← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 9174.e · 生成模型:GPT-6

循环状态递推压缩过去信息

理解等级:中级

基本RNN用ht=tanh(W h{t−1}+U x_t)递推隐藏状态,再由输出层生成预测。历史通过状态进入下一步,初始状态和重置边界也属于模型输入契约;输出Softmax只适合相应分类任务,不是所有RNN必须采用。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      To overcome this limitation, it is useful to introduce a notion of state in a neural network. In this case, the detection of an event such as “getting stuck” may be used to modify the network state in some way. This is accomplished using so-called recurrent neural networks. A recurrent neural network uses a special kind of neuron that sums the input xt at time

    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:摘录中关于循环状态递推压缩过去信息的定义、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:To overcome this limitation, it is useful to introduce a notion of state in a neural network. In this case, the detection of an event such as “getting stuck” may be used to modify the network state in some way. This is accomplished using so-called recurrent neural networks. A recurrent neural network uses a special kind of neuron that sums the input xt at time