Scholarlink

Simply deepening neural networks leads to several challenges in learning. First and foremost, in the optimization computations of neural networks with many parameters, we cannot use efficient optimization methods utilizing second-order differentials such as Newton’s method.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Fundamental Technologies Supporting Deep Learning

  • 章节

  • 发表日期：

    2025年

查看全文
  • Takeshi Okadome

摘要

Simply deepening neural networks leads to several challenges in learning. First and foremost, in the optimization computations of neural networks with many parameters, we cannot use efficient optimization methods utilizing second-order differentials such as Newton’s method.