Topological Feature
Topological features, derived from topological data analysis (TDA), capture the shape and connectivity of data, offering insights beyond traditional feature extraction methods. Current research focuses on integrating TDA with machine learning models, such as graph neural networks and transformers, to improve the robustness and interpretability of various applications, including vulnerability detection, anomaly detection in deep learning, and protein structure prediction. This interdisciplinary approach is proving valuable across diverse fields, enhancing model performance and providing new avenues for understanding complex data structures.
Papers
Current Topological and Machine Learning Applications for Bias Detection in Text
Colleen Farrelly, Yashbir Singh, Quincy A. Hathaway, Gunnar Carlsson, Ashok Choudhary, Rahul Paul, Gianfranco Doretto, Yassine Himeur, Shadi Atalls, Wathiq Mansoor
Detecting out-of-distribution text using topological features of transformer-based language models
Andres Pollano, Anupam Chaudhuri, Anj Simmons