Transfer Entropy

Transfer entropy (TE) quantifies the directional flow of information between time-varying systems, aiming to identify causal relationships and dependencies. Current research focuses on applying TE within machine learning models, such as convolutional neural networks and transformers, to improve model performance, feature selection, and causal discovery in diverse applications. This involves developing efficient TE estimation methods and integrating TE calculations into existing algorithms for tasks ranging from classifying eSports player skill levels to root cause analysis in industrial systems. The broader impact of this work lies in enhancing our understanding of complex systems and improving the accuracy and efficiency of various machine learning techniques.