Showing papers by "Deakin University published in 2020"

A Review of Uncertainty Quantification in Deep Learning: Techniques, Applications and Challenges

Abstract: Uncertainty quantification (UQ) plays a pivotal role in reduction of uncertainties during both optimization and decision making processes. It can be applied to solve a variety of real-world applications in science and engineering. Bayesian approximation and ensemble learning techniques are two most widely-used UQ methods in the literature. In this regard, researchers have proposed different UQ methods and examined their performance in a variety of applications such as computer vision (e.g., self-driving cars and object detection), image processing (e.g., image restoration), medical image analysis (e.g., medical image classification and segmentation), natural language processing (e.g., text classification, social media texts and recidivism risk-scoring), bioinformatics, etc. This study reviews recent advances in UQ methods used in deep learning. Moreover, we also investigate the application of these methods in reinforcement learning (RL). Then, we outline a few important applications of UQ methods. Finally, we briefly highlight the fundamental research challenges faced by UQ methods and discuss the future research directions in this field.

Deep Reinforcement Learning for Multiagent Systems: A Review of Challenges, Solutions, and Applications

Abstract: Reinforcement learning (RL) algorithms have been around for decades and employed to solve various sequential decision-making problems. These algorithms, however, have faced great challenges when dealing with high-dimensional environments. The recent development of deep learning has enabled RL methods to drive optimal policies for sophisticated and capable agents, which can perform efficiently in these challenging environments. This article addresses an important aspect of deep RL related to situations that require multiple agents to communicate and cooperate to solve complex tasks. A survey of different approaches to problems related to multiagent deep RL (MADRL) is presented, including nonstationarity, partial observability, continuous state and action spaces, multiagent training schemes, and multiagent transfer learning. The merits and demerits of the reviewed methods will be analyzed and discussed with their corresponding applications explored. It is envisaged that this review provides insights about various MADRL methods and can lead to the future development of more robust and highly useful multiagent learning methods for solving real-world problems.

Scalable Manufacturing of Free‐Standing, Strong Ti 3 C 2 T x MXene Films with Outstanding Conductivity

Abstract: Free-standing films that display high strength and high electrical conductivity are critical for flexible electronics, such as electromagnetic interference (EMI) shielding coatings and current collectors for batteries and supercapacitors. 2D Ti3 C2 Tx flakes are ideal candidates for making conductive films due to their high strength and metallic conductivity. It is, however, challenging to transfer those outstanding properties of single MXene flakes to macroscale films as a result of the small flake size and relatively poor flake alignment that occurs during solution-based processing. Here, a scalable method is shown for the fabrication of strong and highly conducting pure MXene films containing highly aligned large MXene flakes. These films demonstrate record tensile strength up to ≈570 MPa for a 940 nm thick film and electrical conductivity of ≈15 100 S cm-1 for a 214 nm thick film, which are both the highest values compared to previously reported pure Ti3 C2 Tx films. These films also exhibit outstanding EMI shielding performance (≈50 dB for a 940 nm thick film) that exceeds other synthetic materials with comparable thickness. MXene films with aligned flakes provide an effective route for producing large-area, high-strength, and high-electrical-conductivity MXene-based films for future electronic applications.

Country Responses and the Reaction of the Stock Market to COVID-19—a Preliminary Exposition

Abstract: As the coronavirus pandemic (COVID-19) has amplified so has country responses to it. With COVID-19 taking its toll on humans, as reflected in the number of people infected by, and deaths from, COVI...

A global outlook to the interruption of education due to COVID-19 pandemic: Navigating in a time of uncertainty and crisis

Abstract: Uncertain times require prompt reflexes to survive and this study is a collaborative reflex to better understand uncertainty and navigate through it. The Coronavirus (Covid-19) pandemic hit hard and interrupted many dimensions of our lives, particularly education. As a response to interruption of education due to the Covid-19 pandemic, this study is a collaborative reaction that narrates the overall view, reflections from the K-12 and higher educational landscape, lessons learned and suggestions from a total of 31 countries across the world with a representation of 62,7% of the whole world population. In addition to the value of each case by country, the synthesis of this research suggests that the current practices can be defined as emergency remote education and this practice is different from planned practices such as distance education, online learning or other derivations. Above all, this study points out how social injustice, inequity and the digital divide have been exacerbated during the pandemic and need unique and targeted measures if they are to be addressed. While there are support communities and mechanisms, parents are overburdened between regular daily/professional duties and emerging educational roles, and all parties are experiencing trauma, psychological pressure and anxiety to various degrees, which necessitates a pedagogy of care, affection and empathy. In terms of educational processes, the interruption of education signifies the importance of openness in education and highlights issues that should be taken into consideration such as using alternative assessment and evaluation methods as well as concerns about surveillance, ethics, and data privacy resulting from nearly exclusive dependency on online solutions.

Progress, opportunity, and perspective on exosome isolation - efforts for efficient exosome-based theranostics.

Abstract: Exosomes are small extracellular vesicles with diameters of 30-150 nm. In both physiological and pathological conditions, nearly all types of cells can release exosomes, which play important roles in cell communication and epigenetic regulation by transporting crucial protein and genetic materials such as miRNA, mRNA, and DNA. Consequently, exosome-based disease diagnosis and therapeutic methods have been intensively investigated. However, as in any natural science field, the in-depth investigation of exosomes relies heavily on technological advances. Historically, the two main technical hindrances that have restricted the basic and applied researches of exosomes include, first, how to simplify the extraction and improve the yield of exosomes and, second, how to effectively distinguish exosomes from other extracellular vesicles, especially functional microvesicles. Over the past few decades, although a standardized exosome isolation method has still not become available, a number of techniques have been established through exploration of the biochemical and physicochemical features of exosomes. In this work, by comprehensively analyzing the progresses in exosome separation strategies, we provide a panoramic view of current exosome isolation techniques, providing perspectives toward the development of novel approaches for high-efficient exosome isolation from various types of biological matrices. In addition, from the perspective of exosome-based diagnosis and therapeutics, we emphasize the issue of quantitative exosome and microvesicle separation.

ENIGMA and global neuroscience: A decade of large-scale studies of the brain in health and disease across more than 40 countries

Abstract: This review summarizes the last decade of work by the ENIGMA (Enhancing NeuroImaging Genetics through Meta Analysis) Consortium, a global alliance of over 1400 scientists across 43 countries, studying the human brain in health and disease. Building on large-scale genetic studies that discovered the first robustly replicated genetic loci associated with brain metrics, ENIGMA has diversified into over 50 working groups (WGs), pooling worldwide data and expertise to answer fundamental questions in neuroscience, psychiatry, neurology, and genetics. Most ENIGMA WGs focus on specific psychiatric and neurological conditions, other WGs study normal variation due to sex and gender differences, or development and aging; still other WGs develop methodological pipelines and tools to facilitate harmonized analyses of "big data" (i.e., genetic and epigenetic data, multimodal MRI, and electroencephalography data). These international efforts have yielded the largest neuroimaging studies to date in schizophrenia, bipolar disorder, major depressive disorder, post-traumatic stress disorder, substance use disorders, obsessive-compulsive disorder, attention-deficit/hyperactivity disorder, autism spectrum disorders, epilepsy, and 22q11.2 deletion syndrome. More recent ENIGMA WGs have formed to study anxiety disorders, suicidal thoughts and behavior, sleep and insomnia, eating disorders, irritability, brain injury, antisocial personality and conduct disorder, and dissociative identity disorder. Here, we summarize the first decade of ENIGMA's activities and ongoing projects, and describe the successes and challenges encountered along the way. We highlight the advantages of collaborative large-scale coordinated data analyses for testing reproducibility and robustness of findings, offering the opportunity to identify brain systems involved in clinical syndromes across diverse samples and associated genetic, environmental, demographic, cognitive, and psychosocial factors.

Ultra-processed foods and the nutrition transition: Global, regional and national trends, food systems transformations and political economy drivers.

Abstract: Understanding the drivers and dynamics of global ultra-processed food (UPF) consumption is essential, given the evidence linking these foods with adverse health outcomes. In this synthesis review, we take two steps. First, we quantify per capita volumes and trends in UPF sales, and ingredients (sweeteners, fats, sodium and cosmetic additives) supplied by these foods, in countries classified by income and region. Second, we review the literature on food systems and political economy factors that likely explain the observed changes. We find evidence for a substantial expansion in the types and quantities of UPFs sold worldwide, representing a transition towards a more processed global diet but with wide variations between regions and countries. As countries grow richer, higher volumes and a wider variety of UPFs are sold. Sales are highest in Australasia, North America, Europe and Latin America but growing rapidly in Asia, the Middle East and Africa. These developments are closely linked with the industrialization of food systems, technological change and globalization, including growth in the market and political activities of transnational food corporations and inadequate policies to protect nutrition in these new contexts. The scale of dietary change underway, especially in highly populated middle-income countries, raises serious concern for global health.

Effect of Vitamin C, Hydrocortisone, and Thiamine vs Hydrocortisone Alone on Time Alive and Free of Vasopressor Support Among Patients With Septic Shock: The VITAMINS Randomized Clinical Trial

Abstract: Importance It is unclear whether vitamin C, hydrocortisone, and thiamine are more effective than hydrocortisone alone in expediting resolution of septic shock. Objective To determine whether the combination of vitamin C, hydrocortisone, and thiamine, compared with hydrocortisone alone, improves the duration of time alive and free of vasopressor administration in patients with septic shock. Design, Setting, and Participants Multicenter, open-label, randomized clinical trial conducted in 10 intensive care units in Australia, New Zealand, and Brazil that recruited 216 patients fulfilling the Sepsis-3 definition of septic shock. The first patient was enrolled on May 8, 2018, and the last on July 9, 2019. The final date of follow-up was October 6, 2019. Interventions Patients were randomized to the intervention group (n = 109), consisting of intravenous vitamin C (1.5 g every 6 hours), hydrocortisone (50 mg every 6 hours), and thiamine (200 mg every 12 hours), or to the control group (n = 107), consisting of intravenous hydrocortisone (50 mg every 6 hours) alone until shock resolution or up to 10 days. Main Outcomes and Measures The primary trial outcome was duration of time alive and free of vasopressor administration up to day 7. Ten secondary outcomes were prespecified, including 90-day mortality. Results Among 216 patients who were randomized, 211 provided consent and completed the primary outcome measurement (mean age, 61.7 years [SD, 15.0]; 133 men [63%]). Time alive and vasopressor free up to day 7 was 122.1 hours (interquartile range [IQR], 76.3-145.4 hours) in the intervention group and 124.6 hours (IQR, 82.1-147.0 hours) in the control group; the median of all paired differences was –0.6 hours (95% CI, –8.3 to 7.2 hours;P = .83). Of 10 prespecified secondary outcomes, 9 showed no statistically significant difference. Ninety-day mortality was 30/105 (28.6%) in the intervention group and 25/102 (24.5%) in the control group (hazard ratio, 1.18; 95% CI, 0.69-2.00). No serious adverse events were reported. Conclusions and Relevance In patients with septic shock, treatment with intravenous vitamin C, hydrocortisone, and thiamine, compared with intravenous hydrocortisone alone, did not significantly improve the duration of time alive and free of vasopressor administration over 7 days. The finding suggests that treatment with intravenous vitamin C, hydrocortisone, and thiamine does not lead to a more rapid resolution of septic shock compared with intravenous hydrocortisone alone. Trial Registration ClinicalTrials.gov Identifier:NCT03333278

The impact of the COVID-19 pandemic on eating disorder risk and symptoms.

Abstract: The current COVID-19 pandemic has created a global context likely to increase eating disorder (ED) risk and symptoms, decrease factors that protect against EDs, and exacerbate barriers to care. Three pathways exist by which this pandemic may exacerbate ED risk. One, the disruptions to daily routines and constraints to outdoor activities may increase weight and shape concerns, and negatively impact eating, exercise, and sleeping patterns, which may in turn increase ED risk and symptoms. Relatedly, the pandemic and accompanying social restrictions may deprive individuals of social support and adaptive coping strategies, thereby potentially elevating ED risk and symptoms by removing protective factors. Two, increased exposure to ED-specific or anxiety-provoking media, as well as increased reliance on video conferencing, may increase ED risk and symptoms. Three, fears of contagion may increase ED symptoms specifically related to health concerns, or by the pursuit of restrictive diets focused on increasing immunity. In addition, elevated rates of stress and negative affect due to the pandemic and social isolation may also contribute to increasing risk. Evaluating and assessing these factors are key to better understanding the impact of the pandemic on ED risk and recovery and to inform resource dissemination and targets.

Ultra-Processed Foods and Health Outcomes: A Narrative Review

Abstract: The nutrition literature and authoritative reports increasingly recognise the concept of ultra-processed foods (UPF), as a descriptor of unhealthy diets. UPFs are now prevalent in diets worldwide. This review aims to identify and appraise the studies on healthy participants that investigated associations between levels of UPF consumption and health outcomes. This involved a systematic search for extant literature; integration and interpretation of findings from diverse study types, populations, health outcomes and dietary assessments; and quality appraisal. Of 43 studies reviewed, 37 found dietary UPF exposure associated with at least one adverse health outcome. Among adults, these included overweight, obesity and cardio-metabolic risks; cancer, type-2 diabetes and cardiovascular diseases; irritable bowel syndrome, depression and frailty conditions; and all-cause mortality. Among children and adolescents, these included cardio-metabolic risks and asthma. No study reported an association between UPF and beneficial health outcomes. Most findings were derived from observational studies and evidence of plausible biological mechanisms to increase confidence in the veracity of these observed associations is steadily evolving. There is now a considerable body of evidence supporting the use of UPFs as a scientific concept to assess the ‘healthiness’ of foods within the context of dietary patterns and to help inform the development of dietary guidelines and nutrition policy actions.

Oil price news and COVID-19—Is there any connection?

Abstract: We evaluate the relative importance of COVID-19 infections and oil price news in influencing oil prices. We show that when the number of new COVID-19 infections surpasses 84,479, COVID-19 exerts a bigger effect on oil prices. Oil price news when conditioned on COVID-19 cases have limited effects on price. When oil price volatility is used as a threshold, at higher levels of volatility, both COVID-19 cases and negative oil price news influence oil prices.

Reflection Backdoor: A Natural Backdoor Attack on Deep Neural Networks

Abstract: Recent studies have shown that DNNs can be compromised by backdoor attacks crafted at training time. A backdoor attack installs a backdoor into the victim model by injecting a backdoor pattern into a small proportion of the training data. At test time, the victim model behaves normally on clean test data, yet consistently predicts a specific (likely incorrect) target class whenever the backdoor pattern is present in a test example. While existing backdoor attacks are effective, they are not stealthy. The modifications made on training data or labels are often suspicious and can be easily detected by simple data filtering or human inspection. In this paper, we present a new type of backdoor attack inspired by an important natural phenomenon: reflection. Using mathematical modeling of physical reflection models, we propose reflection backdoor (Refool) to plant reflections as backdoor into a victim model. We demonstrate on 3 computer vision tasks and 5 datasets that, Refoolcan attack state-of-the-art DNNs with high success rate, and is resistant to state-of-the-art backdoor defenses.

Psychometric Validation of the Bangla Fear of COVID-19 Scale: Confirmatory Factor Analysis and Rasch Analysis.

Abstract: The recently developed Fear of COVID-19 Scale (FCV-19S) is a seven-item uni-dimensional scale that assesses the severity of fears of COVID-19. Given the rapid increase of COVID-19 cases in Bangladesh, we aimed to translate and validate the FCV-19S in Bangla. The forward-backward translation method was used to translate the English version of the questionnaire into Bangla. The reliability and validity properties of the Bangla FCV-19S were rigorously psychometrically evaluated (utilizing both confirmatory factor analysis and Rasch analysis) in relation to socio-demographic variables, national lockdown variables, and response to the Bangla Health Patient Questionnaire. The sample comprised 8550 Bangladeshi participants. The Cronbach α value for the Bangla FCV-19S was 0.871 indicating very good internal reliability. The results of the confirmatory factor analysis showed that the uni-dimensional factor structure of the FCV-19S fitted well with the data. The FCV-19S was significantly correlated with the nine-item Bangla Patient Health Questionnaire (PHQ-90) (r = 0.406, p < 0.001). FCV-19S scores were significantly associated with higher worries concerning lockdown. Measurement invariance of the FCV-19S showed no differences with respect to age or gender. The Bangla version of FCV-19S is a valid and reliable tool with robust psychometric properties which will be useful for researchers carrying out studies among the Bangla speaking population in assessing the psychological impact of fear from COVID-19 infection during this pandemic.

Federated Learning With Blockchain for Autonomous Vehicles: Analysis and Design Challenges

Abstract: We propose an autonomous blockchain-based federated learning (BFL) design for privacy-aware and efficient vehicular communication networking, where local on-vehicle machine learning (oVML) model updates are exchanged and verified in a distributed fashion. BFL enables oVML without any centralized training data or coordination by utilizing the consensus mechanism of the blockchain. Relying on a renewal reward approach, we develop a mathematical framework that features the controllable network and BFL parameters (e.g., the retransmission limit, block size, block arrival rate, and the frame sizes) so as to capture their impact on the system-level performance. More importantly, our rigorous analysis of oVML system dynamics quantifies the end-to-end delay with BFL, which provides important insights into deriving optimal block arrival rate by considering communication and consensus delays. We present a variety of numerical and simulation results highlighting various non-trivial findings and insights for adaptive BFL design. In particular, based on analytical results, we minimize the system delay by exploiting the channel dynamics and demonstrate that the proposed idea of tuning the block arrival rate is provably online and capable of driving the system dynamics to the desired operating point. It also identifies the improved dependency on other blockchain parameters for a given set of channel conditions, retransmission limits, and frame sizes. 1 However, a number of challenges (gaps in knowledge) need to be resolved in order to realise these changes. In particular, we identify key bottleneck challenges requiring further investigations, and provide potential future reserach directions. 1 An early version of this work has been accepted for presentation in IEEE WCNC Wksps 2020 [1] .

Predicting psychological and subjective well-being from personality: A meta-analysis.

Abstract: This study reports the most comprehensive assessment to date of the relations that the domains and facets of Big Five and HEXACO personality have with self-reported subjective well-being (SWB: life satisfaction, positive affect, and negative affect) and psychological well-being (PWB: positive relations, autonomy, environmental mastery, purpose in life, self-acceptance, and personal growth). It presents a meta-analysis (n = 334,567, k = 462) of the correlations of Big Five and HEXACO personality domains with the dimensions of SWB and PWB. It provides the first meta-analysis of personality and well-being to examine (a) HEXACO personality, (b) PWB dimensions, and (c) a broad range of established Big Five measures. It also provides the first robust synthesis of facet-level correlations and incremental prediction by facets over domains in relation to SWB and PWB using 4 large data sets comprising data from prominent, long-form hierarchical personality frameworks: NEO PI-R (n = 1,673), IPIP-NEO (n = 903), HEXACO PI-R (n = 465), and Big Five Aspect Scales (n = 706). Meta-analytic results highlighted the importance of Big Five neuroticism, extraversion, and conscientiousness. The pattern of correlations between Big Five personality and SWB was similar across personality measures (e.g., BFI, NEO, IPIP, BFAS, Adjectives). In the HEXACO model, extraversion was the strongest well-being correlate. Facet-level analyses provided a richer description of the relationship between personality and well-being, and clarified differences between the two trait frameworks. Prediction by facets was typically around 20% better than domains, and this incremental prediction was larger for some well-being dimensions than others. (PsycINFO Database Record (c) 2020 APA, all rights reserved).

A Game-Theoretical Approach for User Allocation in Edge Computing Environment

Abstract: Edge Computing provides mobile and Internet-of-Things (IoT) app vendors with a new distributed computing paradigm which allows an app vendor to deploy its app at hired edge servers distributed near app users at the edge of the cloud. This way, app users can be allocated to hired edge servers nearby to minimize network latency and energy consumption. A cost-effective edge user allocation (EUA) requires maximum app users to be served with minimum overall system cost. Finding a centralized optimal solution to this EUA problem is NP-hard. Thus, we propose EUAGame, a game-theoretic approach that formulates the EUA problem as a potential game. We analyze the game and show that it admits a Nash equilibrium. Then, we design a novel decentralized algorithm for finding a Nash equilibrium in the game as a solution to the EUA problem. The performance of this algorithm is theoretically analyzed and experimentally evaluated. The results show that the EUA problem can be solved effectively and efficiently.

Hourly Oil Price Volatility: The Role of COVID-19

Abstract: In this paper, we study the evolution of hourly oil price volatility. Using multiple measures of oil price volatility, we conclude that volatility increased following the onset of COVID-19. After controlling for conventional predictors of oil price volatility, we show that COVID-19 cases and deaths led to an increase in daily oil price volatility by between 8% and 22%. Our results pass a battery of robustness tests.

TON_IoT Telemetry Dataset: A New Generation Dataset of IoT and IIoT for Data-Driven Intrusion Detection Systems

Abstract: Although the Internet of Things (IoT) can increase efficiency and productivity through intelligent and remote management, it also increases the risk of cyber-attacks. The potential threats to IoT applications and the need to reduce risk have recently become an interesting research topic. It is crucial that effective Intrusion Detection Systems (IDSs) tailored to IoT applications be developed. Such IDSs require an updated and representative IoT dataset for training and evaluation. However, there is a lack of benchmark IoT and IIoT datasets for assessing IDSs-enabled IoT systems. This paper addresses this issue and proposes a new data-driven IoT/IIoT dataset with the ground truth that incorporates a label feature indicating normal and attack classes, as well as a type feature indicating the sub-classes of attacks targeting IoT/IIoT applications for multi-classification problems. The proposed dataset, which is named TON_IoT, includes Telemetry data of IoT/IIoT services, as well as Operating Systems logs and Network traffic of IoT network, collected from a realistic representation of a medium-scale network at the Cyber Range and IoT Labs at the UNSW Canberra (Australia). This paper also describes the proposed dataset of the Telemetry data of IoT/IIoT services and their characteristics. TON_IoT has various advantages that are currently lacking in the state-of-the-art datasets: i) it has various normal and attack events for different IoT/IIoT services, and ii) it includes heterogeneous data sources. We evaluated the performance of several popular Machine Learning (ML) methods and a Deep Learning model in both binary and multi-class classification problems for intrusion detection purposes using the proposed Telemetry dataset.

Innovation can accelerate the transition towards a sustainable food system

Abstract: Future technologies and systemic innovation are critical for the profound transformation the food system needs. These innovations range from food production, land use and emissions, all the way to improved diets and waste management. Here, we identify these technologies, assess their readiness and propose eight action points that could accelerate the transition towards a more sustainable food system. We argue that the speed of innovation could be significantly increased with the appropriate incentives, regulations and social licence. These, in turn, require constructive stakeholder dialogue and clear transition pathways.

SARS-CoV-2 genomic and subgenomic RNAs in diagnostic samples are not an indicator of active replication.

Abstract: Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) was first detected in late December 2019 and has spread worldwide. Coronaviruses are enveloped, positive sense, single-stranded RNA viruses and employ a complicated pattern of virus genome length RNA replication as well as transcription of genome length and leader containing subgenomic RNAs. Although not fully understood, both replication and transcription are thought to take place in so-called double-membrane vesicles in the cytoplasm of infected cells. Here we show detection of SARS-CoV-2 subgenomic RNAs in diagnostic samples up to 17 days after initial detection of infection and provide evidence for their nuclease resistance and protection by cellular membranes suggesting that detection of subgenomic RNAs in such samples may not be a suitable indicator of active coronavirus replication/infection.