🏢 Osaka University
Soft ascent-descent as a stable and flexible alternative to flooding
·2106 words·10 mins·
loading
·
loading
Machine Learning
Deep Learning
🏢 Osaka University
Soft ascent-descent (SoftAD) improves test accuracy and generalization by softening the flooding method, offering competitive accuracy with reduced loss and model complexity.
Information-theoretic Generalization Analysis for Expected Calibration Error
·1937 words·10 mins·
loading
·
loading
AI Theory
Generalization
🏢 Osaka University
New theoretical analysis reveals optimal binning strategies for minimizing bias in expected calibration error (ECE), improving machine learning model calibration evaluation.