News
At the invitation of Professor Chrysostomos Stylios, the Director of the Institute of Industrial Systems Research under the Innovation Research Department of the Hellenic Ministry of Development and Investment, Professor X. San Liang visited the Industrial Systems Research Institute located at the Athena Research Center, Patras, Greece. During the visit, Professor Liang delivered a presentation entitled "Quantitative Causality Analysis, Causality-Aided Learning, and Causal AI-Based Prediction."
In his presentation, Professor Liang firstly highlighted the limitations, particularly, the lack of causal inference in the current machine learning algorithms, that may hinder the progress in the field of artificial intelligence. He then shared with the audience his advanced causality analysis tool, the "Liang-Kleeman Information Flow" analysis method, which has been rigorously derived and well validated in all kinds of real world applications across scientific disciplines. With its quantitative nature and computational efficiency, it proves to be promising for the development of the next generation of artificial intelligence algorithms. Professor Liang's team has already employed the information flow-based causality analysis to develop a causal artificial intelligence forecasting model, which testified to success at the 8th National Weather Forecast Contest (June-October, 2022).
During the visit, Professor Stylios and his team member, Marios Tyrovolas, presented significant progress in their development of causal artificial intelligence algorithms based on the Liang Information Flow: one notable limitation of traditional Fuzzy Cognitive Maps (FCMs) is their susceptibility to unknowingly incorporating spurious correlations from data, leading to reduced predictive capacity and interpretability, while the introduction of Liang's information flow and causality analysis as constraints in FCM learning process effectively mitigates these issues by incorporating actual causal relationships, thereby enhancing the overall predictive and explanatory capabilities of the model. Both the host and guest are anticipating further exchanges and collaborations in the near future.