📄 中文摘要
在前一篇文章中，已经将一些值传递给 RNN 并获得了输出。为了更好地理解随着输入数量的增加，如何处理这些数据，需要对神经网络进行展开。展开后的 RNN 结构清晰地展示了时间序列数据的处理过程，使得每个时间步的输入、隐藏状态和输出都能一目了然。这种展开的方式有助于理解 RNN 的工作原理，尤其是在处理长序列数据时，能够更直观地看到信息如何在时间维度上流动和更新。
理解 RNN – 第三部分：展开递归神经网络
出处: Understanding RNNs – Part 3: Unrolling a Recurrent Neural Network
发布: 2026年2月18日
📄 English Summary
Understanding RNNs – Part 3: Unrolling a Recurrent Neural Network
In the previous article, some values were passed to the RNN, and the output was obtained. To better understand how to handle increasing amounts of input data, it is necessary to unroll the neural network. The unrolled structure of the RNN clearly illustrates the processing of sequential data, making each time step's input, hidden state, and output easily observable. This unrolling approach aids in understanding the workings of RNNs, especially when dealing with long sequences, as it allows for a more intuitive view of how information flows and updates over time.