Homogeneity Bias
Homogeneity bias, the tendency for systems to represent certain groups more uniformly than others, is a significant concern across various fields, from AI and machine learning to image analysis and even social sciences. Current research focuses on detecting and mitigating this bias through improved model architectures (e.g., incorporating attention mechanisms, multiplex networks, or latent factor models) and algorithmic approaches (e.g., developing robust unmixing frameworks, and employing novel regularization techniques). Understanding and addressing homogeneity bias is crucial for ensuring fairness and accuracy in AI systems, improving the reliability of scientific analyses, and promoting equitable outcomes in diverse applications.
Papers
SCALE: Self-regulated Clustered federAted LEarning in a Homogeneous Environment
Sai Puppala, Ismail Hossain, Md Jahangir Alam, Sajedul Talukder, Zahidur Talukder, Syed Bahauddin
Mew: Multiplexed Immunofluorescence Image Analysis through an Efficient Multiplex Network
Sukwon Yun, Jie Peng, Alexandro E. Trevino, Chanyoung Park, Tianlong Chen