Neural Networks
Representations, computational graphs, learning
Look inside the learning machine.
From a single neuron to attention: read the idea, move the numbers, and see the computation change.
神经网络、卷积与序列模型 · 完整交互讲义 · 9 段教学视频 · 中英字幕
Complete study course for the three supplied 2025 lectures (09–11): 9 full video parts, derivations, experiments and worked exercises. The new semester's slides are not available yet; source-page references make later updates straightforward.
Convolutional Neural Networks
Local connections, shared weights, spatial structure
RNN and Transformer
State, sequence order, attention
A continuous learning path
L09: learn a representation and differentiate a computation graph.
L10: share local detectors across an image.
L11: carry state through time, then connect positions with attention.
Use it as a study lab
Predict an answer before revealing it. Step through an example. Change one variable at a time. Finish with a short quiz and worked exercises.
Each lecture contains three full narrated parts, nine worked exercises, executable labs and a Chinese terminology guide. Source-page references and factual qualifications appear beside the relevant explanation.