Advanced Machine Learning with Bilevel Optimization. There is an urgent need to develop a new machine learning (ML) paradigm that can overcome data-privacy and model-size constraints in real-world applications. This project aims to develop an advanced paradigm of ML with bilevel optimisation, called bilevel ML. A theoretically-guaranteed fast approximate solver and a new fuzzy bilevel learning framework will be developed to achieve the aim in complex situations; a methodology to transfer knowled ....Advanced Machine Learning with Bilevel Optimization. There is an urgent need to develop a new machine learning (ML) paradigm that can overcome data-privacy and model-size constraints in real-world applications. This project aims to develop an advanced paradigm of ML with bilevel optimisation, called bilevel ML. A theoretically-guaranteed fast approximate solver and a new fuzzy bilevel learning framework will be developed to achieve the aim in complex situations; a methodology to transfer knowledge and an approach to fast-adapt bilevel optimization solutions when required computing resources change. The anticipated outcomes should significantly improve the reliability of ML with benefits for safety learning and computing resource optimisation in ML-based data analytics.Read moreRead less
Discovery Early Career Researcher Award - Grant ID: DE240101089
Funder
Australian Research Council
Funding Amount
$436,847.00
Summary
Trustworthy Hypothesis Transfer Learning. It is urgent to develop a new hypothesis transfer learning scheme that can overcome potential risks when finetuning unreliable large-scale pre-trained models. This project aims to develop an advanced and reliable scheme of hypothesis transfer learning, called Trustworthy Hypothesis Transfer Learning (TrustHTL). A new theoretically guaranteed heterogeneous hypothesis transfer learning framework will be developed to handle heterogeneous situations; a metho ....Trustworthy Hypothesis Transfer Learning. It is urgent to develop a new hypothesis transfer learning scheme that can overcome potential risks when finetuning unreliable large-scale pre-trained models. This project aims to develop an advanced and reliable scheme of hypothesis transfer learning, called Trustworthy Hypothesis Transfer Learning (TrustHTL). A new theoretically guaranteed heterogeneous hypothesis transfer learning framework will be developed to handle heterogeneous situations; a methodology to disinherit risks of pre-trained models and a new fuzzy relation based distributional discrepancy in heterogeneous transfer learning scenarios. The outcomes should significantly improve the reliability of machine learning with benefits for safety learning in data analytics.Read moreRead less
Democratisation of Deep Learning: Neural Architecture Search at Low Cost. The need to manually design Deep Learning-based Neural Networks (DNNs) limits their usage to AI experts and hinders the exploitation of their true potential more broadly, e.g., in farming, humanities. We aim to replace this tedious process through novel AI methods capable of generating DNNs that can perform significantly better and at a lower computational cost than manually designed DNNs. We further expand this idea to so ....Democratisation of Deep Learning: Neural Architecture Search at Low Cost. The need to manually design Deep Learning-based Neural Networks (DNNs) limits their usage to AI experts and hinders the exploitation of their true potential more broadly, e.g., in farming, humanities. We aim to replace this tedious process through novel AI methods capable of generating DNNs that can perform significantly better and at a lower computational cost than manually designed DNNs. We further expand this idea to solve complex real-world problems with both labelled and unlabelled data found in various applications including energy and climate change. The expected outcomes include the novel AI methods, highly trained AI researchers and a number of critical applications that will bring significant benefits to Australia and the world.Read moreRead less
Accelerated Finite-time Learning and Control in Cyber-Physical Systems. Efficient learning and control in cyber-physical systems such as smart grids and robotic systems are very important for achieving economic and social benefits. This project aims to establish a breakthrough accelerated finite-time dynamics theory and technology to assist in delivering efficient learning and control. Expected outcomes include new distributed accelerated finite-time dynamics based learning and control algorithm ....Accelerated Finite-time Learning and Control in Cyber-Physical Systems. Efficient learning and control in cyber-physical systems such as smart grids and robotic systems are very important for achieving economic and social benefits. This project aims to establish a breakthrough accelerated finite-time dynamics theory and technology to assist in delivering efficient learning and control. Expected outcomes include new distributed accelerated finite-time dynamics based learning and control algorithms and tools for optimal operations in cyber-physical systems. This should provide significant benefits including a practical technology for industry applications in smart grids and robotic systems, and training of the next generation engineers in this technology for Australia.Read moreRead less
Variable Structure Complex Network Systems with Smart Grid Applications. This project aims to establish a breakthrough theory and technology to help deliver reliability and security of complex network systems, which are subject to structure changes, against faults and cyberattacks. Expected outcomes include a new theory that lays the foundation for understanding such systems, innovative algorithms and tools for their design, and a practical software platform used for ensuring reliability and sec ....Variable Structure Complex Network Systems with Smart Grid Applications. This project aims to establish a breakthrough theory and technology to help deliver reliability and security of complex network systems, which are subject to structure changes, against faults and cyberattacks. Expected outcomes include a new theory that lays the foundation for understanding such systems, innovative algorithms and tools for their design, and a practical software platform used for ensuring reliability and security of such systems. It will be applied directly to critical infrastructure such as the national power grid to help maintain lifeline resilience and achieve economic benefits. It will also provide an opportunity to train the next generation engineers in this cutting-edge technology for Australia.Read moreRead less
Learning deep resilient behaviour for uncertainty-aware autonomy. This research project aims to propose a novel framework for developing uncertainty-aware autonomous systems using deep learning. There are fundamental gaps in our knowledge of deep uncertainty quantification and its application for risk-aware decision making. Novel algorithms will be proposed to reliably generate deep uncertainty estimates with low computational overhead. These estimates will be then exploited by safety-critical s ....Learning deep resilient behaviour for uncertainty-aware autonomy. This research project aims to propose a novel framework for developing uncertainty-aware autonomous systems using deep learning. There are fundamental gaps in our knowledge of deep uncertainty quantification and its application for risk-aware decision making. Novel algorithms will be proposed to reliably generate deep uncertainty estimates with low computational overhead. These estimates will be then exploited by safety-critical systems such as autonomous robots to identify risky actions and avoid catastrophise. Developed algorithms will be implemented on an autonomous robotic system to make it averse to uncertainties. The outcomes will greatly increase reliable telerobotic applications in mining, manufacturing, defence, and health.Read moreRead less
Collaborative Sensing and Learning for Maritime Situational Awareness. We aim to demonstrate coordinated autonomous sensing of naval assets in dynamic maritime environments, reducing the operational load required to deliver a high quality maritime situational awareness. A realistic simulation based approach will help us develop novel artificial intelligence technology including: self-adaptive strategies for dynamic asset allocation, embedded smart sensing capabilities for naval observation syste ....Collaborative Sensing and Learning for Maritime Situational Awareness. We aim to demonstrate coordinated autonomous sensing of naval assets in dynamic maritime environments, reducing the operational load required to deliver a high quality maritime situational awareness. A realistic simulation based approach will help us develop novel artificial intelligence technology including: self-adaptive strategies for dynamic asset allocation, embedded smart sensing capabilities for naval observation systems and novel approaches to continuous collaborative learning from multi-spectral media. In addition to the emerging partnership between participants, the project will advance sovereign capability to develop maritime intelligence gathering technology for the Royal Australian Navy to underpin stability in our region. Read moreRead less
Discovery Early Career Researcher Award - Grant ID: DE200100063
Funder
Australian Research Council
Funding Amount
$394,398.00
Summary
Nonmonotone Algorithms in Operator Splitting, Optimisation and Data Science. This project aims to develop the mathematical foundations for the analysis and development of optimisation algorithms used in data science. Despite their now ubiquitous use, machine learning software packages routinely rely on a number of algorithms from mathematical optimisation which are not properly understood. By moving beyond the traditional realms of Fejér monotone algorithms, this project expects to develop the m ....Nonmonotone Algorithms in Operator Splitting, Optimisation and Data Science. This project aims to develop the mathematical foundations for the analysis and development of optimisation algorithms used in data science. Despite their now ubiquitous use, machine learning software packages routinely rely on a number of algorithms from mathematical optimisation which are not properly understood. By moving beyond the traditional realms of Fejér monotone algorithms, this project expects to develop the mathematical theory required to rigorously justify the use of such algorithms and thereby ensure the integrity of the decision tools they produce. This mathematical framework is also expected to produce new algorithms for optimisation which benefit consumers of data science such as the health-care and cybersecurity sectors.Read moreRead less
Switching Dynamics Approach for Distributed Global Optimisation . This project aims to create a breakthrough switching dynamics approach and new technology to speed up finding optimal solutions. It will develop a distributed switching dynamics based optimisation scheme for global optimisation problems in industrial big-data environments where timely decision making is required. It will result in a practical technology for industry optimisation problems such as economic energy dispatch in smart g ....Switching Dynamics Approach for Distributed Global Optimisation . This project aims to create a breakthrough switching dynamics approach and new technology to speed up finding optimal solutions. It will develop a distributed switching dynamics based optimisation scheme for global optimisation problems in industrial big-data environments where timely decision making is required. It will result in a practical technology for industry optimisation problems such as economic energy dispatch in smart grids and optimal charging and discharging tasks in a large network of electric vehicles, helping Australian power industry improve efficiency and security, as well as training the next generation scientists and engineers for Australia in this emerging field.Read moreRead less
Generating Knowledge from High-dimensional and Incrementally Acquired Data. Complex data from emergencies, e.g., data acquired from an ongoing viral outbreak or actively moving bush fire are often received progressively. The analysis of such situations cannot wait until the complete data set is available at the end of the emergency. The aim of this project is to overcome this serious deficiency of current AI tools by developing innovative Neural Network based methods that can learn from continu ....Generating Knowledge from High-dimensional and Incrementally Acquired Data. Complex data from emergencies, e.g., data acquired from an ongoing viral outbreak or actively moving bush fire are often received progressively. The analysis of such situations cannot wait until the complete data set is available at the end of the emergency. The aim of this project is to overcome this serious deficiency of current AI tools by developing innovative Neural Network based methods that can learn from continuous data streams and extract and interpret the hidden knowledge either semantically or mathematically. The expected outcomes of this project include the development of novel methods, highly trained AI researchers and a number of critical real applications that will bring significant benefits to Australia and the world.Read moreRead less