Optimal online prediction in adversarial environments

In many prediction problems, including those that arise in computer security and computational finance, the process generating the data is best modelled as an adversary with whom the predictor competes. Even decision problems that are not inherently adversarial can be usefully modeled in this way, since the assumptions are sufficiently weak that effective prediction strategies for adversarial settings are very widely applicable.

Optimal online prediction in adversarial environments

In many prediction problems, including those that arise in computer security and computational finance, the process generating the data is best modelled as an adversary with whom the predictor competes. Even decision problems that are not inherently adversarial can be usefully modeled in this way, since the assumptions are sufficiently weak that effective prediction strategies for adversarial settings are very widely applicable.

Optimal strategies and minimax lower bounds for online convex games

A number of learning problems can be cast as an Online Convex Game: on each round, a learner makes a prediction x from a convex set, the environment plays a loss function f, and the learner’s long-term goal is to minimize regret. Algorithms have been proposed by Zinkevich, when f is assumed to be convex, and Hazan et al., when f is assumed to be strongly convex, that have provably low regret. We consider these two settings and analyze such games from a minimax perspective, proving minimax strategies and lower bounds in each case. These results prove that the existing algorithms are essentially optimal.

Inequitable Provision of Optimal Services for Patients with Chronic Heart failure: A National Geo-Mapping Study

Objective: To compare the location and accessibility of current Australian chronic heart failure (CHF) management programs and general practice services with the probable distribution of the population with CHF. Design and setting: Data on the prevalence and distribution of the CHF population throughout Australia, and the locations of CHF management programs and general practice services from 1 January 2004 to 31 December 2005 were analysed using geographic information systems (GIS) technology. Outcome measures: Distance of populations with CHF to CHF management programs and general practice services. Results: The highest prevalence of CHF (20.3–79.8 per 1000 population) occurred in areas with high concentrations of people over 65 years of age and in areas with higher proportions of Indigenous people. Five thousand CHF patients (8%) discharged from hospital in 2004–2005 were managed in one of the 62 identified CHF management programs. There were no CHF management programs in the Northern Territory or Tasmania. Only four CHF management programs were located outside major cities, with a total case load of 80 patients (0.7%). The mean distance from any Australian population centre to the nearest CHF management program was 332 km (median, 163 km; range, 0.15–3246 km). In rural areas, where the burden of CHF management falls upon general practitioners, the mean distance to general practice services was 37 km (median, 20 km; range, 0–656 km). Conclusion: There is an inequity in the provision of CHF management programs to rural Australians.

Identifying optimal lipid raft characteristics required to promote nanoscale protein-protein interactions on the plasma membrane

The dynamic lateral segregation of signaling proteins into microdomains is proposed to facilitate signal transduction, but the constraints on microdomain size, mobility, and diffusion that might realize this function are undefined. Here we interrogate a stochastic spatial model of the plasma membrane to determine how microdomains affect protein dynamics. Taking lipid rafts as representative microdomains, we show that reduced protein mobility in rafts segregates dynamically partitioning proteins, but the equilibrium concentration is largely independent of raft size and mobility. Rafts weakly impede small-scale protein diffusion but more strongly impede long-range protein mobility. The long-range mobility of raft-partitioning and raft-excluded proteins, however, is reduced to a similar extent. Dynamic partitioning into rafts increases specific interprotein collision rates, but to maximize this critical, biologically relevant function, rafts must be small (diameter, 6 to 14 nm) and mobile. Intermolecular collisions can also be favored by the selective capture and exclusion of proteins by rafts, although this mechanism is generally less efficient than simple dynamic partitioning. Generalizing these results, we conclude that microdomains can readily operate as protein concentrators or isolators but there appear to be significant constraints on size and mobility if microdomains are also required to function as reaction chambers that facilitate nanoscale protein-protein interactions. These results may have significant implications for the many signaling cascades that are scaffolded or assembled in plasma membrane microdomains.

Optimal allocation strategies for the dark pool problem

We study the problem of allocating stocks to dark pools. We propose and analyze an optimal approach for allocations, if continuous-valued allocations are allowed. We also propose a modification for the case when only integer-valued allocations are possible. We extend the previous work on this problem to adversarial scenarios, while also improving on their results in the iid setup. The resulting algorithms are efficient, and perform well in simulations under stochastic and adversarial inputs.

Evaluating multivariate volatility forecasts : how effective are statistical and economic loss functions?

Multivariate volatility forecasts are an important input in many financial applications, in particular portfolio optimisation problems. Given the number of models available and the range of loss functions to discriminate between them, it is obvious that selecting the optimal forecasting model is challenging. The aim of this thesis is to thoroughly investigate how effective many commonly used statistical (MSE and QLIKE) and economic (portfolio variance and portfolio utility) loss functions are at discriminating between competing multivariate volatility forecasts. An analytical investigation of the loss functions is performed to determine whether they identify the correct forecast as the best forecast. This is followed by an extensive simulation study examines the ability of the loss functions to consistently rank forecasts, and their statistical power within tests of predictive ability. For the tests of predictive ability, the model confidence set (MCS) approach of Hansen, Lunde and Nason (2003, 2011) is employed. As well, an empirical study investigates whether simulation findings hold in a realistic setting. In light of these earlier studies, a major empirical study seeks to identify the set of superior multivariate volatility forecasting models from 43 models that use either daily squared returns or realised volatility to generate forecasts. This study also assesses how the choice of volatility proxy affects the ability of the statistical loss functions to discriminate between forecasts. Analysis of the loss functions shows that QLIKE, MSE and portfolio variance can discriminate between multivariate volatility forecasts, while portfolio utility cannot. An examination of the effective loss functions shows that they all can identify the correct forecast at a point in time, however, their ability to discriminate between competing forecasts does vary. That is, QLIKE is identified as the most effective loss function, followed by portfolio variance which is then followed by MSE. The major empirical analysis reports that the optimal set of multivariate volatility forecasting models includes forecasts generated from daily squared returns and realised volatility. Furthermore, it finds that the volatility proxy affects the statistical loss functions’ ability to discriminate between forecasts in tests of predictive ability. These findings deepen our understanding of how to choose between competing multivariate volatility forecasts.

Optimal budget allocation in budget-based access control

In dynamic and uncertain environments, where the needs of security and information availability are difficult to balance, an access control approach based on a static policy will be suboptimal regardless of how comprehensive it is. Risk-based approaches to access control attempt to address this problem by allocating a limited budget to users, through which they pay for the exceptions deemed necessary. So far the primary focus has been on how to incorporate the notion of budget into access control rather than what or if there is an optimal amount of budget to allocate to users. In this paper we discuss the problems that arise from a sub-optimal allocation of budget and introduce a generalised characterisation of an optimal budget allocation function that maximises organisations expected benefit in the presence of self-interested employees and costly audit.

Bioimpedance in the assessment of unilateral lymphedema of a limb: the optimal frequency

Background: Bioimpedance techniques provide a reliable method of assessing unilateral lymphedema in a clinical setting. Bioimpedance devices are traditionally used to assess body composition at a current frequency of 50 kHz. However, these devices are not transferable to the assessment of lymphedema, as the sensitivity of measuring the impedance of extracellular fluid is frequency dependent. It has previously been shown that the best frequency to detect extracellular fluid is 0 kHz (or DC). However, measurement at this frequency is not possible in practice due to the high skin impedance at DC, and an estimate is usually determined from low frequency measurements. This study investigated the efficacy of various low frequency ranges for the detection of lymphedema. Methods and Results: Limb impedance was measured at 256 frequencies between 3 kHz and 1000 kHz for a sample control population, arm lymphedema population, and leg lymphedema population. Limb impedance was measured using the ImpediMed SFB7 and ImpediMed L-Dex® U400 with equipotential electrode placement on the wrists and ankles. The contralateral limb impedance ratio for arms and legs was used to calculate a lymphedema index (L-Dex) at each measurement frequency. The standard deviation of the limb impedance ratio in a healthy control population has been shown to increase with frequency for both the arm and leg. Box and whisker plots of the spread of the control and lymphedema populations show that there exists good differentiation between the arm and leg L-Dex measured for lymphedema subjects and the arm and leg L-Dex measured for control subjects up to a frequency of about 30 kHz. Conclusions: It can be concluded that impedance measurements above a frequency of 30 kHz decrease sensitivity to extracellular fluid and are not reliable for early detection of lymphedema.

The use of semantic priming as a paradigm for the detection of faking on self-report psychological tests

It is recognised that individuals do not always respond honestly when completing psychological tests. One of the foremost issues for research in this area is the inability to detect individuals attempting to fake. While a number of strategies have been identified in faking, a commonality of these strategies is the latent role of long term memory. Seven studies were conducted in order to examine whether it is possible to detect the activation of faking related cognitions using a lexical decision task. Study 1 found that engagement with experiential processing styles predicted the ability to fake successfully, confirming the role of associative processing styles in faking. After identifying appropriate stimuli for the lexical decision task (Studies 2A and 2B), Studies 3 to 5 examined whether a cognitive state of faking could be primed and subsequently identified, using a lexical decision task. Throughout the course of these studies, the experimental methodology was increasingly refined in an attempt to successfully identify the relevant priming mechanisms. The results were consistent and robust throughout the three priming studies: faking good on a personality test primed positive faking related words in the lexical decision tasks. Faking bad, however, did not result in reliable priming of negative faking related cognitions. To more completely address potential issues with the stimuli and the possible role of affective priming, two additional studies were conducted. Studies 6A and 6B revealed that negative faking related words were more arousing than positive faking related words, and that positive faking related words were more abstract than negative faking related words and neutral words. Study 7 examined whether the priming effects evident in the lexical decision tasks occurred as a result of an unintentional mood induction while faking the psychological tests. Results were equivocal in this regard. This program of research aligned the fields of psychological assessment and cognition to inform the preliminary development and validation of a new tool to detect faking. Consequently, an implicit technique to identify attempts to fake good on a psychological test has been identified, using long established and robust cognitive theories in a novel and innovative way. This approach represents a new paradigm for the detection of individuals responding strategically to psychological testing. With continuing development and validation, this technique may have immense utility in the field of psychological assessment.