Research paper analyzes noise sensitivity in hierarchical functions

A new research paper explores the learning complexity of functions with hierarchical structures, particularly in the context of deep learning. The study demonstrates that functions with tree-like hierarchical structures exhibit exponentially small noise stability in relation to their depth, especially when deviating from linearity. These findings have implications for agnostic learning, providing super-polynomial lower bounds for learning hierarchical functions in both Boolean and Gaussian settings. AI

IMPACT Provides theoretical insights into the learning complexity of hierarchical functions, potentially influencing future model architectures.

RANK_REASON The cluster contains a research paper published on arXiv detailing theoretical findings in machine learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Abbe
• Bengio
• Cornacchiam
• Dachman-Soled
• Diakonikolas
• Kane
• Kleinberg
• Lisa Feldman Barrett
• Lotfi
• Pittas
• Raghu
• Rupert Liebl
• Wimmer
• Zarifis
• Zhang

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →