Search

Machine Learning is a relatively new and emergent field in video game development domain. Despite considerable relevance to the video game industry, there has yet to be a significant commercial product utilising machine learning in its design or function. Although previous research has shown significant potential in the use of video games in developing and testing Artificial Intelligence, the reverse i.e., using Artificial Intelligence to develop and test video games is far less common. This paper provides a survey of existing techniques and reviews current and future applications of machine learning in the field of video game development, both as a tool to streamline development and management processes and as an integrated part of video game end products themselves. This paper also explores a number of machine learning technologies not yet applied to use in the video game field and discusses their potential in future research and product development. Despite the relative newness and lack of development in this field, this paper finds that there is potential for machine learning to significantly improve and expedite production in the video game industry. Machine Learning can potentially be harnessed to develop new or improved products or automate development processes in conventional video game products.

The Cooper Basin of Australia is challenged by high deviatory stress conditions in the strike-slip to reverse stress regimes, which increase borehole breakout during drilling. This paper studies the applicability of failure criteria such as Mohr-Coulomb, Mogi-Coulomb, and Drucker-Prager for wellbore collapse in a low permeability sandstone well in the Cooper Basin, Australia. The influences of well conditions and rock mechanical properties on the wellbore collapse modeling are also investigated. The 1D mechanical earth model (MEM) from well Tamarama-1 is compiled and extracted to provide four sections of data with borehole breakouts. First, the stresses around the borehole are calculated using the Kirsch equations. Then, different failure criteria (i.e., Mohr-Coulomb, Mogi-Coulomb, and Drucker-Prager) are used to evaluate the areas of borehole breakout. Finally, a parametric analysis is presented to show the influences of well conditions and rock mechanical properties on the prediction of borehole breakouts. The results show that the Mogi-Coulomb criterion and Drucker-Prager criterion have a good match with borehole breakouts from field observations. The Mohr-Coulomb criterion overestimates the wellbore collapse pressure and breakout areas because it ignores the strengthening effect of the intermediate principal stress. The parametric study shows that the maximum horizontal stress, azimuth, inclination, and friction angle can affect borehole breakouts in terms of magnitude and orientation. Rock mechanical properties such as Young's Modulus do not affect the borehole breakout, and Poisson's ratio has a minor effect on the borehole breakout. A better understanding of borehole stability is crucial and beneficial to drilling and completions of oil and gas wells. This paper provides an integrated workflow to determine which failure criterion is most applicable for given conditions and provides a framework for a more detailed sensitivity analysis on borehole breakout. It

The area of human performance improvement has become a greater focus in recent military contexts as evidenced by Australian Defence Force projects. The aim of the design proposed in this paper is to develop a Performance Management System using Machine Learning (PMSML) to enhance the physical human performance of individual warfighters in combat situations through 1) early recognition and self-management of acute health events in the field; 2) forecasting of soldier (user) failure; and 3) proactive self-management of longer-term health outcomes during prolonged manoeuvres or combat situations. This paper proposes a high-level design and approach using machine learning algorithms to assess the feasibility of improving metrics such as health data accuracy and efficiency when transmitting data from sensors to the cloud in military networks. The significance of this design is to predict health conditions of users on a personalised basis for an individual's physical and mental health performance without compromising performance metrics using machine learning algorithms. Results show that machine learning algorithms outperformed other existing methods, which must compromise between certain metrics.

Digital twin technology today is diverse and emerging and its full potential is not yet widely understood. The concept of a digital twin allows for the analysis, design, optimisation and evolution of systems to take place fully digital, or in conjunction with a cyber-physical system to improve speed, accuracy and efficiency when compared to traditional engineering approaches. Digital twins continue to be a technology that enables new paradigms, such as Industry 4.0 and Factories of the Future as well as generating improved efficiencies within existing systems. The development of digital twin technology in traditional industries such as manufacturing, construction, the automotive industry, agriculture and transportation has highlighted its potential, but often insufficiently explored the risks associated with their integration. In this paper we explore risks relating to the cyber-security of systems employing digital twin technology and also consider the opportunities for digital twins themselves to mitigate cyber-security risks and become an integral part of a security in-depth defence.

This paper presents a system for assessing men's physique, using advanced computer vision and deep learning methods. The pipeline involves the segmenting and gauging of the condition of various muscle groups. The proposed system is composed of two different deep learning models working in tandem, the first is an Instance Segmentation Mask R-CNN and the second is GoogLeNet. The Mask R-CNN is used for the accurate multi-class identification and detection of different muscle groups, such as chest, biceps, abs, and shoulders. GoogLeNet then further classifies each muscle group into various levels (level 1, level 2, up to level n). Furthermore, performance metrics, such as accuracy, precision, F1-measure and confusion matrices are used to evaluate the effectiveness of the proposed system. We believe, that as this system provides information regarding the physical shape/ form of a person, it can be used to augment Diet and Exercise Recommendation Systems (DERS) and can have many commercial as well as clinical applications.

Aim This review assessed the effectiveness of interventions using a goal-setting approach on glycaemic control for people diagnosed with prediabetes or type 2 diabetes. Design A systematic review guided by the Joanna Briggs Institute methodology for conducting systematic reviews of primary research studies was conducted. Data Sources Randomized controlled trials and experimental studies with a minimum follow-up period of 6 months were considered for inclusion. The primary outcome was change in glycaemic control as measured by glycated haemoglobin (%) and/or fasting plasma glucose (mg/dl). A systematic search of seven electronic databases was completed in October 2020 Review Methods Papers meeting the inclusion criteria were critically appraised using the Joanna Briggs Institute tools for critical appraisal followed by data extraction. A Grading of Recommendations Assessment, Development and Evaluation assessment was conducted to assess the overall certainty of the evidence. Fixed-effect meta-analyses were completed to demonstrate the mean effect for each outcome of interest. Results Twenty one studies were included in this review. Goal setting was more effective than usual care for glycaemic control in prediabetes at 6 months and at 12 months for fasting plasma glucose (mg/dl) and glycated haemoglobin (%). Goal setting was more effective than usual care for glycaemic control in type 2 diabetes for fasting plasma glucose (mg/dl) at 6 months, fasting plasma glucose (mg/dl) at 12 months, glycated haemoglobin (%) at 6 months and glycated haemoglobin (%) at 12 months. Conclusion The evidence suggests goal setting is effective in supporting people to achieve glycaemic targets in prediabetes and type 2 diabetes.

This paper measures and inclusively analyses the extent of Australian public universities’ sustainability disclosures on three key reporting media, that is annual reports, stand-alone sustainability reports, and websites. The study also compares the extent of the disclosures across university groupings. Content analysis was used to analyse the data by utilising the updated Graphical Assessment of Sustainability in Universities (GASU) indicators together with a weighted scoring scheme. Mann–Whitney and Kruskal–Wallis tests were applied in the comparison tests. The results of this study indicate that the extent of sustainability disclosures made by Australian public universities is overall fairly low, and the evaluation and comparison of the extent of sustainability disclosures among different university groupings overall show no significant difference. Overall, the findings suggest that there is a need to increase both the level of awareness and recognition of benefits resulted from performing sustainable development activities and the way these performances are better communicated to stakeholders.

In recent years, it has been demonstrated that deep learning has great success in a variety of computer vision applications. Deep learning-based Faster R-CNN algorithm depends on region proposal network that provides state-of-the-art object detection performance. To date, a limited number of Faster R-CNN approaches have been attempted to detect seagrass from underwater digital images. This paper proposes an improved seagrass detector that enhances the detection performance by combining the Faster R-CNN framework with the NASNet-A backbone. This seagrass detector achieves a high mean average precision (mAP) of 0.412 on ECUHO-2 dataset, which is significantly better than state-of-the-art Halophila ovalis detection performance on this dataset.

This paper introduces a semi-supervised two-stage framework that leverages a large collection of unlabelled seagrass data with the guidance of a small amount of labelled seagrass data. First, a state-of-the-art Convolutional Neural Network (CNN) named EfficientNet is trained on a small number of labelled seagrass image patches. Next, the trained classifier generates pseudo labels for the unlabelled seagrass image patches. Finally, this classifier is retrained on the combination of labelled and pseudo labelled seagrass image patches. This approach achieves 98.0% multi-species detection accuracy on a newly developed seagrass dataset. Our approach on the publicly available 'DeepSeagrass' dataset achieves an overall accuracy of 91.3% and 93.0%, which outperforms the state-of-the-art multi-species seagrass detection accuracy of 88.2% and 92.4% for four and five classes, respectively.

Public services face increasing and more varied demands, but public service managers often lack the resources to develop their staff to effectively deal with them. This paper identifies how public sector line managers can support their subordinates to not just cope with challenges, but also to grow and develop in the midst of them. It primarily draws on employee attributions of leader effectiveness, to identify growth-oriented leadership behaviours. The framework is based on empirical data from two sequential phases of interviews, focus groups, and one-on-one discussions with employees and managers in the public sector. Five categories of growth-oriented leadership behaviours were identified: managing the whole team, enabling self-management, managing safe failures, supporting both career and personal growth, and recognising individual needs and contributions. Causation coding identified a set of processes through which these leadership behaviours supported staff to use networks, learn, and adapt. The developed framework can be used to enhance growth and development of public sector employees.

This study examines policy implementation as it unfolds from policy, to the contractors and then to its impact on employees delivering the social service. Multilevel statistical analysis is used to examine survey data from 36 contractors and 542 employees. The use of multilevel analysis provides a methodological tool and subsequent analysis new to the discipline of public administration. The results indicate a strong positive association between resource (in)adequacy decision-making at the contractor level and employee outcomes. The contribution of the paper is new evidence to support the widening of existing quality care indicators to include organisational support metrics for contractors.

This paper aims to study and evaluate the performance of a novel control scheme for Dynamic Voltage Restores (DVR) with Transformer-based Impedance Sourse Inverter (Trans-ZSI) configuration. An Improved Maximum Constant Boost Control (IMCBC) method is proposed to enhance the overall performance of solar PV fed TransZSI-DVRs. The proposed technique contributes to THD reduction and voltage stress mitigation across the switches. A model developed in Matlab/SIMULINK is tested under different types of voltage disturbance. The results show the effectiveness of the proposed control method as a corrective approach in compensating the voltage at the point of common coupling.

Cardiovascular Disease (CVD) is the leading cause of death worldwide. Calcification in the Abdominal Aorta is a stable marker of CVD development and, hence, it's early detection is considered crucial to saving lives. Imaging techniques such as Computed Tomography (CT) and Digital X-Ray Imaging can be used to accurately predict and localize Abdominal Aortic Calcification (AAC), however, these methods are not only expensive but also expose the patients to high ionizing radiation. In contrast, Dual Energy X-ray Absorptiometry (DXA) is an efficient, cost-effective and low radiation exposure-based imaging alternative, but with challenges like low resolution and vague vertebral boundaries. This poses a bottleneck in identifying the vertebrae and their boundaries which is crucial in manual as well as automatic scoring of AAC from DXA scans. In this paper, we address this research gap by proposing a framework which first localizes the vertebrae T12, L1, L2, L3, L4 and L5 and then generates Inter-Vertebral Guides (IVGs) between them. Our deep model is trained on lateral view DXA spine images and shows promising results in generating IVGs with high accuracy, which we believe can greatly reduce inter-observer variability in AAC scoring in DXA imaging domain.

Patients are often encouraged to make use of wearable devices for remote collection and monitoring of health data. This adoption of wearables results in a significant increase in the volume of data collected and transmitted. The battery life of the devices is then quickly diminished due to the high processing requirements of the device. Given the importance attached to medical data, it is imperative that all transmitted data adhere to strict integrity and availability requirements. Reducing the volume of healthcare data for network transmission may improve sensor battery life without compromising accuracy. There is a trade-off between efficiency and accuracy which can be controlled by adjusting the sampling and transmission rates. This paper demonstrates that machine learning can be used to analyse complex health data metrics such as accuracy and efficiency of data transmission to overcome the trade-off problem. The study uses time series nonlinear autoregressive neural network algorithms to enhance both metrics by taking fewer samples to transmit. The algorithms were tested with a standard heart rate dataset to compare their accuracy and efficiency. The result showed that the Levenberg-Marquardt algorithm was the best performer with an efficiency of 3.33 and accuracy of 79.17%, which is similar to other algorithms’ accuracies but demonstrates improved efficiency. This proves that machine learning can improve without sacrificing a metric over the other compared to the existing methods with high efficiency.

This paper investigates how family events impact founding entrepreneurs’ psychological affect during the new venture creation process and whether their ventures commenced before such events continue to be viable. We also explore the impact of the interaction of family events and overconfidence on new venture survival. We use panel data drawn from the Australian Household, Income and Labor Dynamics (HILDA) survey, focusing on non-venture related, family event-induced psychological affect that entrepreneurs may experience as a predictor of their new venture survival. Our accelerated failure time model results are consistent with past arguments in the social psychology literature, that ceteris parabis, positive events have a positive influence and negative events have a negative influence on survival. However, when entrepreneurial overconfidence is considered, although the interaction between negative family events and entrepreneur overconfidence spurs cautious behaviour to a small degree as anticipated, it is positive events interacting with overconfidence that makes the biggest impact (negative) on the new venture. The research contributes to our understanding of the embeddedness of family in the entrepreneurial process – specifically how major family events, such as the birth and death of a child and marriage and divorce, which seemingly have little or nothing to do with the new venture, can impact its survival. New founders experiencing a major family event(s) can be psychologically affected through their feelings, emotions and moods. The viability of the startups they helm depend on their ability to manage these stressors. It challenges past assumptions by revealing how positive family events can have comparatively greater negative impact on new venture survival than negative ones.

Malware detection is an important area of cyber security. Computer systems rely on malware detection applications to prevent malware attacks from succeeding. Malware detection is not a straightforward task, as new variants of malware are generated at an increasing rate. Machine learning (ML) has been utilised to generate predictive classification models to identify new malware variants which conventional malware detection methods may not detect. Machine learning, has however, been found to be vulnerable to different types of adversarial attacks, in which an attacker is able to negatively affect the classification ability of the ML model. Several defensive measures to prevent adversarial poisoning attacks have been developed, but they often rely on the use of a trusted clean dataset to help identify and remove adversarial examples from the training dataset. The defence in this paper does not require a trusted clean dataset, but instead, identifies intentional false negatives (zero day malware classified as benign) at the testing stage by examining the activation weights of the ML model. The defence was able to identify 94.07% of the successful targeted poisoning attacks.

An intrusion detection system's (IDS) key role is to recognise anomalous activities from both inside and outside the network system. In literature, many machine learning techniques have been proposed to improve the performance of IDS. To create a good IDS, a single classifier might not be powerful enough. To overcome this bottleneck researchers focus on hybrid/ensemble techniques. Such methods are more complex and computation intensive, but they provide greater accuracy and lower false alarm rates (FAR). In this paper, we propose a bagging ensemble that improves the performance of IDS in terms of accuracy and FAR where the NSL-KDD dataset has been used to classify benign and abnormal traffic. We have also applied the information gain-based feature selection method to select highly relevant features for improving the accuracy of the proposed technique and achieved 84.93 % accuracy and 2.45 % FAR on the test dataset.

This paper explores activism enacted through Silk Road, a now defunct cryptomarket where illicit drugs were sold in the dark web. Drawing on a digital ethnography of Silk Road, we develop the notion of constructive activism to extend the lexicon of concepts available to discuss forms of online activism. Monitoring of the cryptomarket took place between June 2011 and its closure in October 2013. Just before and after the closure of the marketplace we conducted anonymous online interviews with 17 people who reported buying drugs on Silk Road (1.0). These interviews were conducted synchronously and interactively through encrypted instant messaging. Participants discussed harnessing and developing the technological tools needed to access Silk Road and engage within the Silk Road community. For participants Silk Road was not just a market for trading drugs: it facilitated a shared experience of personal freedom within a libertarian philosophical framework, where open discussions about stigmatized behaviours were encouraged and supported. Tensions between public activism against drug prohibition and the need to hide one's identity as a drug user from public scrutiny were partially resolved through community actions that internalized these politics, rather than engaging in forms of online activism that are intended to have real-world political effects. Most aptly described through van de Sande's (2015) concept of prefigurative politics, they sought to transform their values into built environments that were designed to socially engineer a more permissive digital reality, which we refer to as constructive activism.