知识卡片
学习率影响迭代而非表达边界
学习率控制一次参数修正的大小,需与数据尺度、优化器和停止预算一起选择。原文“小就永远找不到、大就必跳过可分解”不是感知机的一般定理;可分性与所用算法的收敛条件应分开判断。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 人工神经网络
原书出处 · 1 处
- · 查看原书依据
Although it is very simple, this learning algorithm has still a lot in common with state-of-the-art algorithms. First, weights are updated in an iterative process using small increments governed by the parameter r, which is referred to as the learning rate. By changing w in small increments, the algorithm is literally rotating and translating the separating line in a direction that minimizes the loss, given
参考来源
- 位置:《Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)》第10章《人工神经网络》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)/_epub-src/OEBPS/xhtml/chapter_10.xhtml)
- 结论依据:摘录中关于学习率影响迭代而非表达边界的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:Although it is very simple, this learning algorithm has still a lot in common with state-of-the-art algorithms. First, weights are updated in an iterative process using small increments governed by the parameter r, which is referred to as the learning rate. By changing w in small increments, the algorithm is literally rotating and translating the separating line in a direction that minimizes the loss, given