Hyun woo Lee

Robust Weight Initialization for Tanh Neural Networks with Fixed Point Analysis

As a neural network's depth increases, it can improve generalization performance. However, training deep networks is challenging due to gradient and signal propagation issues. To address these challenges, extensive theoretical research and various methods have been introduced. Despite these advances…