Toward Human-guided Safe Reinforcement Learning in the Real World. This project aims to investigate human-guided safe reinforcement learning (RL). Safe RL is an important topic that could enable real applications of RL systems by addressing safety constraints. Existing safe RL assumes the availability of specified safety constraints in mathematical or logical forms. This project proposes to study learning safety objectives from information provided directly by humans or indirectly via language m ....Toward Human-guided Safe Reinforcement Learning in the Real World. This project aims to investigate human-guided safe reinforcement learning (RL). Safe RL is an important topic that could enable real applications of RL systems by addressing safety constraints. Existing safe RL assumes the availability of specified safety constraints in mathematical or logical forms. This project proposes to study learning safety objectives from information provided directly by humans or indirectly via language models, and human-guided continuous correction for safety improvements. The established theories and developed algorithms will advance frontier technologies in AI and contribute to a wide range of real applications of safe RL, such as robotics and autonomous driving, bringing enormous social and economic benefits. Read moreRead less
Generative Visual Pre-training on Unlabelled Big Data. This project aims to develop a generative visual pre-training of large-scale deep neural networks on unlabelled big data. Developing pre-trained visual models that are accurate, robust, and efficient for downstream tasks is a keystone of modern computer vision, but it poses challenges and knowledge gaps to existing unsupervised representation learning. Expected outcomes include new theories and algorithms for unsupervised visual pre-training ....Generative Visual Pre-training on Unlabelled Big Data. This project aims to develop a generative visual pre-training of large-scale deep neural networks on unlabelled big data. Developing pre-trained visual models that are accurate, robust, and efficient for downstream tasks is a keystone of modern computer vision, but it poses challenges and knowledge gaps to existing unsupervised representation learning. Expected outcomes include new theories and algorithms for unsupervised visual pre-training, which are anticipated to deepen our understanding of visual representation and make it easier to build and deploy computer vision applications and services. Examples of benefits include modernising machines in manufacturing and farming with visual intelligence. Read moreRead less
Data Complexity and Uncertainty-Resilient Deep Variational Learning. Enterprise data present increasingly significant characteristics and complexities, such as multi-aspect, heterogeneous and hierarchical features and interactions, and evolving dependencies and multi-distributions. They continue to significantly challenge the state-of-the-art probabilistic and neural learning systems with limited to insufficient capabilities and capacity. This research aims to develop a theory of flexible deep v ....Data Complexity and Uncertainty-Resilient Deep Variational Learning. Enterprise data present increasingly significant characteristics and complexities, such as multi-aspect, heterogeneous and hierarchical features and interactions, and evolving dependencies and multi-distributions. They continue to significantly challenge the state-of-the-art probabilistic and neural learning systems with limited to insufficient capabilities and capacity. This research aims to develop a theory of flexible deep variational learning transforming new deep probabilistic models with flexible variational neural mechanisms for analytically explainable, complexity-resilient analytics of real-life data. The outcomes are expected to fill important knowledge gaps and lift critical innovation competencies in wide domains.Read moreRead less
Exploiting Geometries of Learning for Fast, Adaptive and Robust AI. This project aims to uniquely exploit geometric manifolds in deep learning to advance the frontier of Artificial Intelligence (AI) research and applications in cybersecurity and general cognitive tasks. It expects to develop new theories, algorithms, tools, and technologies for machine learning systems that are fast, adaptive, lifelong and robust, even with limited supervision. Expected outcomes will enhance Australia's capabili ....Exploiting Geometries of Learning for Fast, Adaptive and Robust AI. This project aims to uniquely exploit geometric manifolds in deep learning to advance the frontier of Artificial Intelligence (AI) research and applications in cybersecurity and general cognitive tasks. It expects to develop new theories, algorithms, tools, and technologies for machine learning systems that are fast, adaptive, lifelong and robust, even with limited supervision. Expected outcomes will enhance Australia's capability and competitiveness in AI, and deliver robust and trustworthy learning technology. The project should provide significant benefits not only in advancing scientific and translational knowledge but also in accelerating AI innovations, safeguarding cyberspace, and reducing the burden on defence expenses in Australia.Read moreRead less
Learning to Reason in Reinforcement Learning. Deep Reinforcement Learning (RL) uses deep neural networks to represent and learn optimal decision-making policies for intelligent agents in complex environments. However, most RL approaches require millions of episodes to converge to good policies, making it difficult for RL to be applied in real-world scenarios taking significant resources. This project aims to equip RL with capabilities such as counterfactual reasoning and outcome anticipation to ....Learning to Reason in Reinforcement Learning. Deep Reinforcement Learning (RL) uses deep neural networks to represent and learn optimal decision-making policies for intelligent agents in complex environments. However, most RL approaches require millions of episodes to converge to good policies, making it difficult for RL to be applied in real-world scenarios taking significant resources. This project aims to equip RL with capabilities such as counterfactual reasoning and outcome anticipation to significantly reduce the number of interactions required, improve generalisation, and provide the agent with the capability to consider the cause-effects. These improvements would narrow the gap between AI and human capabilities and broaden the adoption of RL in real-world applications.Read moreRead less
Causal Knowledge-Empowered Adaptive Federated Learning. Federated learning tools are a promising framework for collaborative machine learning (ML) that also maintain data privacy; however, their ability to model heterogeneous data remains a key challenge. This project aims to develop a new learning scheme for coordinated training of ML models that successfully bridges variable data distributions. The framework proposed will be the first globally that can use causal knowledge to 1) handle data he ....Causal Knowledge-Empowered Adaptive Federated Learning. Federated learning tools are a promising framework for collaborative machine learning (ML) that also maintain data privacy; however, their ability to model heterogeneous data remains a key challenge. This project aims to develop a new learning scheme for coordinated training of ML models that successfully bridges variable data distributions. The framework proposed will be the first globally that can use causal knowledge to 1) handle data heterogeneity across devices and 2) address the real-world challenges when only a subset of devices have labelled data. Expected outcomes and benefits include the theoretical underpinnings and algorithms of causality-based collaborative training of ML models while better preserving the users’ data privacy.Read moreRead less
Tracking towards a complete model of skilled reading comprehension. This project aims to promote the development of the first complete computational model of reading comprehension. Many computational models of sub-components of reading have been developed, but none fully explain the complex co-ordination of perceptual, attentional and cognitive processes required for successful comprehension. The project intends to use eye tracking studies to test and refine Über-Reader, a new computational mode ....Tracking towards a complete model of skilled reading comprehension. This project aims to promote the development of the first complete computational model of reading comprehension. Many computational models of sub-components of reading have been developed, but none fully explain the complex co-ordination of perceptual, attentional and cognitive processes required for successful comprehension. The project intends to use eye tracking studies to test and refine Über-Reader, a new computational model that aims to provide a complete account of the memory systems and cognitive processes involved in reading comprehension and how they differ with reading skill. The outcomes will advance understanding of the causes of success and failure in reading and contribute to diagnosing and remediating reading difficulties.Read moreRead less
A Generic Framework for Verifying Machine Learning Algorithms. This project aims to discover new ways to verify whether decisions made by Artificial Intelligence and Machine Learning algorithms are as per the specifications set by their designers and/or regulatory bodies. The project also provides new methods to align algorithm decisions when they are found to be non-abiding. The outcomes will include new machine learning theories and frameworks for algorithmic assurance. The significance of the ....A Generic Framework for Verifying Machine Learning Algorithms. This project aims to discover new ways to verify whether decisions made by Artificial Intelligence and Machine Learning algorithms are as per the specifications set by their designers and/or regulatory bodies. The project also provides new methods to align algorithm decisions when they are found to be non-abiding. The outcomes will include new machine learning theories and frameworks for algorithmic assurance. The significance of the project is that it will offer a crucial platform for certifying algorithms and thus benefit society and businesses in deciding the right Artificial Intelligence algorithms. Read moreRead less
Making Meta-learning Generalised . This project aims to develop novel machine learning techniques, termed generalised meta-learning, to make machines better utilise past experience to solve new tasks with few data. It expects to reduce the undesirable dependence of current machine learning on labelled data and significantly expand its application scope. Expected outcomes of the project consist of new theoretical results on meta-learning and a set of innovative algorithms that can support the bui ....Making Meta-learning Generalised . This project aims to develop novel machine learning techniques, termed generalised meta-learning, to make machines better utilise past experience to solve new tasks with few data. It expects to reduce the undesirable dependence of current machine learning on labelled data and significantly expand its application scope. Expected outcomes of the project consist of new theoretical results on meta-learning and a set of innovative algorithms that can support the building of next generation of computer vision systems to work in open and dynamic environments. This should be able to produce solid benefits to the science, society, and economy of Australian via the application of these advanced intelligent systems.Read moreRead less
3D Vision Geometric Optimisation in Deep Learning. This project aims to develop a methodology for integrating the algorithms of 3D Vision Geometry and Optimization into the framework of Machine Learning and demonstrate the wide applicability of the new methods on a variety of challenging fundamental problems in Computer Vision. These include 3D geometric scene understanding, and estimation and prediction of human 2D/3D pose and activity. Applications of this technology are to be found in Intell ....3D Vision Geometric Optimisation in Deep Learning. This project aims to develop a methodology for integrating the algorithms of 3D Vision Geometry and Optimization into the framework of Machine Learning and demonstrate the wide applicability of the new methods on a variety of challenging fundamental problems in Computer Vision. These include 3D geometric scene understanding, and estimation and prediction of human 2D/3D pose and activity. Applications of this technology are to be found in Intelligent Transportation, Environment Monitoring, and Augmented Reality, applicable in smart-city planning and medical applications such as computer-enhanced surgery. The goal is to build Australia's competitive advantage in the forefront of ICT research and technology innovation.Read moreRead less