Is Higher Singular Value Entropy Always Better for Matrix Parameters?
科学空间3839 字 (约 16 分钟)
92
Higher singular value entropy is not always better; via geometric modeling and mean-field approximation, the optimal entropy is found to be approximately log(n) - 1 (where n is matrix dimension), corresponding to an effective rank of ~e·n, balancing expressiveness and redundancy.
入选理由：奇异值熵最大值为 log(n)，但最优值约为 log(n) - 1，对应有效秩 ≈ e·n（e≈2.718）
FeaturedArticle#Singular Value Entropy#Effective Rank#Matrix Decomposition#Information Theory#Deep Learning Optimization中文