AI and machine learning
Developing AI-based solutions that enhance industry performance
Under the theme of AI and machine learning, we are developing theories that explain the generalisation ability of (deep) machine learning models that will enable research applicable to a range of different technologies and applications, particularly for industry engagement. Our team of researchers will:
- Find principled designs for AI and machine learning algorithms to enhance the trustworthiness of AI and machine learning techniques and tools
- Develop algorithms that deal with weakly supervised information, causally responsible representations, heterogeneous information fusion, visual plausible data generation, energy cost efficient computation, and model robustness and scalability in the wide
This research will have broader applications in eCommerce, health, cybersecurity, logistics and supply chain, and streamlined manufacturing.
Explore our research
We are also interested in how to reduce the side effect of noise on the instance, which may be caused by the failure of sensors or even malicious attacks. We human have the ability to correctly recognise the objects even there are noise (e.g., we can easily recognise human faces under extreme illumination conditions, when partially occluded, or even with heavy makeup); while current machine learning algorithms may not. Recent studies also show that an imperceptible noise on the instance will lead machines to make wrong decisions. All those mean that we human and machines are using different feature extraction mechanisms for making decisions. What are the differences? And how to align them? Answering those questions is very important to build robust and trustworthy machine learning algorithms.
Generative Adversarial Networks (GANs) were called as the most interesting idea in the last 10 years in machine learning by Turing award recipient Yann LeCun. Their most significant impact has been observed in many challenging problems, such as plausible image generation, image-to-image translation, facial attribute manipulation and similar domains. However, there are still many research questions around GANs. For example, the minimax optimisation problem underlying GANs increases the training difficulty, how can we precisely manipulate the content creation in GANs, and whether the synthetic data from GANs can be reliable for the use in the finance or medical domains.
Deep learning algorithms have given exciting performances, e.g., painting pictures, beating Go champions, and autonomously driving cars, among others, showing that they have very good generalisation abilities (small differences between training and test errors). These empirical achievements have astounded yet confounded their human creators. Why do deep learning algorithms generalise so well on unseen data? It lacks mathematical elegance. We do not know the underlying principles that guarantee its success. Let alone to interpret or pertinently strengthen its generalisation ability. In this project, we aim to analyse error bounds, e.g., generalisation error bound and excess risk bound, by measuring the complexity of the predefined (or algorithmic) hypothesis class. An algorithmic hypothesis class is a subset of the predefined hypothesis class that a learning algorithm will (or is likely to) output.
Learning with noisy labels becomes a more and more important topic recently. The reason is that, in the era of big data, datasets are becoming larger and larger. Often, large-scale datasets are infeasible to be annotated accurately due to the cost and time, which naturally brings us cheap datasets with noisy labels. However, the noisy dataset can severely degenerate the performance of machine learning models, especially for the deep neural networks, as they easily memorise and eventually fit label noise. In this project, we are interested to model the noise and then eliminate the side-effect of label noise, i.e., obtaining the optimal classifier defined by the clean data by exploiting the noisy data.
Since the development of the first real deep neural network AlexNet in 2012, deep learning has made great progress in computer vision and natural language processing. Lots of these breakthroughs often come alone with the new architecture design of deep neural networks. We are interested in pushing the boundary of deep learning performance by advancing the design of neural architectures, while taking the trade-off between the accuracy of the network and its computation cost into consideration.
Core research team
Tongliang Liu is currently a Lecturer and director of the Trustworthy Machine Learning Lab with School of Computer Science at the University of Sydney. He is broadly interested in the fields of trustworthy machine learning and its interdisciplinary applications, with a particular emphasis on learning with noisy labels, transfer learning, adversarial learning, unsupervised learning, and statistical deep learning theory. He has published papers on various top conferences and journals, such as NeurIPS, ICML, ICLR, CVPR, ECCV, KDD, IJCAI, AAAI, IEEE TPAMI, IEEE TNNLS, IEEE TIP, and IEEE TMM. He received the ICME 2019 best paper award. He is a recipient of Discovery Early Career Researcher Award (DECRA) from Australian Research Council (ARC); the Cardiovascular Initiative Catalyst Award by the Cardiovascular Initiative; and was named in the Early Achievers Leadboard of Engineering and Computer Science by The Australian in 2020.
Chang Xu is Senior Lecturer and ARC DECRA Fellow at the School of Computer Science, University of Sydney. He received the Ph.D. degree from Peking University, China. His research interests lie in machine learning algorithms and related applications in computer vision. He has published over 100 papers in prestigious journals and top tier conferences. He has received several paper awards, including Distinguished Paper Award in IJCAI 2018. He regularly severed as the Senior PC member or PC member for many conferences, e.g. NIPS, ICML, CVPR, ICCV, IJCAI and AAAI. He has been recognised as "Top Ten Distinguished Senior PC Member" in IJCAI 2017.