Jie Lu

Australian Artificial Intelligence Institute,

University of Technology Sydney, NSW, 2007, Australia  

Email：  Jie.Lu@uts.edu.au

Abstract: The research will present how machine learning can innovatively and effectively learn from data to support data-driven decision-making in uncertain and dynamic situations. A set of new fuzzy transfer learning theories, methodologies and algo-rithms will be presented that can transfer knowledge learnt in one or more source domains to target domains by building latent space, mapping functions and self-training to overcome tremendous uncertainties in data, learning processes and decision outputs (classification and regression). Another set of concept drift the-ories, methodologies and algorithms will be discussed about how to handle ever-changing dynamic data stream environments with unpredictable stream pattern drifts by effectively and accurately detecting, understanding, and adapting con-cept drift in an explanatory way, indicating when, where and how concept drift occurs and reacting accordingly. These new developments enable advanced ma-chine learning and therefore enhance data-driven prediction and decision sup-port systems in uncertain and dynamic real-world environments.

2. Making Machine Learning Fairer

Ho-fung Leung 

Department of Computer Science and Engineering and Department of Sociology,

The Chinese University of Hong Kong

Email: lhf@cuhk.edu.hk

Abstract:  In this talk we shall present our recent works on multi-agent reinforcement learning. In multi-agent reinforcement learning, agents interact with one another in a multi-agent system. They continuously revise their decision policies by learning from their experiences of interacting with other agents. Generally, a social norm of action will emerge at some point. I shall describe our research results in multi-agent reinforcement learning, and discuss what we can learn from these results. I shall also highlight some theoretical results on mathematical modelling of multi-agent reinforcement learning.

Bao-Liang Lu

Center for Brain-Like Computing and Machine Intelligence

Department of Computing Science and Engineering

Shanghai Jiao Tong University, Shanghai 200240, China

bllu@sjtu.edu.cn

Abstract:  Brain-computer interface (aBCI) is a type of brain-computer interface that can recognize and/or regulate emotions. In particular, according to whether to regulate emotions, aBCI can be divided into two categories. The first category is emotion recognition BCI, which can recognize emotions based on the brain signals collected by external devices. The second category is emotion recognition and regulation BCI, which can not only recognize emotions but also regulate emotions by stimulating specific brain areas. Currently, most research is focused on emotion recognition BCI. The study on emotion regulation BCI is highly limited. This talk will introduce our recent work on emotion recognition BCI and applications. Specifically, we will introduce a multimodal affective BCI framework of combining EEG signals and eye movement signals, a plug-and-play domain adaptation for cross-subject EEG-based Emotion Recognition, GAN-based methods for EEG data augmentation and the practical application of emotion recognition BCI to depression evaluation.

Dongrui Wu

School of Artificial Intelligence and Automation,

Huazhong University of Science and Technology, China

Email: drwu@hust.edu.cn

Abstract: Brain-computer interface (BCI) is a direct communication pathway between the brain and an external device. Because of individual differences and non-stationarity of brain signals, a BCI usually needs subject-specific calibration, which is time-consuming and user unfriendly. Sophisticated machine learning approaches can help reduce or even completely eliminate calibrations, improving the utility of BCIs. Recent studies also found that machine learning models in BCIs are vulnerable to adversarial attacks, and brain signals also contain lots of private information, so the security and privacy of BCIs are also important considerations in their commercial applications. This talk will introduce transfer learning approaches for expedite BCI calibration, and their adversarial attack and privacy protection approaches. The ultimate goal is to implement accurate, secure and privacy-preserving BCIs.

Abstract: While reinforcement learning (RL) has shown super-human decision-making ability in playing games, RL was extremely difficult to come to the real world. An apparent cause of the difficulty is the missing of a cost-free playing ground, i.e., an environment model, for RL. However, learning an effective environment model from data is inhibited by high compounding error, which was theoretically proved as the simulation lemma found in 2002. Nearly 20 years later, we now have established a new theory with compounding error eliminated, which allows us to learn effective environment models. We show that environment model learning enables truly offline RL, which makes zero trial errors to train a policy. We also show that such offline RL can be applied in a wide range of real-world tasks.

Biography： Yang Yu is a Professor of Artificial Intelligence School, Nanjing University. His research interesting is in reinforcement learning. He received CCF-IEEE Early Career Scientist Award in 2020, was recognized as an AI's 10 to Watch by IEEE Intelligent Systems in 2018, received Early Career Award by PAKDD in 2018, and invited to give an Early Career Spotlight talk in IJCAI’18.

Abstract: In recent years, researchers began to study how to apply neural network to graph data, forming a research boom of graph neural network (GNN). Most GNNs are equivalent to a low-pass filter, which aggregates the feature information of neighbor nodes along the network structure, and realizes the effective fusion of network structure and attribute features. This report will rethink some key operations of GNNs, such as message aggregation mechanism, reliable graph structure, and the role of low-pass filtering, and report the recent progress of GNNs. 

Biography: Chuan Shi is the professor in Beijing University of Posts and Telecommunications, deputy director of Beijing Key Lab of Intelligent Telecommunication Software and Multimedia. The main research interests include data mining, machine learning, artificial intelligence and big data analysis. He has published more than 100 refereed papers, including top journals and conferences in data mining and machine learning, such as IEEE TKDE, ACM TKDD, KDD, WWW, NeurIPS, AAAI and IJCAI. He has been honored as the best paper award in ADMA 2011/ADMA 2018 and the best paper nomination in the WebConf 2021. He has won several awards, such as the second prize of Natural Science of Beijing/CCF (1st) and the first prize of artificial intelligence science and technology progress of Wu Wenjun (3rd).