Deep Learning 深度学习知识地图
Personal notes for neural-network foundations, probability, information theory, optimisation, PyTorch practice, convolutional architecture, and formula review. 这里按知识点整理神经网络基础、概率、信息论、优化、PyTorch 练习、卷积网络结构和公式汇总，不按任何真实课程资料编号整理。
Neural Network Foundations神经网络基础
Perceptrons, decision boundaries, logical functions, XOR, hidden units, and why nonlinear layers matter.感知机、分界线、逻辑函数、XOR、隐藏层，以及为什么非线性激活函数重要。
Probability & Information概率与信息量
Bayes' rule, entropy, KL divergence, Gaussian uncertainty, and Wasserstein distance.贝叶斯公式、熵、KL 散度、高斯分布的不确定性和 Wasserstein 距离。
Formula Summary公式汇总
A compact memory page for the formulas used across these notes.把这些笔记里反复出现的公式集中放在一页，方便复习和背熟。
Theory Quiz理论练习
40 original practice questions for terminology, formula use, and design judgement.40 道原创练习题，覆盖术语、公式使用和概念判断。
Focused Topics专题笔记
Optimisation & Backpropagation优化与反向传播
Gradient descent, learning rate behaviour, momentum, forward pass, loss, gradients, and parameter updates.梯度下降、学习率、动量、前向传播、loss、梯度和参数更新。
PyTorch & XOR PracticePyTorch 与 XOR 练习
Tensor basics, small neural networks, local minima, initialisation, and why repeated runs can behave differently.Tensor 基础、小神经网络、局部最小值、初始化，以及为什么多次运行结果不同。
Convolutional Architecture卷积网络结构
Filter shapes, stride, layer size, parameter counting, weight initialisation, batch normalisation, and skip connections.卷积核、stride、层尺寸、参数计数、权重初始化、批归一化和跳连结构。
How to Study This Area建议复习路线
Start with perceptrons and XOR so the idea of a decision boundary is clear. Then review probability and information theory, because loss functions and classification probabilities depend on them. After that, study optimisation and backpropagation, then PyTorch practice, then convolutional architecture. Use the formula summary as the memory sheet and the quiz as the final check.先从感知机和 XOR 开始，把分界线和隐藏层的意义搞清楚；再看概率和信息论，因为分类概率和 loss 会用到这些概念；然后看优化和反向传播，再到 PyTorch 练习，最后看卷积网络结构。公式汇总用来背熟，quiz 用来检查是否真的会用。