task relatedness by constraining the comparator to use a lesser number of best experts than the number of tasks. We show how this corresponds naturally to learning under spectral or structural matrix constraints, and propose regularization techniques to enforce the constraints. The regularization techniques proposed here are interesting in their own right and multitask learning is just one application for the ideas. A theoretical analysis of one such regularizer is performed, and a regret bound that shows benefits of this setup is reported.

Multitask learning with expert advice

We consider the problem of prediction with expert advice in the setting where a forecaster is presented with several online prediction tasks. Instead of competing against the best expert separately on each task, we assume the tasks are related, and thus we expect that a few experts will perform well on the entire set of tasks. That is, our forecaster would like, on each task, to compete against the best expert chosen from a small set of experts. While we describe the "ideal" algorithm and its performance bound, we show that the computation required for this algorithm is as hard as computation of a matrix permanent. We present an efficient algorithm based on mixing priors, and prove a bound that is nearly as good for the sequential task presentation case. We also consider a harder case where the task may change arbitrarily from round to round, and we develop an efficient approximate randomized algorithm based on Markov chain Monte Carlo techniques.

Learning the Kernel Matrix with Semi-Definite Programming

Kernel-based learning algorithms work by embedding the data into a Euclidean space, and then searching for linear relations among the embedded data points. The embedding is performed implicitly, by specifying the inner products between each pair of points in the embedding space. This information is contained in the so-called kernel matrix, a symmetric and positive definite matrix that encodes the relative positions of all points. Specifying this matrix amounts to specifying the geometry of the embedding space and inducing a notion of similarity in the input space -- classical model selection problems in machine learning. In this paper we show how the kernel matrix can be learned from data via semi-definite programming (SDP) techniques. When applied to a kernel matrix associated with both training and test data this gives a powerful transductive algorithm -- using the labelled part of the data one can learn an embedding also for the unlabelled part. The similarity between test points is inferred from training points and their labels. Importantly, these learning problems are convex, so we obtain a method for learning both the model class and the function without local minima. Furthermore, this approach leads directly to a convex method to learn the 2-norm soft margin parameter in support vector machines, solving another important open problem. Finally, the novel approach presented in the paper is supported by positive empirical results.

Margin-adaptive model selection in statistical learning

A classical condition for fast learning rates is the margin condition, first introduced by Mammen and Tsybakov. We tackle in this paper the problem of adaptivity to this condition in the context of model selection, in a general learning framework. Actually, we consider a weaker version of this condition that allows one to take into account that learning within a small model can be much easier than within a large one. Requiring this “strong margin adaptivity” makes the model selection problem more challenging. We first prove, in a general framework, that some penalization procedures (including local Rademacher complexities) exhibit this adaptivity when the models are nested. Contrary to previous results, this holds with penalties that only depend on the data. Our second main result is that strong margin adaptivity is not always possible when the models are not nested: for every model selection procedure (even a randomized one), there is a problem for which it does not demonstrate strong margin adaptivity.

Application of machine learning in fault diagnostics of mechanical systems

A diagnostic method based on Bayesian Networks (probabilistic graphical models) is presented. Unlike conventional diagnostic approaches, in this method instead of focusing on system residuals at one or a few operating points, diagnosis is done by analyzing system behavior patterns over a window of operation. It is shown how this approach can loosen the dependency of diagnostic methods on precise system modeling while maintaining the desired characteristics of fault detection and diagnosis (FDD) tools (fault isolation, robustness, adaptability, and scalability) at a satisfactory level. As an example, the method is applied to fault diagnosis in HVAC systems, an area with considerable modeling and sensor network constraints.

REGAL : a regularization based algorithm for reinforcement learning in weakly communicating MDPs

We provide an algorithm that achieves the optimal regret rate in an unknown weakly communicating Markov Decision Process (MDP). The algorithm proceeds in episodes where, in each episode, it picks a policy using regularization based on the span of the optimal bias vector. For an MDP with S states and A actions whose optimal bias vector has span bounded by H, we show a regret bound of ~ O(HS p AT ). We also relate the span to various diameter-like quantities associated with the MDP, demonstrating how our results improve on previous regret bounds.

An ecological footprint for an early learning centre : identifying opportunities for early childhood sustainability education through interdisciplinary research

In this study, engineers and educators worked together to adapt and apply the ecological footprint (EF) methodology to an early learning centre in Brisbane, Australia. Results were analysed to determine how environmental impact can be reduced at the study site and more generally across early childhood settings. It was found that food, transport and energy consumption had the largest impact on the centre’s overall footprint. In transport and energy, early childhood centres can reduce their impact through infrastructure and cultural change, in association with changed curriculum strategies. Building design, the type of energy purchased and appliance usage can all be modified to reduce the energy footprint. The transport footprint can be reduced through more families using active and public transport, which can be encouraged by providing information, support and facilities and appropriate siting of new centres. Introducing the concept of ecological footprint in early childhood education may be an effective way to educate children, staff and parents on the links between the food they eat, land usage and environmental impact. This study responds directly to the call in this journal for research focused on early childhood education and for more to be made of interdisciplinary research opportunities.

Objectifying early-number : a visual nomenclature to express mathematical domain knowledge

To address issues of divisive ideologies in the Mathematics Education community and to subsequently advance educational practice, an alternative theoretical framework and operational model is proposed which represents a consilience of constructivist learning theories whilst acknowledging the objective but improvable nature of domain knowledge. Based upon Popper’s three-world model of knowledge, the proposed theory supports the differentiation and explicit modelling of both shared domain knowledge and idiosyncratic personal understanding using a visual nomenclature. The visual nomenclature embodies Piaget’s notion of reflective abstraction and so may support an individual’s experience-based transformation of personal understanding with regards to shared domain knowledge. Using the operational model and visual nomenclature, seminal literature regarding early-number counting and addition was analysed and described. Exemplars of the resultant visual artefacts demonstrate the proposed theory’s viability as a tool with which to characterise the reflective abstraction-based organisation of a domain’s shared knowledge. Utilising such a description of knowledge, future research needs to consider the refinement of the operational model and visual nomenclature to include the analysis, description and scaffolded transformation of personal understanding. A detailed model of knowledge and understanding may then underpin the future development of educational software tools such as computer-mediated teaching and learning environments.

A Popperian consilience : modelling mathematical knowledge and understanding

Goldin (2003) and McDonald, Yanchar, and Osguthorpe (2005) have called for mathematics learning theory that reconciles the chasm between ideologies, and which may advance mathematics teaching and learning practice. This paper discusses the theoretical underpinnings of a recently completed PhD study that draws upon Popper’s (1978) three-world model of knowledge as a lens through which to reconsider a variety of learning theories, including Piaget’s reflective abstraction. Based upon this consideration of theories, an alternative theoretical framework and complementary operational model was synthesised, the viability of which was demonstrated by its use to analyse the domain of early-number counting, addition and subtraction.

Using ‘think aloud’ as a strategy for learning clinical reasoning in high fidelity case-based simulation for undergraduate nursing students

AIMS This paper reports on the implementation of a research project that trials an educational strategy implemented over six months of an undergraduate third year nursing curriculum. This project aims to explore the effectiveness of ‘think aloud’ as a strategy for learning clinical reasoning for students in simulated clinical settings. BACKGROUND Nurses are required to apply and utilise critical thinking skills to enable clinical reasoning and problem solving in the clinical setting [1]. Nursing students are expected to develop and display clinical reasoning skills in practice, but may struggle articulating reasons behind decisions about patient care. For students learning to manage complex clinical situations, teaching approaches are required that make these instinctive cognitive processes explicit and clear [2-5]. In line with professional expectations, nursing students in third year at Queensland University of Technology (QUT) are expected to display clinical reasoning skills in practice. This can be a complex proposition for students in practice situations, particularly as the degree of uncertainty or decision complexity increases [6-7]. The ‘think aloud’ approach is an innovative learning/teaching method which can create an environment suitable for developing clinical reasoning skills in students [4, 8]. This project aims to use the ‘think aloud’ strategy within a simulation context to provide a safe learning environment in which third year students are assisted to uncover cognitive approaches that best assist them to make effective patient care decisions, and improve their confidence, clinical reasoning and active critical reflection on their practice. MEHODS In semester 2 2011 at QUT, third year nursing students will undertake high fidelity simulation, some for the first time commencing in September of 2011. There will be two cohorts for strategy implementation (group 1= use think aloud as a strategy within the simulation, group 2= not given a specific strategy outside of nursing assessment frameworks) in relation to problem solving patient needs. Students will be briefed about the scenario, given a nursing handover, placed into a simulation group and an observer group, and the facilitator/teacher will run the simulation from a control room, and not have contact (as a ‘teacher’) with students during the simulation. Then debriefing will occur as a whole group outside of the simulation room where the session can be reviewed on screen. The think aloud strategy will be described to students in their pre-simulation briefing and allow for clarification of this strategy at this time. All other aspects of the simulations remain the same, (resources, suggested nursing assessment frameworks, simulation session duration, size of simulation teams, preparatory materials). RESULTS Methodology of the project and the challenges of implementation will be the focus of this presentation. This will include ethical considerations in designing the project, recruitment of students and implementation of a voluntary research project within a busy educational curriculum which in third year targets 669 students over two campuses. CONCLUSIONS In an environment of increasingly constrained clinical placement opportunities, exploration of alternate strategies to improve critical thinking skills and develop clinical reasoning and problem solving for nursing students is imperative in preparing nurses to respond to changing patient needs. References 1. Lasater, K., High-fidelity simulation and the development of clinical judgement: students' experiences. Journal of Nursing Education, 2007. 46(6): p. 269-276. 2. Lapkin, S., et al., Effectiveness of patient simulation manikins in teaching clinical reasoning skills to undergraduate nursing students: a systematic review. Clinical Simulation in Nursing, 2010. 6(6): p. e207-22. 3. Kaddoura, M.P.C.M.S.N.R.N., New Graduate Nurses' Perceptions of the Effects of Clinical Simulation on Their Critical Thinking, Learning, and Confidence. The Journal of Continuing Education in Nursing, 2010. 41(11): p. 506. 4. Banning, M., The think aloud approach as an educational tool to develop and assess clinical reasoning in undergraduate students. Nurse Education Today, 2008. 28: p. 8-14. 5. Porter-O'Grady, T., Profound change:21st century nursing. Nursing Outlook, 2001. 49(4): p. 182-186. 6. Andersson, A.K., M. Omberg, and M. Svedlund, Triage in the emergency department-a qualitative study of the factors which nurses consider when making decisions. Nursing in Critical Care, 2006. 11(3): p. 136-145. 7. O'Neill, E.S., N.M. Dluhy, and C. Chin, Modelling novice clinical reasoning for a computerized decision support system. Journal of Advanced Nursing, 2005. 49(1): p. 68-77. 8. Lee, J.E. and N. Ryan-Wenger, The "Think Aloud" seminar for teaching clinical reasoning: a case study of a child with pharyngitis. J Pediatr Health Care, 1997. 11(3): p. 101-10.