[IJCAI 2022] Conference Review and Deep Graph Learning Research Trends - LG AI Research BLOG

IJCAI 2022, held in Vienna, Austria, is a prestigious AI conference known for its high paper acceptance rate, and it was conducted in a hybrid online/offline format. This conference covered a wide range of topics, from deep learning methodologies to robotics and systems research.

Looking at the major research trends, papers related to Reinforcement Learning were overwhelmingly numerous. Various sessions including Agent/Multi-Agent Systems, Planning, and Safe RL were held, and research combined with robotics was also actively introduced.

In addition, Federated Learning and Reasoning research stood out. In particular, in the NLP field, research on reasoning based on knowledge graphs was prominent, and the importance of spatio-temporal data was once again emphasized. Unlike in the past, instead of the keyword Big Data, technology for gaining insights from scarce data emerged as the key focus.

In the Invited Talk, Professor Judea Pearl emphasized the necessity of Causal Inference. To solve the interpretability problem of existing black-box models, a direction was proposed to visibly encode causal relationships in order to solve prediction, interpretation, and counterfactual problems, and to increase the efficiency of data integration.