Sofia Chaudry, Murdoch University
Olanike Adeoye, Victoria Institute of Strategic Economic Studies
Co-author: Professor Sardar Islam, Victoria Institute of Strategic Economic Studies
There still exists a need for compelling analysis and solutions to the identification of an optimal capital structure in the presence of information asymmetry. The study deploys the differential game theory framework as a dynamic technique which analyses the conflicts of interests between the shareholder and manager. This is done with respect to the company’s optimal capital structure. The model was designed as an incentive contract between the shareholder and the manager to mitigate the agency problem of unobservable effort and cash diversion while maximizing the value of the firm which then leads to a good corporate governance system. To simulate the model, secondary data were obtained from the financial statements of Asaleo Care Limited, a consumer staple sector company in Australia. Preliminary results of the model in a Nash open-loop solution concept for Asaleo Care Limited depicts that in the absence of the incentive contract designed, market value of the company declines while the investment of the company and its earnings becomes negative over time, thereby emphasizing the significance of such robust incentive contract to manage the cash diversion and unobservable efforts problems to attain optimal mix of finance.
Dr Philipp Braun, University of Newcastle
Co-author: Professor Christopher M. Kellett, Professor Steven R. Weller, University of Newcastle
For his work in the economics of climate change, Professor William Nordhaus was a co-recipient of the 2018 Nobel Memorial Prize for Economic Sciences. A core component of the work undertaken by Nordhaus is the Dynamic Integrated model of Climate and Economy, known as the DICE model. The DICE model is a discrete-time model which is primarily used in conjunction with a particular optimal control problem in order to estimate optimal pathways for reducing greenhouse gas emissions. In this Poster we give an introduction to the DICE model from a systems and control perspective. In addition, we indicate challenges and open problems of potential interest.
Rixon Crane, The University of Queensland
Co-author: Dr Fred Roosta, The University of Queensland
Continuous technological and communication advancements have enabled the collection of and access to ever-growing large scale datasets. There is a significant amount of research and development being devoted to machine learning problems in general, and the underlying optimization algorithms in particular, that are formulated on these large scale datasets. However, lack of adequate computational resources, in particular storage, can severely limit, or even prevent, any attempts at solving such optimization problems in a traditional stand-alone way, e.g., using a single machine. This can be remedied through distributed computing, in which resources across a network of stand-alone computational nodes are “pooled” together to scale to the problem at hand. Optimization algorithms designed for distributed environments can efficiently leverage the computational resources of a network of machines. A significant issue with this framework of computing is that inter-machine communication can be expensive. Therefore, in such environments it is necessary to design algorithms that use a low number of communication rounds. There is significant potential for communication-efficient distributed second-order optimization methods. In my research we developed such methods that remedy many of the shortcomings of the existing methods.
Yang Liu, The University of Queensland
To alleviate several shortcomings of the classical Newton’s method and its (inexact) Newton-CG variant, while preserving their various desirable theoretical and algorithmic properties, Newton-MR has recently been introduced, which extends Newton-CG in the same manner that MINRES extends CG.
Recently, stability of Newton-CG under Hessian perturbations, i.e., inexact curvature information, have been extensively studied. Such stability analysis has been leveraged in designing variants of Newton-CG, in which as a way to reduce the computational costs involving the Hessian matrix, the curvature is suitably approximated. Here, we do that for Newton-MR. Unlike the stability analysis of Newton-CG, which relies on spectrum preserving perturbations in the sense of Lowner partial order, our work here draws from matrix perturbation theory to estimate the distance between the underlying exact and perturbed sub-spaces. Numerical experiments demonstrate great degree of stability for Newton-MR, amounting to a highly efficient algorithm in large-scale problems.
Kobamelo Mashaba, Curtin University
Co-author: Dr Xu Honglei, Curtin University
The existence of critical jobs in manufacturing production line has resulted in a need of new unprecedented control tool over the hybrid manufacturing system. The aim of optimal control is primarily to gain insight of production system faster, and aid manufactures to make informed decision, on operations and implementation of lean tools, and other optimization methods. We achieve this by developing a hybrid model and a new smoothing algorithm that tackles the cost balancing between a job quality and job tardiness by finding optimal service time of the system
Ramin Rakhsha, The University of Western Australia
Mining projects are capital intensive and inherently carry uncertainties and risks with them.
General Mine Planning techniques and software often use a static and linear (constant commodity price, discount factor, and a static risk profile) approach in regards with the risks and uncertainties for the mine plans. Almost all of the current General Mine Planning software use the Discounted Cash Flow (DCF) valuation method to maximise (optimise) the Net Present Value (NPV) as the objective for the Mine Production Plan.
This project is centred on quantification of risks and uncertainties within the mine planning cycle and integration of mine planning optimisation through non-linear means. The aim here is to use System Dynamics (SD) -an engineering/mathematical simulation method based on feedback theory- to generate a mine optimisation technique considering uncertainties and risks in the mine plan using Real Option valuation (ROV).
The project has initially tested SD applications to realise the impact of price uncertainties with a series of NPV driven mine production plans generated with a General Mine Planning (GMP) software, and will subsequently be developing ROV within SD environment for comparison reasons against the NPV and to understand the practicality of ROV and SD for Mine Planning purposes.
Zoe Renwick, RMIT University
As a renewable energy source, solar radiation is attracting the attention of researchers. Estimation of the amount of solar radiation reaching the ground in predefined time intervals is required for all kinds of planning activities around solar irradiance. For example, to synchronize solar electricity generators with microgrids, we need to estimate the amount of solar irradiance in very short time intervals. Solar irradiance (the power, per square meter, received from the sun) is measured by sensors called pyranometers. Compared with other meteorological devices, pyranometers are exceptionally prone to errors. Due to the errors inherent in the measurement of solar irradiance, quality control checks are necessary to ensure we have reliable and accurate data. Currently, the most commonly used quality control checks applied to solar irradiance data are range tests and clear-sky models. These methods have a number of limitations which can be addressed by the use of Bayes’ theorem. Bayesian Control charts can provide more accurate estimates by continually updating control limits as information is gathered. Preliminary results of our research have shown that the Bayesian method does have a higher rate of accuracy in detecting erroneous data points above currently used methods.
Ekta Sharma, University of Southern Queensland
Co-author: Dr Ravinesh Deo, Professor Alfio Parisi, University of Southern Queensland
Dr Ramendra Prasad, University of Fiji
Development of practical and efficient air quality regulating mechanisms is indeed a challenge. A pertinent air pollutant responsible for recurrent health-care costs with increased respiratory induced mortality in Australia is Particulate Matter (PM). In such a pressing public health problem, artificial intelligence can provide promising solution via developing models to obtain predictions through novel learning algorithms. This research project, unique in its own kind, focuses on generating real-time air quality forecasts of PM-2.5, PM-10 and the overall lower atmospheric visibility. To establish a robust optimisation model, an online sequential-extreme learning machine (OS-ELM), a powerful artificial intelligence algorithm was integrated with improved empirical mode decomposition with adaptive noise (ICEEMDAN) as a data pre-processing tool. The resulting novel data-driven, hybrid predictive model: ICEEMDAN-OS-ELM registers good predictive ability, particularly appropriate for near real-time forecasts of PM-2.5, PM-10 and visibility at key test sites in Australia. The excellent performance of the OS-ELM hybrid model indicates its suitability as a decision-support systems tool in air quality monitoring, forecasting and subsequent health risk mitigation. Future study aspires to develop robust predictive frameworks in facilitating air quality strategies in mitigating Australia’s public health risks in a coordinated way through advanced practical development of artificial intelligence approaches.
Herbert Taco Arana, Curtin University
Co-author: Professor Louis Caccetta, Curtin University
The use of optimisation and simulation models by ore producers, to support critical strategic supply chain management decisions is becoming standard industry practice. The performance of such models is crucial for meeting global demand, whilst maximising the profitability of a supply chain across its lifespan. However, their performance also often hinges on inputs which require varying levels of user intervention. This increases the risk of introducing errors which can significantly impact the performance of a model. Typical examples of this may be observed in bulk material export shipping operations where, for certain supply chain models, the inputs required consists of ship stem data. To improve the predictive value of these models, users often intervene to modify the data to reflect that which is anticipated in the future. In this research, we explore the use of machine-learning to assist in the generation of semi-synthetic data. We present a hybrid optimisation-neural-network based decision support tool that builds on current methods employed by bulk material exporters and mitigates the risk to model-mediated decisions associated with user intervention.